EN中文

Journal of Geographical Sciences >

2018 , Vol. 28 >Issue 10: 1538 - 1559

DOI: https://doi.org/10.1007/s11442-018-1560-3

Orginal Article

Changes in land use/cover mapped over 80 years in the Highlands of Northern Ethiopia

  • ETEFA Guyassa , 1, 2 ,
  • Amaury FRANKL 1, 3 ,
  • Sil LANCKRIET 1 ,
  • BIADGILGN Demissie 1, 4 ,
  • GEBREYOHANNES Zenebe 5 ,
  • AMANUEL Zenebe 2 ,
  • Jean POESEN 6 ,
  • Jan NYSSEN 1
Expand
  • 1. Department of Geography, Ghent University, Belgium
  • 2. Department of Land Resources Management and Environmental Protection, Mekelle University, Ethiopia
  • 3. Research Foundation Flanders (FWO), Brussels, Belgium
  • 4. Department of Geography and Environmental Studies, Mekelle University, Ethiopia
  • 5. Institute of Geo-information and Earth Observation Sciences, Mekelle University, Ethiopia
  • 6. Division of Geography, Department of Earth and Environmental Sciences, KU Leuven, Belgium;

Author: Etefa Guyassa, PhD, specialized in physical geography. E-mail:

Received date: 2017-04-12

  Accepted date: 2017-09-13

  Online published: 2018-10-25

Copyright

Journal of Geographical Sciences, All Rights Reserved
Fold

Abstract

Despite many studies on land degradation in the Highlands of Northern Ethiopia, quantitative information regarding long-term changes in land use/cover (LUC) is rare. Hence, this study aims to investigate the LUC changes in the Geba catchment (5142 km2), Northern Ethiopia, over 80 years (1935-2014). Aerial photographs (APs) of the 1930s and Google Earth (GE) images (2014) were used. The point-count technique was utilized by overlaying a grid on APs and GE images. The occurrence of cropland, forest, grassland, shrubland, bare land, built-up areas and water body was counted to compute their fractions. A multivariate adaptive regression spline was applied to identify the explanatory factors of LUC and to create fractional maps of LUC. The results indicate significant changes of most types, except for forest and cropland. In the 1930s, shrubland (48%) was dominant, followed by cropland (39%). The fraction of cropland in 2014 (42%) remained approximately the same as in the 1930s, while shrubland significantly dropped to 37%. Forests shrank further from a meagre 6.3% in the 1930s to 2.3% in 2014. High overall accuracies (93% and 83%) and strong Kappa coefficients (89% and 72%) for point counts and fractional maps respectively indicate the validity of the techniques used for LUC mapping.

Key words： fractional map; Google Earth; land use/cover; multivariate adaptive regression; Italian aerial photographs

Cite this article

ETEFA Guyassa , Amaury FRANKL , Sil LANCKRIET , BIADGILGN Demissie , GEBREYOHANNES Zenebe , AMANUEL Zenebe , Jean POESEN , Jan NYSSEN . Changes in land use/cover mapped over 80 years in the Highlands of Northern Ethiopia[J]. Journal of Geographical Sciences, 2018 , 28(10) : 1538 -1559 . DOI: 10.1007/s11442-018-1560-3

1 Introduction

Land use/cover (LUC) has transformed (on different spatial scales) across the world (Lambin et al., 2003; Lepers et al., 2005; Maitima et al., 2009; FAO, 2010), including Ethiopia (Alemayehu et al., 2009; Asmamaw et al., 2011; Mengistu et al., 2012; Meire et al., 2013; de Muelenaere et al., 2014). At global and regional scales, an increase of the cropland area has been a common phenomenon at the expense of the forest and shrubland area during the last few centuries (Ramankutty and Foley, 1999; Goldewijk, 2001; Goldewijk and Ramankutty, 2004). In Northern Ethiopia, severe land degradation occurred until the large-scale implementation of soil and water conservation (SWC) measures, that started in circa 1991 (Nyssen et al., 2004; Frankl et al., 2011; Hurni and Wiesmann, 2010). Forests which once covered large country areas vanished, although there are no reliable data on the original forest cover, about which many speculations exist (Pankhurst, 1995; Woien, 1995).
The explanation of the causes of land use/cover change extends from simple to multiple factors and their complex interactions occur on different spatial and temporal scales (Lambin et al., 2001). The rapid population growth, changing economic conditions and institutional factors, local and global uses are identified as the main factors for the transformation of the earth’s surface on a global scale (Turner et al., 1994; Lambin et al., 2003; Lambin et al., 2001). The agriculture-based economy (Nyssen et al., 2004), low income (Kidane et al., 2010; FAO, 2011), rapid population growth (CSA, 2008) and recurrent drought occurrences are assumed to be the prominent drivers for land cover changes in the Tigray region, Northern Ethiopia.
Despite the presence of ancient human settlements and agricultural practices in Northern Ethiopia (Bard et al., 2000; Nyssen et al., 2004), the long-term evolution of change in LUC has received a little attention in the region, particularly on a local scale. Most often, surface processes (LUC change, gully erosion, surface runoff and implementation of soil and water conservation) have occurred at large spatial and temporal scales in Northern Ethiopia (Hurni et al., 2005). Although some studies on LUC change are available in the region (e.g. Munro et al., 2008; Alemayehu et al., 2009; Teka et al., 2013; de Muelenaere et al., 2014), they all cover the period from 1965 to 2014 and generally focus on small catchments. Nyssen et al. (2009) and Meire et al. (2013) have compared the land use type and environmental changes in the region by using repeated historical terrestrial photographs over 140 years (from 1868 to 2008). Despite the wide application of terrestrial photographs for the study of LUC changes, the inaccessibility of remote sites constrains the spatial representation of terrestrial photographs as compared to aerial photography and satellite images.
Earlier reports on land cover, e.g. vegetation cover, are based on subjective descriptions and cause inconclusive reports on the original cover and deforestation rate in Ethiopia (Woien, 1995). In this regard, the shortage of historical datasets on LUC is claimed as a problem to measure the long-term environmental changes. Although the 1930s Aerial Photographs (APs), which are the oldest aerial photographs of the region taken during the Italian occupation of Ethiopia (1935-1941) (Nyssen et al., 2016), are available, they have rarely been used to compare the historical LUC with the current situation. An exception to that is the local study by Frankl et al. (2015). These Italian aerial photographs are important historical datasets for spatio-temporal analyses of land use, cover and management (Nyssen et al., 2016). Aerial photographs taken for military purposes during the World War II have been widely used for civilian purposes, such as land use change studies (Zeimetz et al., 1976).
Long-term data are required for reliable LUC change assessments and quantifications (Lambin et al., 2003). A good knowledge of historic land use trends with adequate coverage is important so as to understand the current and future developments (Goldewijk, 2001). A better understanding of the distribution of LUC allows reliable predictions and improves the scenario of future environmental changes (Lambin et al., 2003). Large-scale LUC changes can have substantial biophysical consequences and influence the water balance of a catchment, which in turn determines the economic development activities (e.g. dam construction, ground water use) in the area. Moreover, the map of LUC on the scale of the Geba catchment (ca. 5142 km2), Northern Ethiopia, particularly in the 1930s, is very important for the analysis of environmental changes over a long period. The Geba catchment is characterized by diversified biophysical elements. Hence, the study area was selected due to its representativeness for the Northern Ethiopian Highlands in terms of its biophysical characteristics (i.e. climate, geology, soils, topography and LUC). Therefore, this study aims: 1) to quantify the long-term LUC changes of the Geba catchment (5142 km2), Northern Ethiopia, by comparing the areal fractions of different LUC classes on the oldest aerial photographs (1935-1936) to the Google Earth images of 2014; 2) to analyze the major explanatory factors, which determine the LUC occurrence in the two periods, and 3) to create fractional maps of the major LUC of the Geba catchment in the 1930s and in 2014.

2 Materials and methods

2.1 Study area

The study area, the Geba catchment (Figure 1), is located in the Tigray region, Northern Ethiopian Highlands, between 13°16' to 14°15'N and 38°38' to 39°48'E. The topography of the catchment was obtained from the DEM-SRTM data with a 30 m × 30 m resolution (USGS, 2016). It covers an area of 5142 km2 with elevations ranging from about 900 m a.s.l. at its outlet in the southwest to 3300 m a.s.l. near Adigrat in the north. The Geba River is one of the major tributaries of the Tekeze River which finally joins the Atbara River and the Nile in Sudan (Figure 1).
Figure 1 Oro-hydrography of the Geba catchment and flight lines of Italian aerial photographs
The Geba catchment is characterized by a diversified geology: Precambrian basement, Palaeozoic and Mesozoic sedimentary rocks, Tertiary volcanic rocks and Quaternary deposits (Tesfamichael et al., 2010). A larger part of the catchment (55%) is classified as sloping (6%-15%) and over 30% is steep (>15%) while most of the flat areas are found around Mekelle, Wukro, Sinkata and Atsbi accounting for only about 15% (Tielens, 2012). Lithologic and topographic relations are very important factors for explaining the soil type distribution (Van de Wauw et al., 2008; Tielens, 2012). The dominant soils of the catchment include Leptosols, Cambisols and Regosols, with the most developed and deepest soil types on gentle slopes (Vertisols, Luvisols, Cambisols) (Nyssen et al., 2008a; Tielens, 2012). Soil erosion, landslide and deposition processes in the catchment result in the formation of young soils particularly on foot slopes (HTS, 1976; Nyssen et al., 2008a; Van de Wauw et al., 2008). On steep slopes and plateaus, shallow soils (like Leptosols) and bare rock are found, while more fine-textured soils with alluvial, stagnic or vertic properties are found in the valley bottoms (Tielens, 2012).
The Geba catchment has a tropical and semi-arid climate with a bi-annual rainfall distribution (“belg” = small rain and “kremt” = main rain season). The main rainy season, which accounts for circa 80% of the annual precipitation is short, restricted to mid-July - early September, but is characterized by intense showers and large drop sizes (Nyssen et al., 2005; Virgo and Munro, 1978). The rainfall is highly variable with an annual depth of 555-1200 mm. It is mainly determined by topographical factors such as slope aspect, aspect of the valley and slope gradient over longer distances (Nyssen et al., 2005), but lacks a significant correlation to elevation (Amanuel, 2009; Gebresamuel et al., 2010; Nyssen et al., 2005; Taye et al., 2013; Vanmaercke et al., 2014; Virgo and Munro, 1978). The annual evapotranspiration depth exceeds the precipitation depth, except during the rainy season, ranging from 905 to 2538 mm (Hadush, 2012). The mean annual maximum average air temperature ranges from 21 to 31ºC and the mean annual minimum air temperature ranges from 3 to 16ºC (Araya et al., 2010) whereas the monthly average varies between 12 and 19ºC.
The economy of the rural communities (85% of population) depends on agriculture, hence cropland dominates land use followed by bushland and bare land (Amanuel, 2009; Gebresamuel et al., 2010; Taye et al., 2013). Rangeland is mainly found on steep slopes and severely degraded due to overgrazing. The study area lost its native forests a long time ago (Aerts et al., 2016; Gebru et al., 2009; Nyssen et al., 2004) and the remnant patches of forests are usually limited to inaccessible areas and around churches (HTS, 1976; Munro et al., 2008; Gebru et al., 2009; Aerts et al., 2016). Over the last century, the Tigray region (including the study area) has faced various man-made and natural problems at different times. The war (by the Italian) in the 1930s in which chemical weapons were used, famines and population displacement due to the drought and civil war (1974-1991) (Lanckriet et al., 2015) and the peak land degradation from the 1930s to the second half of the 20th century (Nyssen et al., 2015) were among the major shocks which took place in the region. Over the last few decades, massive soil and water conservation activities including afforestation had been organized in Northern Ethiopia in order to reverse the environmental degradation (Nyssen et al., 2004; Descheemaeker et al., 2006a; Vancampenhout et al., 2006; Munro et al., 2008; Nyssen et al., 2008b; Frankl et al., 2011, 2013a; Taye et al., 2013).

2.2 Land use/cover analysis by the point counting method

Despite its obvious importance for the study of land use, the 1930s photography, which represent the oldest APs of Ethiopia remain underutilized in Ethiopia (Frankl et al., 2015; Nyssen et al., 2016), mainly due to their recent rediscovery. The Italian Military Geographical Institute (IGM) took these black and white APs with a scale ranging from 1:11,500 (nearly vertical photograph) to 1:13,000-1:18,000 (oblique photographs) during the Italian occupation of Ethiopia (1930s-1940s) and the sets were recovered by Nyssen et al. (2016). These APs cover large areas of Northern Ethiopia, including parts of the Geba catchment except the eastern and northern parts of the catchment.
Due to the obliqueness of these photographs and small areas covered by the stereo pairs of the vertical photographs, creating orthophotographs from these photographs using the conventional photogrammetry method is difficult to apply on large regions (Frankl et al., 2015). Alternatively, the point counting approach allows to extract information easily and to estimate the fractions of LUC from the APs as applied by e.g. Bellhouse (1981) and Daniels et al. (1968). In this technique, different land covers that occur under the points of a grid superimposed on an area are being identified and counted (Bellhouse, 1981). This technique has been chosen for resource assessment and to examine land cover and use changes (Saebo, 1983; Shockey, 1969; Zeimetz et al., 1976). Hence, we applied this point counting technique to Italian aerial photographs to assess the fraction of LUC in 1935-36, while the LUC data in 2014 were obtained from Google Earth imageries captured in the same season (December to February) as the APs. The Google Earth images of the study area for the period from December 2013 to February 2014 are based on Pleiades 1A and Pleiades 1B satellites (2 m resolution) and SPOT 6 satellite (1.5 m and 6 m resolution).
To assess LUC, a point grid with 270 points (18×15 points) with a spacing of circa 130 m between the points (Figure 2) was superimposed on the scanned vertical and one oblique AP and GE whose area ranges from 2.39 km2 to 12.36 km2 with an average area of 4.66 (±1.72) km2. The size and location of GE is approximately the same as the APs. In total, 134 aerial photographs, which cover about 12% of the study area, taken along 15 flight lines fairly distributed over the study area were used (Figure 1). To cover the northern and eastern part of the study area where APs do not exist (Figure 1), five additional locations were selected where data were available for LUC in 1965 for Dergajen, Hadnet and Sinkata (Teka et al.,2015) and for upper Agulae watershed (Alemayehu et al., 2009) and known (permanently forested) places in Dessa’a throughout the 20th century. We assumed that the 1965 interpretation was the best (though not perfect) supplement for the missing Italian APs of that area, while the error of using the 1965’s information for the 1930s LUC condition is assumed to be small. Hence, sample locations were well represented by the dominant agro-ecological zones and lithologies of the study area as shown in Supplementary Figure S1.
Figure S1 Proportion of elevation (classified based on traditional agro-ecology) and lithology for the observed and predicted map. For explanation on variables see Table 1.
LUC categories on APs and GE were counted on the screen by three experts. The experts had to count different APs and GE in order to do the point counting in a relatively short time. Seven common LUC classes, forest, shrubland, cropland, grassland, bare lands, water bodies and built-up areas are considered in this study (Table 1). Important photo interpretation elements such as pattern, shape, association, tone and location of LUC were used to count each LUC (Loelkes et al., 1983). Keys were prepared for this counting (Supplementary Table S1)

2.3 Explanatory factors for land use/cover changes

Climatic variations, geomorphic settings and socio-economic effects are among the major explanatory factors of LUC occurrences. Northern Ethiopia is characterized by a large climatic and biophysical heterogeneity. Potential environmental factors believed to influence the LUC occurrences and changes in the Geba catchment are climate (precipitation, air temperature, evapotranspiration) and topography (elevation, slope), lithology and soil types. Despite the availability of long-term climate data at meteorological stations in and nearby the study area, a further analysis regarding the effects of climate variables on the distribution of LUC is not done in this study as the interpolation of the point meteorological data resulted in large errors due to strong topographic effects on the rainfall distribution. Other existing databases for spatial rainfall data (e.g. Rainfall Estimate (RFE), from the National Oceanic and Atmospheric Administration Climate Prediction Centre (NOAA-CPC)) was checked if it can be used but the validation of this database using the meteorological stations has resulted in a weak correlation (R2 = 0.38, n = 15). But climatic variables used as explanatory factors have a strong correlation with topography (such as elevation and slope).
In order to analyze the effect of lithology on LUC distribution, the detailed classification of the catchment was grouped into four major lithological categories based on their mineral composition and physical properties (Table 1). The soil map of the study area produced by Tielens (2012) was used for this study by grouping the soils into three major classes based on their suitability for crop production or plant growth (Supplementary Figure S2a and Table 1). This suitability classification was done using the soil structure, soil drainage and depth characterization of each soil unit (Tielens, 2012).
Figure S2 Geba catchment showing: a) lithology (extracted from Tesfamichael et al., 2010; Tesfaye and Gebretsadik, 1982) and soil suitability for cultivation (based on Tielens, 2012); b) slope gradient extracted from DEM - SRTM (USGS, 2014). AP: Location of Italian aerial photo
Socio-economic variables which anticipate to influence the LUC occurrence are mapped for the Geba catchment and are to be used in the LUC modelling. Population density, distance to main roads and distance to towns are the variables used to investigate if socio-economic factors are affecting LUC occurrences. The population density map for 2014 at 1 km grid was extracted from the Centre for International Earth Science Information Network (CIESIN, 2016). Mapping the distance of cells to main roads and towns in 2014 was based on the Google Earth database (January 10, 2014), while the map of distance to towns in the 1930s was created based on the1930s topographic maps of the study area (IGM, 2012) (see Supplementary Figure S3).
Table 1 Variables and their descriptions
Factor Variable Description Source/reference
Land use/cover Bare land Land with no vegetation cover, rock outcrop, quarry WBISPP, 2003
Built-up area Land under settlement, roads
Cropland Cultivated land (irrigated and non-irrigated) including open and regularly ploughed with or without shrub or tree line (boundary) and scattered trees, fallow with and without bushes/trees
Forest Land covered with dense trees or open; woodland; riparian trees, plantation (large scale and woodlot), church forest
Grassland Land covered with grasses used as grazing area
Shrubland Land covered with bushes: open, open with trees, dense, dense with trees, exclosures
Water body Land covered with water: lake, pond, river including dry river bed
Topography Alt (m a.s.l.) Average elevation at different locations DEM-SRTM at 30× 30 m resolution
Slope (%) Average slope gradient at different locations. DEM-SRTM at 30 ×30 m resolution
Soil type SS Soils suitable for cultivation, well to perfectly drained, fertile, moderately deep to deep soil. e.g. Vertic Cambisols, Calcaric Vertisols, Vertic Phaeozems Tielens et al. (2012)
SMS Soils moderately suitable for cultivation, shallow to moderately deep, moderate fertility, moderately drained. e.g. Eutric Regosols, Eutric Cambisols, Calcic Luvisols, Calcaric Cambisols
SNS Soils not suitable for cultivation, very shallow, rock outcrop, stony, excessively or poorly drained. e.g. Leptosols, Gleysols
Lithology LV Volcanics (intrusive and extrusive): trap series; Mekelle dolerite. Contain wide range of minerals, which enhance growth of tree and crop Tesfamichael et al. (2010); Tesfaye and Gebretsadik (1982)
LSC Sedimentary rock dominated by calcium carbonate: metalimestone; Antalo limestone; Agula shale. The land has a dry aspect because of karst and high infiltration.
LSNC Sedimentary rocks (non-carbonate): slates, metavolcanics; Edaga Arbi glacials; alluvium. Fine-texture, results in slow infiltration and relatively fertile soils
LSS Sandstones: metaconglomerate; major intrusive (these are granites); Enticho sandstone; Amba Aradom sandstone; Adigrat sandstone. These rocks have coarse texture, silica dominated. Sandy weathering materials where water will easily infiltrate; the domination of Si further makes that few minerals are available for vegetation growth
Socio-economy Pd (#/km2) Population density in 2010 CIESIN, 2016
DT (km) Distance to town in the 1930s (DT30 and distance to town in 2014 (DT14)
DR14 (km) Distance to road in 2014

2.4 Accuracy assessment of the point counting method and spatial map

An accuracy assessment was necessary so as to evaluate the validity of points counted from APs and GE images. Confusion (error) matrixes (Congalton et al., 1983; Hoffer, 1975) were prepared for the point counting and the fractional mapping accuracy measurements (Table 2). For the accuracy assessment of the point counting 275 ground control points that correspond to the points counted on APs/GE were randomly selected from 55 APs (i.e. 3-7 points per photo) and used for verification of both the 1930s and 2014 LUC. For the 1930s LUC, the verification was done based on the information obtained from old people who have lived in the area and have information on the historical LUC (from their parents). For 2014, field observations were made to validate whether the different LUC types were correctly identified on the screen. The assessment of accuracy of the fractional maps of the dominant LUC of the study area was done using the Google Earth image (Fritz et al., 2009; Bastin et al., 2013; Annys et al., 2016). Using the 2 km × 2 km grid, 233 points have been randomly selected utilizing ArcMap from the map of the Geba catchment. Then the values of the different LUC maps of 2014 corresponding to the random points were extracted. Based on these values the dominant LUC at each random point was selected and contrasted to the dominant LUC in the corresponding point in the fractional maps. The same random points were overlaid on the 2014 GE image so as to count the classes that were correctly allocated on the map. The dominant LUC in the random point corresponds to a 1.5 km × 1.5 km pixel on the fractional map, while the observation on GE was just at a point which can result in the error of disagreement by selecting the non-dominant class. An accuracy check for the historical LUC map could not be made as no field database exists.
Table 2 Commission-omission error matrix of LUC point count and mapping
Ground control / Google Earth
Cropland Shrubland Forest Other
cover		 Sum Commission
error
LUC count on screen Cropland 132 1 0 2 135 0.02
Shrubland 6 88 3 2 99 0.13
Forest 0 0 6 0 6 0.00
Other 3 2 0 30 35 0.10
Sum 141 91 9 34 275
Omission error 0.06 0.03 0.33 0.12
Overall accuracy 0.93
Kappa coefficient 0.89
LUC
map		 Cropland 106 10 1 6 123 0.16
Shrubland 12 68 1 1 82 0.21
Forest 1 1 3 0 5 0.67
Other 4 2 0 17 23 0.35
Sum 123 81 5 24 233
Omission error 0.14 0.16 0.40 0.29
Overall accuracy 0.83
Kappa coefficient 0.72

2.5 Data analyses

The fractions of the LUC classes, described as the density or weight of LUC, were calculated for each scene (AP and GE) as the ratio of the total number of points for each class to the total number of points of all LUC counted per AP/GE (Daniels et al., 1968). Points where the LUC is unknown or invisible on APs due to the clouds or damage of the photographs were excluded from further analyses and the corresponding points on GE omitted, too. The significance of the differences in the fraction of each LUC type between the 1930s and 2014 were tested by using the Mann-Whitney U test (α = 0.05), while the significance in the frequency of spatial distribution of LUC across the sample area (scene) was checked by using the Chi-square test (α = 0.05). The spatial distribution of LUC refers to the binary description of the LUC occurrence (i.e absence or presence) in the sample scene. A change matrix was prepared between LUC in the 1930s and 2014 so as to analyze the transformation of LUC in 2014 as compared to the 1930s.
Rasters of the selected explanatory factors with a resolution of 1.5 km × 1.5 km were used in order to fit approximately to the scale from which the LUC data were obtained from the APs and GE images (ca. 4.66 km2). This grid size was selected based on the smallest scene (2.4 km2), so that all observations have at least 1 cell. Average values of different explanatory factors at the location of all scenes were extracted from their spatial maps created for the Geba catchment. After checking the non-linear relation (using multiple linear regression) between LUC and the explanatory factors, Multivariate Adaptive Regression Splines (MARS) (Friedman, 1991), a non-parametric regression was used to model the LUC classes. It is a stepwise (forward and backward pass) method and a powerful and flexible non-parametric regression for modelling the complex multivariate datasets (Friedman, 1991) widely applied in various disciplines including the LUC change studies (e.g. Quiros et al., 2009; Tayyebi and Pijanowski, 2014). It uses the basis functions as predictors in place of the original data by breaking the data into different regions for fitting. The generalization of the model is checked using the Generalized Cross Validation (GCV) and by its normalized value (GCV R2) statistics which are also used to avoid over-fitting training data (Friedman and Silverman, 1989). The relative importance of the explanatory variables is ranked using the three criteria method of the MARS model (number of subset, GCV and RSS) which varies for different LUCs (Table 4).
The MARS equations developed for different LUC types in the 1930s and 2014 were used to create fractional maps of the LUCs, using pixel size of 1.5 km × 1.5 km that are approximately the same size as the smallest scene area (2.4 km2) of the interpreted APs. In a fractional map, for every pixel, the fraction (0.00 -1.00) or the percentage (0%-100%) of a particular LUC class is represented (Romanov et al., 2003).
Point count and mapping of LUC were validated by computing the commission errors (measure of the producer’s accuracy), the omission error (measure of the user’ accuracy) and overall accuracy from the error matrix table (Hoffer, 1975; Congalton et al. 1983). Furthermore, the degree of agreement between the accuracy of the counting and the field observation and between the mapping and Google Earth was measured by computing the Kappa coefficient of agreement from the confusion matrix (Fleiss, 1971; Stehman, 1997). This coefficient is used for the correction of the chance (expected) agreement. The multiple Kappa coefficient, i.e. the Fleiss Kappa coefficient, was applied by taking into account the multiple random points and the different LUC classes used for this precision assessment (Fleiss, 1971). Hence, the Kappa coefficient is calculated from the LUC error matrix as follows:
where Pii = Nij/N (i.e. proportion of the correctly counted LUC or ratio of diagonal value to the total number of observation); Pi+ = Ni/N (i.e, proportion of the marginal row total); P+i = Mi/N (i.e. proportion of the marginal column total). Statistical analyses were carried out in R 3.3.2 and SPSS ver. 21 software packages while all mapping was done in ArcGIS 10.1.

3 Results

3.1 Geomorphic settings

Sedimentary rocks with calcium carbonate dominate the study area followed by sedimentary rocks without calcium carbonate such as sandstones (17%), while volcanic rock covers a smaller area (10%) (Supplementary Figure S2a). The soil type of southern and central Geba is categorized as suitable (34%) and the northern and southwestern parts are moderately suitable (35%) while eastern areas are dominantly non-suitable soils (31%) (Supplementary Figure S2a). The distribution of the soil suitability classes is related to the geology and topography of the catchment (Tielens, 2012; IUSS, 2015).
Figure 2 An example of a point grid superimposed on: a) aerial photograph of January 3, 1936 with coordinate of center of vertical photo 13.561547°N and 39.024014°E (i.e. south of Abiy Adi); b) Google Earth image of January 4, 2014. Significant conversion of LUC occurred at this location as shown by: c) terrestrial photo around grid point N16 where land cover changed from dense forest in 1936 (d) to open forest in 2014 (e), and f) terrestrial photo around grid point F12 where land cover changed from open forest in 1936 (g) to cropland with trees in 2014 (h)

3.2 Change of LUC fraction from the 1930s-2014

Despite the difficulty of stereo viewing due to high obliquity and small, overlapping area of vertical photos, the 1930s APs are found to be an important source of historical information. Field observation for the verification of the point counting method has resulted in a high overall accuracy (93%) for the recent LUC, as shown in the error matrix (Table 2). The count of shrubland and cropland led to a high accuracy (i.e. a low omission error), 97% and 93% respectively, while the forest count resulted in a low accuracy. The error matrix shows that other land (summation of bare land, housing, waterbody, grassland) was identified with a high accuracy (89%) or low errors. Moreover, the error matrix caused a very strong Kappa coefficient (κ = 0.89) (Table 2). The accuracy assessment for the 1930s LUC was not possible to do through a field visit but showed a high correspondence (86%) to the result of the interview carried out on the old people concerning the history of land use in ancient times. The interviewed people with an average age of 69 years have explained the historical LUC of their area with some important landmarks, which they remember or which is based on what they heard from their parents. For example, the forests and shrubs in which they were cutting trees for different purposes such as to build their house and fence, for firewood and farming implements do not exist or are being degraded now.
The results of the quantitative analysis indicate that shrubland and cropland were the dominant LUC in both the 1930s and 2014. In the 1930s, the percentage of shrubland and cropland covers 48% and 39% respectively, but in 2014 the cropland outstretched to 42%, while the shrubland contracted to 37% (Figure 3). All LUC categories (except shrubland and forests) have increased over the last 80 years (Figure 3).
Figure 3 Areal percentage of different land use/cover types in the 1930s and 2014 (n = 34192)
The test of LUC fractions using Mann-Whitney U tests showed significant differences between the 1930s and 2014 for all categories except for cropland and forest. This test revealed that the fraction of shrubland decreased significantly in 2014 as compared to the 1930s while the bare land, grazing land, built-up area and water body had increased significantly in 2014 compared with the 1930s. The forest cover has dropped from about 6.3% to 2.3% during the last 80 years.

3.3 Spatial distribution of land use/cover

In the 1930s, shrubland and cropland were frequently observed (greater than 90%) categories at sample locations while the other categories occurred in less than 45% of the total sample areas. In 2014, all LUC classes had been encountered in about 85% of the observation (n = 139), except forest and water body which occurred in 45% and 71%, respectively (see Supplementary Figure S4). LUC types have undergone dynamic changes (increase or decrease) over a long time period (Figures 2-4). Nevertheless, there are cases where constant fractions and patterns of LUC were observed over this period. The chi-square test for the frequencies of spatial occurrence across the sample areas reveals significant changes in the location of occurrence of LUC except for forests (Supplementary Table S2).
Although cropland did not experience significant changes in fraction between two times, it showed however important spatial changes. In other words, some land that was under agriculture in the 1930s, was abandoned in 2014 and new agricultural lands were created by converting other LUC. The constant fraction of the cropland over a long period was retained at the expense of the other LUC mainly by encroaching and/or expanding into shrubland and forests, and abandonment of marginal, exhausted land, as well as the steepest slopes ban by administrative decision. This study also shows a significant decline in the spatial distribution of shrubland and the expansion of bare lands, grasslands and built-up areas (Supplementary Figure S4, Supplementary Table S2).
The change matrix (based on point counting) revealed a complex transformation of LUC over the last 80 years (Table 3). It should be noted that this transformation matrix is not free from errors due to the obliqueness of the APs and the non-georeferenced grid points on APs, which have resulted in a few meters’ displacement of points on GE as compared to their location on AP. Nevertheless, the matrix showed that all LUC have replaced each other although the transformation intensity is variable (Table 3). The results indicate that about 67% and 54% of the cropland and shrubland respectively, remained spatially unchanged in 80 years, while only 5% of the forest was recorded at its location in the 1930s. A large portion of cropland has transformed into shrubland, built-up area, bare land and grazing land; shrubland transformed into cropland, bare land, built-up area and grazing land while forest mainly changed into shrubland to cropland in the order of importance.
Table 3 Land use/cover transformation matrix from the 1930s to 2014 in Geba catchment
1935/
2014		 Cropland Shrub
land		 Forest Grazing
land		 Built-up Bare
land		 Water
body		 Total
Cropland 9047 2193 116 434 920 714 168 13592
Shrub
land		 4195 9151 351 460 680 1701 289 16827
Forest 514 861 84 36 123 80 17 1715
Grazingland 125 95 12 18 85 21 6 362
Bare land 265 410 12 23 60 226 23 1019
Built-up 79 29 4 7 117 10 1 247
Water body 67 177 5 7 40 75 59 430
Total 14292 12916 584 985 2025 2827 563 34192

3.4 Explanatory factors affecting the land use/cover

The results of the MARS model have revealed the importance of different factors for the occurrence probability of different LUC types in the 1930s and in 2014 (Table 4). The models show that all climatic environmental and socio-economic variables considered in this study have significantly affected one or more LUC types. However, the pressure of the socio-economic factors (i.e. the population density, distance to town) on LUC change have been increasing after the 1930s. For each model, several variables and terms (basis functions) were used to predict the LUC occurrences (Table 4 and Supplementary Table S3). The occurrence of more than one basis function for a single explanatory factor in a model depicts the nonlinear relations between the explanatory factors and LUC categories. The regression models have also illustrated the significant interaction of the explanatory factors on the distribution of LUC (Supplementary Table S3). The results also demonstrated different thresholds for different explanatory variables in the 1930s and 2014.
Table 4 Relative importance of variables in the model using three criteria number of subset (the number of model subset that includes the variables), RSS (the scaled summed decrease of residual sum of squares overall subset) and Generalized Cross Validation (GCV). For the explanation on the variables see Table 1
1930s 2014
Predictor Number of subset GCV RSS Predictor Number of subset GCV RSS
Cropland SS 7 100 100 SS 8 100 100
DT30s 6 75 77 Pd 7 66 69
Slope 6 75 77 Slope 6 55 59
Alt 2 17 26 Alt 5 30 39
SMS 1 12 18 SMS 2 4 18
GCV R2 = 0.52 R2 = 0.64 GCV R2 = 0.63 R2 = 0.73
Shrubland Alt 3 100 100 SS 5 100 100
Slope 2 54 58 Slope 4 51 55
SNS 1 28 33 Alt 3 32 38
SMS 1 13 18
GCV R2 = 0.38 R2 = 0.44 GCV R2 = 0.58 R2 = 0.64
Forest Slope 6 100 100 Alt 6 100 100
Alt 6 100 100 LSC 6 100 100
SNS 6 100 100 Pd 6 100 100
LSC 4 36 49 SS 6 96 97
LSNC 2 25 34
GCV R2 = 0.40 R2 = 0.52 GCV R2 = 0.31 R2 = 0.47
Among all the considerable variables, the slope gradient, elevation, soil suitability for cropping and proximity to town were significantly influencing the cropland fraction in the 1930s explaining 64% of the probability of its occurrences. The soil type which was the main suitable soil for cultivation was the most important factor for the occurrence of a larger fraction of cropland in the 1930s followed by the slope gradient and the distance from town. The result shows that in moderately sloping to flat areas (i.e. a slope gradient of less than 16%), the fraction of cropland was positively affected while the fraction was decreasing when the proximity of the area to town decreased except in suitable soil areas (Supplementary Table S3). Similarly, in 2014, the soil type and slope gradient remained the dominant factors for determining the distribution of cropland in which suitable soil and slope gradients of less than 16% were increasing the fraction of cropland. However, the effects of suitable soils and slope gradients were reversed when combined with the population density. The MARS also showed a larger cropland fraction in mid to high elevation areas except at steep slope gradients during the 1930s and 2014 (Supplementary Table S3). From the MARS model, it is also apparent that the distribution of shrubland was highly dependent on the slope gradient but in reverse direction to the cropland distribution. In flat areas, the fraction of shrubland was negatively affected while in sloping and steep slope areas the coverage had increased in the 1930s and 2014. Moreover, soil types and elevation were also other dominant explanatory factors for the distribution of shrubland. The result showed that suitable soil for cultivation favoured the occurrence of shrubs while suitable and moderately suitable soils affected the distribution of shrubland negatively. Moreover, the results revealed that when elevation increases over circa1800 m the fraction of shrubland had decreased both in past and present times (Supplementary Table S3).
During the 1930s and 2014, elevation, lithologies, soil suitability and slope gradients showed an important relation to the presence of forests although their explanatory power was weak (R2 = 0.52 in the 1930s and R2 = 0.47 in 2014) (Table 4). In both the 1930s and 2014, the elevation was the most important factor for the occurrences of forest in which a larger fraction of forest exists in areas with an elevation of over 1900 m. The interaction of elevation and lithology, soil suitability and population density resulted in different fractions of forest. The result also depicted that it was unlikely to find forest in areas with suitable soil for cultivation, while the increase of population density affected the forest occurrence positively, particularly in 2014.
The multivariate regression models developed for different LUCs in the 1930s and 2014 were used to calculate fraction maps of three major LUC types (i.e. cropland, shrubland and forest) at least five spatial raster layers, which represent the explanatory factors, were prepared to create the fractional maps of cropland, shrubland and forest in the 1930s and 2014. Hence, these fractional maps indicated the spatio-temporal distributions and the changes of LUC. The mapping of LUC of the Geba catchment (Figure 4) was done by 1.5 km × 1.5 km grid of the important explanatory factors in each model. As the values of every pixel are not absolute, percentages of different LUC types, separate maps were produced for the cropland, shrubland and forest. Colour gradients of each LUC have been used to compare the percentages of areal distribution and changes of LUC in the study area over the last 80 years. In these maps high gradients (larger percentages) of a different LUC did not occur on the same location, although mixed LUCs having a smaller percentage on a particular location are observed (Figure 4). Given the several limiting factors for the inaccurate mapping of LUC, the validation of the maps of 2014 using 233 random points on GE showed a high accuracy for cropland (86%) and shrubland (84%) and a lower one for the forest map (60%) and other land covers (71%). Overall, the confusion matrix showed a higher overall accuracy for the fractional map (83%) and a very strong Kappa coefficient (72%) (Table 2).
Figure 4 The spatial distribution and change of three dominant land use/cover of the study area: Cropland (a-c), shrubland (d-f) and forest (g-i). A positive (+) and negative (-) sign in these figures indicates the spatial expansion and shrinkage of LUC types, respectively. Graduated classifications illustrate the percent occurrence of land use type in the 1930s and 2014 that also reveal the degree of spatial changes within each location. Sd, Sf1, Sf2, Sh and Ss are supplementary data at Dergajen, Dessa forest 1, Dessa forest 2, Hadinet and Sinkata respectively.

4 Discussion

4.1 Point counting method, APs and GE images

The high overall accuracy (93%) and a very strong Kappa coefficient agreement (89%) of LUC counting as shown in the error matrix (Table 2), validated that counting LUC on Google Earth was done accurately. A good visualization and a high resolution nature of GE motivates, beside its potential for validation, its direct application in the environmental inventories includes LUC studies (Frankl et al., 2013b; Fritz et al., 2009; Hu et al., 2013). The medium scale and zooming technique make the photographs comparable to the Google Earth image resolution and viewing, which suggests the appropriateness of the comparison of results from the two sources. Frankl et al. (2013b) have used a combination of historical aerial photographs and GE images so as to analyze the temporal change of a gully network in Northern Ethiopia. Further, we are aware of only one study that used aerial photographs produced between 1936 and 1941 at a scale of 1:64,000 for a land use cover change study in Texas and New Mexico, United States (Scanlon et al., 2007).

4.2 Land use/cover fraction

Cropland showed a slight increment over the long time period. Despite the rapid rise of the human population (Nyssen et al., 2009), whose livelihood largely depends on agriculture (Deressa et al., 2008), the area of cultivated land remained almost constant during a long period of time. This indicates that suitable land for agriculture had been entirely occupied for many years, probably centuries, in Northern Ethiopia. Similar results have been reported locally in the catchment (that cultivated lands nearly did not expand over tens of years) (Alemayehu et al., 2009; Meire et al., 2013; Teka et al., 2013). Mitiku et al. (2006) documented that limits to lands (suitable for agriculture) are reached in Northern Ethiopia and that food demands for the increasing population could be met through an intensified use of the existing cropland. On the contrary, a significant expansion of cropland occurred in other parts of Ethiopia (e.g. Rembold et al., 2000; Zeleke and Hurni, 2001; Tsegaye et al., 2010), Africa and worldwide (Lambin et al., 2003; MEA, 2005) during the second half of the 20th century. Dynamic changes (expansion or contraction, rapid or slow) of cropland were reported in the southwest of Ethiopia despite a rapid population growth in the region in the last 50 years (Reid et al., 2000). On the other hand, the most dominant land cover in the 1930s, namely shrubland, has significantly decreased in fraction over the last 80 years. The decline might not only be visible in the land’s percentage under shrub cover but the quality of the shrubland might also have deteriorated compared to history, although it had not been quantified. Field visits demonstrated that in 2014, the shrubs had a low plant density and an open canopy cover. The forest showed a declining trend from its already low percentage in the 1930s (from 6.3% to 2.3%). This small fraction of forest cover in the 1930s indicates that the forest resources had been cleared, even before the 1930s. Despite the strong claims on a dense forest cover in the 1930s in Ethiopia (including our study region), there is no reliable record of forest cover and no precise date and rate of deforestation (Pankhurst, 1995; Woien, 1995). Previous studies in or close to our study area have reported that the forest and shrubland illustrated an increment during the last three to five decades (Alemayehu et al., 2009; Teka et al., 2015; Meire et al., 2013; de Muelenaere et al., 2014; Teka et al., 2013). However, our results may not be contradictory to such findings as their study period only extended to 1965 and covered small parts of the study area. The study carried out by Nyssen et al. (2015) in Northern Ethiopia concerning environmental conditions over the last 145 years indicates the highly variable land degradation. We thought that the woody vegetation cover peak at the end of the 1930s, strongly declined until large-scale soil and water conservation activities started in the 1990s. Hence, our results are in line with those of Nyssen et al. (2015): as compared to the 1930s, the land is still more degraded nowadays in Northern Ethiopia.
Despite extensive forest rehabilitation practices in Northern Ethiopia, there was an exceptionally ongoing deforestation on the remnant natural forest in the region (Munro et al., 2008). Recent deforestation has been detected in the Mt. Lib Amba and Simien Mountains, Northern Ethiopia (Jacob et al., 2015, 2017). Our results also clarify an increase in bare land, grassland, built-up area and water body, which is in agreement with earlier findings (Alemayehu et al., 2009; Meire et al., 2013; Teka et al., 2013), particularly before the start of the SWC practices. The overall decline of vegetation in the last 80 years is also in line with the existence of larger areas where the SWC (including exclosure) has not been fully implemented, for example in the lower parts of the study area, which are relatively remote.

4.3 Spatial change of land use/cover

Considerable spatial changes of LUC have taken place over the last 80 years. In the 1930s, less heterogeneity of land use in aerial photography (scenes) was observed as compared to the data of 2014. In other words, LUCs were less fragmented in the 1930s. By 2014, LUC classes had encroached upon the land that was under different use/cover during the 1930s. Hence, it is not uncommon to observe a mosaic LUC, such as cropland that encroached shrubland, a tree plantation (mainly Eucalyptus) in cropland as patches or in a linear form (Meire et al., 2013), settlement and water body in cropland and so on. However, this study has a limitation of showing the spatio-temporal dynamism of LUC change over the last 80 years in the Geba catchment, due to a lack of intermediate time period data on LUC.
Although cropland did not show a significant change in fraction over a long time (section 4.2), an important spatial change was noted that about 33% of the cropland was transformed to a different LUC, mainly to shrubland, built-up area and bare lands (Table 3). It is obvious that cropland, which lost its productivity due to an exhaustive cultivation, can no longer be used as cropland but is converted to bare land, degraded grazing land or shrubland, unless it is reclaimed. On the other hand, lands that were under shrub, forest and grass in the 1930s was converted to cropland in 2014, probably to search fertile soils. This result is consistent with previous studies (e.g. Zeleke and Hurni, 2001; Alemayehu et al., 2009). There are also some areas where bare lands were converted to cropland which shows the critical shortage of suitable lands for cropping leading an agricultural expansion into marginal lands. Shrubland was also affected by an increase of grazing land, which can be explained by the rise in livestock production associated with the population growth. The transformation of forest into shrubland and cropland in 2014 shows that deforestation had continued over the last 80 years. The conversion of bare lands to forest and shrubland can be linked to the plantation forest and the implementation of exclosures over the last few decades. Considerable fraction of built-up area were recorded in 2014 on the locations previously (1930s) covered by cropland, shrubland and forest which can be linked to a rapid population increase. Other studies also indicate that built-up areas (mainly urban areas) increased at the expense of cropland in Ethiopia during the last few decades (Haregeweyn et al., 2012; Miheretu and Yimer, 2017).

4.4 Explanatory factors of land use/cover distribution and change

Overall, over the last 80 years, the study area had been hit by two major droughts and several famines. The common explanation for LUC change in earlier research carried out in the north or in other parts of Ethiopia, comprises the socio-economic forces, policies and institutions (Alemayehu et al., 2009; Teka et al., 2015; Meire et al., 2013; de Muelenaere et al., 2014; Tadesse et al., 2014; Teka et al., 2013). Physical elements are rarely correlated to LUC changes (Reid et al., 2000; Tadesse et al., 2014), though they trigger significant changes particularly when the land is under stress (Lambin et al., 2001). The present study demonstrated that consideration of topography, soil type and lithology as potential explanatory factors of LUC occurrence, provided a moderate to high model fitting and the validation of results, particularly for cropland and shrubland.
Despite the shortage of suitable cultivation lands in the study area for a long time, steep slope areas remained unsuitable for agriculture. Other reports also indicate that arable lands are frequently observed on level to gentle slope lands (plains, foot slopes and valley floors) (Meire et al., 2013; Teka et al., 2013). But currently, cropland sometimes appears on steep slopes where soil and water conservation (SWC) measures like stone bunds or trenches are executed so as to counter soil erosion. This is in line with a previous study on long-term land use change in the same region (Meire et al., 2013). The current inverse correlation between the population density and the cropland fraction describes the conversion of cropland to built-up (such as housing, roads) areas following a population increase. Ramankutty et al. (2002) explained that the rapid population increase and urbanization resulted in less cropland area per capita. Jacob et al. (2015) have analyzed an increase of the tree line elevation in mountainous areas due to anthropogenic pressure. On the other hand, steep slope lands appear to be reserved for shrub use in the study area, (which is) explained by the extensive conservation measures that had been carried out in degraded areas of the Tigray region during the last two to three decades. Exclosures, as part of SWC, were mostly applied on very steep and degraded slopes, which suffer from a severe soil erosion. Various reports illustrated that sloping and steep areas are often employed for afforestation in Northern Ethiopia (Descheemaeker et al., 2006b). Forests that grew in suitable soils for cropping in the 1930s, had been deforested in 2014. Hence, in 2014, remnants of forests were available in areas where the soils are moderately suitable and non-suitable for cropping, sandstone and sloping, which appear to the fact that afforestation is promoted in less fertile soils. Currently, the forest density has augmented nearby towns which can be linked to the increased awareness of tree growing and management in the Tigray region (de Muelenaere et al., 2014). Clusters of forests planted along farmland boundaries and in villages are commonly encountered in the region (Meire et al., 2013).
The creation of fractional maps on different LUCs in the Geba catchment, which use model developed for each class with a set of explanatory elements, resulted in an approximately similar pattern with the observed fractions (Figure 4). It is important to possess proportional samples of AP/GE (in different explanatory factors) (Supplementary Figure S1) for the prediction of land use/ cover distribution in the entire catchment. The verification of these maps using Google Earth images (2014) showed the model validity in order to predict the occurrence of LUC, particularly for 2014. Overall, 84 % of the LUC was correctly allocated on the fractional map of LUC. Cropland and shrubland were more or less predicted accurately, while large errors had been noticed in the forest prediction. This poor forecast accuracy can be related to the small fraction of forest in each scene which less likely dominates the scene. Moreover, the very strong Kappa coefficient agreement (k = 72) confirms the validity of the created fractional maps. In general, this result suggests that the models we developed for the Geba catchment are reliable to foretell the fraction of LUC change, while using the important explanatory factors selected in each model (Table 4 and Supplementary Table S3).

5 Conclusions

This study demonstrated the usefulness of the analysis of the 1930s APs and GE images for the study of land/cover distribution and changes. The results have demonstrated significant modifications in the fraction and spatial shifts of LUC during the last 80 years. Despite insignificant changes in the fraction of cropland area, 39% in the 1930s to 42% in 2014, it partially shifted its location at the expense of other LUC. The transformation matrix illustrates that 33% of the cropland was given away to different LUC types, mainly to shrubland, bare land, and grassland probably due to its decreasing productivity and change to built-up areas, which can be explained by a rapid population growth. Shrubland is the most affected LUC over this long time period, as it significantly shrank from 48% in the 1930s to 37% in 2014, associated with a shift of cropland and an expansion of the built-up area and grazing land. Forest cover has dropping continuously in the last 80 years, from about 6.3% through an absolute minimum in the 1970s-1980s to less than 2.3% in 2014. The increased frequency of occurrence of different LUC types in observation areas (scenes) shows a more mixed or fragmented LUC system in 2014 compared to the 1930s.
The effects of different forces (environmental and socio-economic variables) on LUC distribution was indicated by non-linear regression analysis. This study also indicates that explanatory factors influence LUC types at different thresholds. Cropland was generally recorded in flat to sloping areas (<16%), while shrubland and forest were often seen on slopes above 5%, although the thresholds change when other important factors exist. The latter (such as elevation, soil suitability, lithology and socio-economic factors) were also very important for influencing the distribution of cropland, shrubland and forest. The comparison of the fractional maps with the observed fraction reveals similar patterns in their distribution. The validation of this fractional map on Google Earth demonstrated a high overall accuracy (83%) and a strong Kappa coefficient (72%), which confirm the usefulness of the databases (GE and explanatory factors) and the MARS model in order to create an accurate LUC fractional map of LUC.
Overall, this study provided useful information regarding the condition of LUC in the Geba catchment in the 1930s and 2014, demonstrating larger areal fractions of shrubland and forest and an approximately constant cropland area in the1930s as compared to 2014. This suggests that more efforts of land management (SWC and exclosure) practices need to be implemented particularly in remote areas. Moreover, further investigation on the historical LUC of the study area could prove the database on the environmental condition in the past so as to evaluate the ongoing SWC interventions or to design new land management strategies.
Appendix
Table S1 Keys used for the classification of LUC during the point counting on AP and GE
Major class Details class Description
X Impossible to interpret (clouds, damage to photo, poor scan quality)
U Unsure, unknown land cover class
Bare land B Bare (bare soil, rock outcrop), mining area
Cropland C0 Cropland fallow (scattered small shrubs)
C1 Cropland (open, regularly ploughed)
C2 Cropland with shrub or tree line (on lynchet or boundary)
C3 Cropland with scattered trees
Forest F0 Forest - open; woodland
F1 Forest - dense

Grazing land

G

Grassland

Built-up

H

Habitat (homestead, houses), road

Shrubland S0 Shrubland - open
S1 Shrubland - open - with trees
S2
S3		 Shrubland - dense
Shrubland - dense - with trees
Water body W Water (lake, river, dry river bed, reservoir)
Table S2 Frequency and percentage of scene at which land use/cover types have shown different changes between the 1930s and 2014; Chi-square test result. n = 139
Land use/cover Increased Decreased No change Sig.
Frequency Percent Frequency Percent Frequency Percent
Bare land 101 73 23 17 15 11 <0.001
Built-up area 113 81 17 12 9 6 <0.001
Cropland 85 61 52 37 2 1 <0.001
Forest 33 24 53 38 53 38 0.088
Grassland 96 69 14 10 29 21 <0.001
Shrubland 29 21 105 76 5 4 <0.001
Water body 73 53 28 20 38 27 <0.001
Table S3 Equations of three major land use/cover (cropland, shrubland and forest) in the 1930s and 2014 which were developed by Multivariate Adaptive Regression Spline model. C1930s = cropland in the 1930s, C2014 = cropland in 2014, S1930s = shrubland in the 1930s, S2014 = shrubland in 2014, F1930s = forest in the 1930s and F2014 = forest in 2014. h=hinge function with zero and a constant (knot) of a factor. For the explanation see Table 1.
Land use/cover 1930s 2014
Cropland C1930s =
0.2232
- 0.01063 * h (0, DT30s - 6)
+ 0.02519 * h (0, 17.4 - slope)
- 0.1589 * h (0, DT30s - 8) * SS
+ 0.09218 * h (0, 6 - DT30s) * SMS
+ 0.04719 * h (0, 7 - slope) * SS
- 0.00141 * h (0, 2037 - alt) * SS
+ 0.1724 * h (0, DT30s - 6) * SS		 C2014 =
0.1196
+ 0.2177 * SS
+ 0.0008321 * h(0, alt - 1859)
+ 0.02293 * h(0, 16 - slope)
- 0.000121 * Pd * SS
- 0.0000118 * Pd * h (0, 16 - slope)
+ 0.0005874 * h (0, 1859 - alt) * SMS
- 0.0000299 * h (0, alt - 1859) * h (0, slope - 7)
- 0.0001047 * h (0, alt - 2150) * h (0, 16 - slope)
GCV 0.041 RSS 4.281 GCV R2 0.53 R2 0.64 GCV 0.028 RSS 2.78 GRSq 0.63 R2 0.73
Shrubland
S1930s =
0.6112
+ 0.1556 * SNS
- 0.000357 * h (0, alt - 1778)
- 0.01745 * h (0, 15.4 - slope)
S2014 =
0.6841
- 0.2899 * SS
- 0.1131 * SMS
- 0.0003771 * h (0, 1778 - alt)
- 0.0002857 * h (0, alt1 - 1859)
- 0.02089 * h (0, 14 - slope)
GCV 0.044 RSS 5.441 GCV R2 0.38 R2 0.44 GCV 0.027 RSS 3.16 GCV R2 0.58 R2 0.64
Forest F1930s =
0.06381
+ 0.08914 * LSNC
- 0.0001301 * h (0, 2361 - alt)
+ 0.0001466 * h (0, slope - 6) * h (0, alt - 1980)
+ 0.0001778 * h (0, slope - 6) * h (0, - 2053)
+ 0.0001651 * slope * max (0, alt - 2361) * LSC
+ 0.000121 * h (0, slope - 6) * h (0, alt - 1980) * SNS		 F2014 =
0.006373
- 0.1273 * SS
+ 0.0002263 * h (0, alt - 1830)
+ 0.0003854 * h (0, alt - 2396)* SNS *LSC
+ 0.02682 * h (0, alt - 2396) * LSC
+ 0.000000614 * h (0, 2396 - alt) * h (0, Pd - 81)
Figure S3 Maps of socio-economic variables
Figure S4 Frequency of occurrence (absence or presence) of different land use/cover types in the 1930s and 2014 scenes, n = 139

The authors have declared that no competing interests exist.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Aerts R, Van Overtveld K, November Eet al., 2016. Conservation of the Ethiopian church forests: Threats, opportunities and implications for their management.Sci. Total Environ., 551: 404-414. doi: 10.1016/ j.scitotenv.2016.02.034.61Natural forest in northern and central Ethiopia is mainly confined to ‘church forests’.61We studied 394 forests in satellite images and field surveyed 78 forests.61Patches are species-poor but communities similar to potential natural vegetation.61Small patch sizes, isolation, edge effects threaten long-term conservation.61Improving management, protection and stakeholder benefits are crucial.

DOI PMID

[2]
Alemayehu F, Taha N, Nyssen Jet al., 2009. The impacts of watershed management on land use and land cover dynamics in Eastern Tigray (Ethiopia).Resour. Conserv. Recy., 53(4): 192-198. doi: 10.1016/j.resconrec.2008. 11.007.Integrated watershed management (IWSM) was implemented to address issues of poverty and land resource degradation in the 14,50002ha upper Agula watershed, in semi-arid Eastern Tigray (Ethiopia), an area known for poverty and resource degradation caused by natural and man-made calamities. The purpose of this study was to assess the impact of IWSM and determine the land use and cover dynamics that it has induced. The change in land use and cover was assessed by integrating remote sensing and geographic information systems (GIS). Two sets of aerial photographs (taken in 1965 and 1994 at scale of 1:50,000) and Landsat ETM+ image (taken in 2000 with 3002m resolution) were used to produce the land use/land cover map and assess land use change.The results reveal significant modification and conversion of land use and cover of the watershed over the last four decades (1965–2005). A significant portion of the watershed was continuously under intensively cultivated (rainfed) land. The area under irrigation increased from 702ha to 222.402ha post-intervention. The area under dense forest increased from 32.402ha to 9802ha.The study further shows that IWSM decreased soil erosion, increased soil moisture, reduced sedimentation and run off, set the scene for a number of positive knock-on effects such as stabilization of gullies and river banks, rehabilitation of degraded lands. IWSM also resulted in increased recharge in the subsurface water.This study reconfirms the importance of IWSM as a key to improve the land cover of watersheds, as a contribution to poverty alleviation and sustainable livelihood.

DOI

[3]
Amanuel Z, 2009. Assessment of spatial and temporal variability of river discharge, sediment yield and sediment-fixed nutrient export in Geba River catchment, northern Ethiopia [D]. K.U. Leuven, Belgium.

[4]
Annys S, Demissie B, Amanuel Zenebeet al., 2017. Land cover changes as impacted by spatio-temporal rainfall variability along the Ethiopian Rift Valley escarpment.Reg. Environ. Change, 17(2): 451-463. doi: 10.1007/ s10113-016-1031-2.Magnitudes of land cover changes nowadays can be assessed properly, but their driving forces are subject to many discussions. Next to the accepted role of human influence, the impact of natural climate variability is often neglected. In this study, the impact of rainfall variability on land cover changes (LCC) is investigated for the western escarpment of the Raya Graben along the northern Ethiopian Rift Valley. First, LCC between 2000 and 2014 were analysed at specific time steps using Landsat imagery. Based on the obtained LCC maps, the link was set with rainfall variability, obtained by means of the satellite-derived rainfall estimates (RFEs) from NOAA-CPC. After a correction by the incorporation of local meteorological station data, these estimates prove to be good estimators for the actual amount of precipitation (RFE1.0 = 0.85, p = 0.00, n = 126; RFE2.0 = 0.76, p = 0.00, n = 934). By performing several linear regression analyses, a significant positive relationship between the precipitation parameter DIFF 5Y (i.e. the at-RFE pixel scale difference in five-year average annual precipitation for the two periods preceding the land cover maps) and the changes in the woody vegetation cover was found (standardised regression coefficient b = 0.23, p = 0.02, n = 108). Despite the dominance of direct human impact, further greening of the study area can be expected for the future concomitantly to a wetter climate, if all other factors remain constant.

DOI

[5]
Araya A, Keesstra SD, Stroosnijder L, 2010. A new agro-climatic classification for crop suitability zoning in northern semi-arid Ethiopia.Agric. For. Meteorol., 150: 1057-1064. doi: 10.1016/j.agrformet.2010.04.003.The agro-climatic resources of Giba catchment in northern Ethiopia were assessed and characterized. The objectives were (i) to ascertain the suitability of the climate for growing teff ( Eragrostis tef) and barley ( Hordeum vulgare); (ii) to determine the onset and length of the growing period (LGP), (iii) to evaluate the traditional method of climate classification, and (iv) to produce comprehensive agro-climatic zones of the Giba catchment. The Ethiopian traditional method of climate classification based on temperature and altitude was found to be less relevant to crop suitability zoning in semi-arid regions of Northern Ethiopia because within this semi-arid drought-prone environment the rainfall is more important for crop growth than temperature. The LGP ranges from 60 to 100 days over the catchment, increasing from north-east to south-west. For the crop suitability zoning, the concept of growing period was introduced into the traditional approach, to produce agro-climatic zones. This method could be used to develop agronomic strategies to cope with the anticipated increase in drought in the semi-arid tropics under climate change. Accordingly, quick maturing and drought-resistant varieties of teff and barley can be grown in the centre and in the east, while medium-maturing cultivars should do well in the south-west. The method requires limited input data and is simple in its use.

DOI

[6]
Asmamaw L, Mohammed A, Lulseged T, 2011. LUC dynamics and their effects in the Gerado catchment, northeastern Ethiopia.Int. J. Environ. Stud., 68(6): 883-900. doi: 10.1080/00207233.2011.637701.This paper analyses the land use/cover dynamics of land degradation through the interpretation of aerial photographs (1958 and 1980) and 2006 SPOT-5 satellite image of the Gerado catchment. Other, non-visual data were gathered from personal interview and focus group discussions conducted in 2010 and 2011 with local elders, farmers and development (agricultural extension) agents. The results identified the presence of cultivated and rural settlement land, shrubland, woodland, bare land, grassland, urban built up area and forest. Throughout the period 1958 2006, urban built-up area, forest and cultivated and rural settlement land expanded at an average rate of 6.85%, 1.85% and 0.14% per year at the expense of shrub, wood and grasslands, which declined by 0.77%, 0.21%, 0.65% per year, respectively. The land use/cover dynamics of 1958 2006 resulted in the reduction/loss of biodiversity, occurrence of high soil erosion and ramification of gullies. The triggers for these changes were population growth, land cultivation, expansion of farmland, inappropriate land management, civil war and fuel wood demand. These led to further land degradation and more food insecurity among many farming households. Land resources have to be used according to their suitability. Thus, the exposed and steep mountains of the area have to be protected from cultivation and should be re-afforested. The paper discusses other implications for management and policy formulation also.

DOI

[7]
Bard K A, Coltorti M, DiBlasi M Cet al., 2000. The environmental history of Tigray (Northern Ethiopia) in the Middle and Late Holocene: A preliminary outline.Afr. Archaeol. Rev., 17(2): 65-86. doi: 10.1023/ A:1006630609041.Abstract&nbsp;&nbsp;This paper outlines the environmental history of the Tigrean Plateau (northern Ethiopia) during the Holocene, based on the available geomorphological, palynological, archaeological, and historical evidence. At present, it seems that (1) the plateau experienced a more humid climate with a denser vegetation cover during the Early Holocene; (2) Soil erosion due to clearing vegetation began in the Middle Holocene; (3) agricultural activity was intensified in the Late Holocene, as a consequence of the rise of a state; (4) demographic pressure increased from the early first millennium BC to the mid&#x2013;first millennium AD, causing soil erosion; (5) environmental degradation and demographic decline occurred in the late first millennium AD; (6) the vegetation cover was regenerated in the early second millennium AD; and (7) progressive vegetation clearance started again in the second half of the second millennium AD.Cet article trace l'histoire ambiante du Plateau Tigr&eacute;en dans l'Holoc&egrave;ne en utilisant les donn&eacute;es g&eacute;omorphologiques, palynologiques, arch&eacute;ologiques et historiques. Il semble que (1) dans l'Holoc&egrave;ne ancien le plateau &eacute;tait caract&eacute;ris&eacute; par une phase humide avec une dense v&eacute;g&eacute;tation; (2) l'&eacute;rosion caus&eacute;e par l'abbattage de la v&eacute;g&eacute;tation commen&cedil;a dans l'Holoc&egrave;ne moyen; (3) l'activit&eacute; agricole s'int&eacute;nsifia &agrave; la fin de l'Holoc&egrave;ne, par cons&eacute;quence de l'essor d'un &eacute;tat; (4) la pression d&eacute;mographique augmenta de plus en plus du d&eacute;but du premier mill&eacute;naire av. J.-Ch. &agrave; la moiti&eacute; du premier mill&eacute;naire ap. J.-Ch.; (5) la d&eacute;gradation ambiante et la diminution d&eacute;mographique se v&eacute;rifi&egrave;rent &agrave; la fin du premier mill&eacute;naire ap. J.-Ch.; (6) une r&eacute;g&eacute;neration de la vegetation se v&eacute;rifia au d&eacute;but du seconde mill&eacute;naire ap. J.-Ch.; et (7) l'abbatage de la v&eacute;g&eacute;tation recommen&cedil;a dans la seconde moiti&eacute; du seconde mill&eacute;naire ap. J.-Ch.Holocene&nbsp;-&nbsp;environment&nbsp;-&nbsp;history&nbsp;-&nbsp;Tigray&nbsp;-&nbsp;Ethiopia

DOI

[8]
Bastin L, Buchanan G, Beresford Aet al., 2013. Open resource mapping and services for web-based land-cover validation.Ecol. Inform., 14: 9-16. doi: 10.1016/j.ecoinf.2012.11.013.78 Web-based tools and mapping services can be used for quick visual assessment of land-cover change. 78 This can increase extent, coverage and quality of shared land-cover change maps. 78 We produce a new Web tool that is cheap, simple and has been successfully used by conservationists. 78 This could be used to monitor sites and inform conservation priorities. 78 Uncertainty information supplied by users helps identify inconsistencies and potential errors.

DOI

[9]
Bellhouse D, 1981. Area estimation by point-counting techniques.Biometrics, 37(2): 303-312. doi: 10.2307/ 2530419.Systematic sampling, the classical approach to the estimation of areas by point-counting techniques, is compared to stratified random sampling and to systematic sampling with multiple random starts. Under an assumed model, we find that stratified sampling is usually more efficient than systematic sampling which, in turn, is more efficient than the multiple-random-starts method.

DOI

[10]
Biadgilgn Demissie, Frankl A, Mitiku Haileet al., 2015. Biophysical controlling factors in upper catchments and braided rivers in drylands.Land Degradation and Development, 26: 748-758. doi: 10.100/lde.2357.

[11]
CIESIN (Center for International Earth Science Information Network)- Columbia University,2016. Gridded Population of the World, Version 4 (GPWv4): Population Density Adjusted to Match 2015 Revision UN WPP Country Totals.Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). Accessed 20 March 2016.

[12]
Congalton R, Oderwald R, Mead R, 1983. Assessing landsat classification accuracy using discrete multivariate analysis statistical techniques.Photogrammetric Engineering and Remote Sensing, 49(12): 1671-1678.The concentration of pppApp was not determined intracellularly by radioimmunoassay.

DOI

[13]
CSA, 2008. Summary and Statistics Report of the 2007 Population and Housing Census. Federal Democratic Republic of Ethiopia Population Census Commission. December 2008, Addis Ababa, Ethiopia, pp113.

[14]
Daniels R B, Gamble E E, Bartelli L Jet al., 1968. Application of the point count method to problems of soil microbiology.Soil Science, 106(2): 149-152.

[15]
de Muelenaere S, Frankl A, Haile Met al., 2014. Historical landscape photographs for calibration of Landsat land use/cover in the Northern Ethiopian Highlands.Land Degrad. Dev., 25(4): 319-335. doi: 10.1002/ldr.2142.ABSTRACTThe combined effects of erosive rains, steep slopes and human land use have caused severe land degradation in the Ethiopian Highlands for several thousand years, but since the 1970s, however, land rehabilitation programmes have been established to try to reverse deterioration. In order to characterize and quantify the transformations in the north Ethiopian Highlands, a study was carried out over 888465km2 of the Tigray Highlands of northern Ethiopia. Using Landsat Multispectral Scanner and later Thematic Mapper imagery (1972, 1984/1986 and 2000), historical terrestrial photographs (1974–1975) and fieldwork (2008), we prepared land use and cover maps. For assessing the use of the historical terrestrial photographs, Landsat images from 1972 were classified using two different methods, namely conventional change detection (image differencing) and ground truthing (using the historical photographs of 1974–1975). Results show that the use of terrestrial photographs is promising, as the classification accuracy based on this method (Kappa coefficient 0·54) is better than the classification accuracy of the method based on image differencing (Kappa coefficient 0·46). Major land use and cover changes indicate the following: (1) a gradual but significant decline in bare ground (32 per cent in 1972 to 8 per cent in 2000); (2) a significant increase of bushland (25 to 43 per cent) and total forest area (including eucalypt plantations, 2·6 to 6·3 per cent); and (3) creation of numerous lakes and ponds. The dominant change trajectory (27 per cent of the study area) indicates a gradual or recent vegetation increase. These changes can be linked to the population growth and the introduction of land rehabilitation initiatives, complemented by growing awareness of land holders. Copyright 08 2012 John Wiley & Sons, Ltd.

DOI

[16]
Deressa T, Hassan R M, Alemu Tet al., 2008. Analyzing the determinants of farmers' choice of adaptation methods and perceptions of climate change in the Nile Basin of Ethiopia. IFPRI Discussion Paper 00798, pp36.

[17]
Descheemaeker K, Nyssen J, Poesen Jet al., 2006a. Runoff on slopes with restoring vegetation: A case study from the Tigray highlands, Ethiopia.J. Hydrol., 331(1): 219-241. doi: 10.1016/j.jhydrol.2006.05.015.Daily runoff depths from 28 plots (5 m x 2 m) recorded during a 2-year period in the semi-arid to subhumid highlands of Tigray were analyzed to study the effect of vegetation restoration in exclosures and to identify other factors influencing runoff production. Plots are distributed over three study sites and located in different land use types and on different combinations of soil type, vegetation cover and slope gradient. Runoff was found to be significantly reduced when a degraded area is allowed to rehabilitate after closure. Runoff depth is significantly correlated with event variables such as rain depth, rainfall intensity, storm duration and soil moisture content. Total vegetation cover is the most important plot variable explaining about 80% of the variation in runoff coefficients through an exponential decay function. Also the runoff generating rainfall threshold has a positive correlation with total vegetation cover. Runoff was found to be negligible when the vegetation cover exceeds 65%. Other important variables affecting runoff production in the study sites are soil organic matter, soil bulk density, litter cover and slope gradient. (c) 2006 Elsevier B.V. All rights reserved.

DOI

[18]
Descheemaeker K, Nyssen J, Rossi Jet al., 2006b. Sediment deposition and pedogenesis in exclosures in the Tigray Highlands, Ethiopia.Geoderma, 132(3): 291-314. doi: 10.1016/j.geoderma.2005.04.027.In the Tigray highlands of Northern Ethiopia, the establishment of exclosures (i.e., areas closed for grazing and agriculture) has become an important measure to combat land degradation and restore vegetative cover. Exclosures are commonly found on steep slopes and downslope from a sediment source area. In this study their sediment trapping capacity and controlling factors were investigated. Total sediment depth turned out to be related to vegetation cover, sediment source area and in some cases slope gradient. Thickness of recent, short-term (&lt; 20 years) sediment deposits was strongly related to distance from the top edge of the closed area, slope gradient, vegetation cover and characteristics of the sediment source area. Mean sediment deposition rates ranged between 26 and 123 Mg ha61 1 yr61 1. Under influence of vegetation and sediment deposition dark soils rich in organic matter (Phaeozems) develop. In view of their high sediment trapping capacity, exclosures are highly valued as efficient soil conservation measures in the Tigray highlands.

DOI

[19]
FAO, 2010. Global Forest Resources Assessment: Main report. Foresty paper 163, FAO. Rome, Italy.

[20]
FAO, 2011. Food and Agriculture Organization Ethiopia Country Programming Framework: 2012-2015. Office of the FAO Representative in Ethiopia to AU and ECA, Addis Ababa.

[21]
Fleiss J L, 1971. Measuring nominal scale agreement among many raters.Psychological Bulletin, 76: 378-382.ABSTRACT Introduced the statistic kappa to measure nominal scale agreement between a fixed pair of raters. Kappa was generalized to the case where each of a sample of 30 patients was rated on a nominal scale by the same number of psychiatrist raters (n = 6), but where the raters rating 1 s were not necessarily the same as those rating another. Large sample standard errors were derived.

DOI

[22]
Frankl A, Nyssen J De Dapper Met al., 2011. Linking long-term gully and river channel dynamics to environmental change using repeat photography (Northern Ethiopia).Geomorphology, 129(3): 238-251. doi: 10.1016/j.geomorph.2011.02.018.In the Highlands of Northern Ethiopia gully occurrence is linked to poverty-driven unsustainable use of the land in a vulnerable semi-arid and mountainous environment, where intensive rainfall challenges the physical integrity of the landscape. Trends in gully and river channel erosion, and their relation to triggering environmental changes can proffer valuable insights into sustainable development in Northern Ethiopia. In order to assess the region-wide change in gully and river channel morphology over 140years, a set of 57 historical photographs taken in Tigray, and, clearly displaying gully cross-sections, were precisely repeated from 2006 till 2009. Ninety-two percent of the gully and river sections (n=38) increased in cross-sectional area during the studied period, especially after 1975. Two repeatedly photographed catchments of Lake Ashenge and Atsela allowed a detailed study of gully development from 1936 until 2009. A conceptual hydrogeomorphic model was devised for these catchments and validated for the Northern Ethiopian Highlands. Three major phases can be distinguished in the hydrological regime of the catchments. In the first phase, between 1868 (or earlier) and ca. 1965, the relatively stable channels showed an oversized morphology inherited from a previous period when external forcing in environmental conditions had caused the channels to shape. In the second phase (ca. 1965 ca. 2000), increased aridity and continued vegetation clearance accelerated the channel dynamics of the gully and river system. The third phase (ca. 2000 present) started after the large-scale implementation of soil and water conservation measures. In 2009, 23% of the gully and river sections were stabilizing. This paper validates previous research indicating severe land degradation in the second half of the 20th century. Additionally, it demonstrates that the recent erosive cycle started around 1965 and, that at the present time, improved land management stabilizes headwater streams.

DOI

[23]
Frankl A, Poesen J, Haile Met al., 2013a. Quantifying long-term changes in gully networks and volumes in dryland environments: The case of Northern Ethiopia.Geomorphology, 201: 254-263. doi: 10.1016/ j.geomorph.2013.06.025.Understanding historical and present gully development is essential when addressing the causes and consequences of land degradation, especially in vulnerable dryland environments. For Northern Ethiopia, several studies exist on the severity of gully erosion, yet few have quantified gully development. In this study, gully network and volume development were quantified over the period 1963-2010 for an area of 123 km(2), representing the regional variability in environmental characteristics. Gully networks were mapped from small-scale aerial photographs and high-resolution satellite images. For the latter, visualizing Google Earth images in 3D proved to be very suitable to investigate gully erosion. From the changes in networks and volumes over the period 1963-2010, the occurrence of one cut-and-fill cycle is apparent. From a largely low-dynamic gully system in the 1960s, network expansion and increased erosion rates in the 1980s and 1990s caused the drainage density and volume to peak in 1994. The average gully density (D-total) was then 2.52 km km(-2) and the area-specific gully volume (V-a) 60 x 10(3) m(3) km(-2). This coincides with soil losses by gully erosion (SLg) of 17.6 t ha(-1) y(-1) over the period 1963-1994. By 2010, improved land management and the region-wide implementation of soil and water conservation measures caused 25% of the gully network to stabilize, resulting in a net infilling of the gully channels over the period 1994-2010. The study validates previous findings that land degradation by gullying was indeed severe in Northern Ethiopia in the second half of the 20th century, but also shows that when proper land management is applied, a gully can be transformed into a linear oasis, which increases the resistance of gullies to further erosion. (C) 2013 Elsevier B.V. All rights reserved.

DOI

[24]
Frankl, A, Zwertvaegher A, Poesen Jet al., 2013b. Transferring Google Earth observations to GIS-software: Example from gully erosion study.Int. J. Digital Earth, 6(2): 196-201. doi: 10.1080/17538947.2012.744777.High-resolution images available on Google Earth are increasingly being consulted in geographic studies. However, most studies limit themselves to visualizations or on-screen measurements. Google Earth allows users to create points, lines, and polygons on-screen, which can be saved as Keyhole Markup Language (KML) files. Here, the use of R statistics freeware is proposed to easily convert these files to the shapefile format [or 090004.shp file format090005], which can be loaded into Geographic Information System (GIS) software (ESRI ArcGIS 9 in our example). The geospatial data integration in GIS strongly increases the analysis possibilities.

DOI

[25]
Frankl A, Seghers V, Stal Cet al., 2015. Using image-based modelling (SfM-MVS) to produce a 1935 ortho-mosaic of the Ethiopian Highlands.Int. J. Digital Earth, 8(5): 421-430. doi: 10.1080/17538947. 2014.942715.Approximately 34,000 aerial photographs covering large parts of Ethiopia and dating back to 19350900091941 have been recently recovered. These allow investigating environmental dynamics for a past period that until now is only accessible from terrestrial photographs or narratives. As the archive consists of both oblique and vertical aerial photographs that cover rather small areas, methods of image-based modelling were used to orthorectify the images. In this study, 9 vertical and 18 low oblique aerial photographs were processed as an ortho-mosaic, covering an area of 25 km2, west of Wukro town in northern Ethiopia. Using 15 control points (derived from Google Earth), a Root Means Square Error of 28.5 m in X 35.4 m in Y were achieved. These values can be viewed as optimal, given the relatively low resolution and poor quality of the imagery, the lack of metadata, the geometric quality of the Google Earth imagery and the recording characteristics. Land use remained largely similar since 1936, with large parts of the land being used as cropland or extensive grazing areas. Most remarkable changes are the strong expansion of the settlements as well as land management improvements. In a larger effort, ortho-mosaics covering large parts of Ethiopia in 19350900091941 will be produced.

DOI

[26]
Friedman J H, 1991. Multivariate adaptive regression splines.The Annals of Statistics, 19(1): 1-67.

[27]
Friedman J H, Silverman B W, 1989. Flexible parsimonious smoothing and additive modeling.Technometrics, 31(1): 3-21A simple method is presented for fitting regression models that are nonlinear in the explanatory variables. Despite its simplicity—or perhaps because of it—the method has some powerful characteristics that cause it to be competitive with and often superior to more sophisticated techniques, especially for small data sets in the presence of high noise. [ABSTRACT FROM AUTHOR]

DOI

[28]
Fritz S, McCallum I, Schill Cet al., 2009. Geo-Wiki. Org: The use of crowdsourcing to improve global land cover.Remote Sensing, 1(3): 345-354. doi: 10.3390/rs1030345.Global land cover is one of the essential terrestrial baseline datasets available for ecosystem modeling, however uncertainty remains an issue. Tools such as Google Earth offer enormous potential for land cover validation. With an ever increasing amount of very fine spatial resolution images (up to 50 cm 50 cm) available on Google Earth, it is becoming possible for every Internet user (including non remote sensing experts) to distinguish land cover features with a high degree of reliability. Such an approach is inexpensive and allows Internet users from any region of the world to get involved in this global validation exercise. The Geo-Wiki Project is a global network of volunteers who wish to help improve the quality of global land cover maps. Since large differences occur between existing global land cover maps, current ecosystem and land-use science lacks crucial accurate data (e.g., to determine the potential of additional agricultural land available to grow crops in Africa), volunteers are asked to review hotspot maps of global land cover disagreement and determine, based on what they actually see in Google Earth and their local knowledge, if the land cover maps are correct or incorrect. Their input is recorded in a database, along with uploaded photos, to be used in the future for the creation of a new and improved hybrid global land cover map.

DOI

[29]
Gebresamuel G, Singh B R, Dick O, 2010. Land-use changes and their impacts on soil degradation and surface runoff of two catchments of Northern Ethiopia.Acta Agr. Scand. BSP, 60(3): 211-226. doi: 10.1080/ 09064710902821741.Land-use/land-cover changes and their associated impact on environment in the period from 1964 to 2006 was studied in two catchments located in the highland of Tigray using geographic information system and remote-sensing approaches supplemented with field measurements. Results show that, for all periods, cultivated land constitutes the most prevalent (>60%) land-use type and shows a total increase of 1.7 ha y(-1) at Gum Selassa and a decrease of 5.5 ha y(-1) at Maileba. Forest and woodland suffered more damage in both areas losing 32.8 ha (100%) and 53 ha (100%), respectively at Gum Selassa; and 1.74 ha (96.7%) and 52.7 ha (100%) at Maileba over four decades. At Gum Selassa, shrubland decreased by 1.26 ha y(-1) while at Maileba it showed a slight positive increment of 0.38 ha y(-1). Area under settlement increased by a greater magnitude at Maileba (6.3 ha y(-1)) and a slight increase at Gum Selassa (1.4 ha y(-1)) in response to the rapid population increase. These changes in land uses/cover brought significant deleterious impacts on land degradation and surface runoff. The cumulative degradation index (DI) was negative for all land uses, with a higher value under Eucalyptus plantation (DI =-282) followed by cultivated land (DI =-260) at Maileba. Changes in land use/cover also decreased the water-storage capacity of soils by 1.63 and 1.09 mm y(-1) at Gum Selassa and Maileba, respectively, with a corresponding increase in surface runoff by 2.7 and 2.3 mm y(-1). Generally, the observed changes in land degradation and surface runoff are highly linked to the change in land use/land cover.

DOI

[30]
Gebru T, Eshetu Z, Huang Yet al., 2009. Holocene palaeovegetation of the Tigray Plateau in northern Ethiopia from charcoal and stable organic carbon isotopic analyses of gully sediments.Palaeogeogr. Palaeoclimatol. Palaeoecol., 282(1): 67-80. doi: 10.1016/j.palaeo.2009.08.011.A long history of supporting sophisticated but unsustainable kingdoms makes the Tigray Plateau of the northern Ethiopian highlands a promising region for the study of relationships between palaeoenvironmental change and the trajectories of human civilizations. The natural vegetation above 2200m elevation is thought to be forests dominated by Juniperus procera. Nonetheless, this hypothesis is not supported in the vegetation cover now and is scarcely studied in the palaeorecord. To examine changes in vegetation, climate, and land use, we identified buried charcoalized wood and estimated the percentage of organic carbon from C 4 plants (% C 4 carbon) from 未 13C values of bulk organic matter in the soils of gully walls in the boundaries of the ancient Aksumite Empire. Charcoal ranged in age from ca. 13,700 to 110cal yr BP. Juniperus procera occurred in even the youngest samples, although at lower percentages of the total charcoal than in older samples. Nevertheless, rapidly regenerating angiosperms usually dominated or co-dominated charcoal even in some of the oldest strata. A shift towards higher % C 4 carbon and % total organic carbon (%TOC) in soils younger than 3300cal yr BP began during a period when agricultural uses of land may have increased in order to support the needs of growing societies. A shift towards higher % C 4 carbon but lower %TOC began at ca. 6000cal yr BP, however, during a period when no charcoal was found and no changes in human societal complexity are known. These results indicate that juniper forest types have long been present at > 2200m in the Tigray Plateau but that they have rarely been the dominant natural vegetation. Furthermore, lack of repeatable correspondence between factors suggests that the causes of similar shifts in vegetation composition were not always the same.

DOI

[31]
Goldewijk K K, 2001. Estimating global land use change over the 1930s 300 years: The HYDE database.Glob. Biogeochem. Cycles, 15(2): 417-433. doi: 10.1029/1999GB001232.

[32]
Goldewijk K K, Ramankutty N, 2004. Land cover change over the last three centuries due to human activities: The availability of new global data sets.GeoJournal, 61(4): 335-344. doi: 10.1007/s10708-004-5050-z.Land use and land cover change is an important driver of global change (Turner et al., 1993). It is recognized that land use change has important consequences for global and regional climates, the global biogeochemical cycles such as carbon, nitrogen, and water, biodiversity, etc. Nevertheless, there have been relatively few comprehensive studies of global, long-term historical changes in land cover due to land use. In this paper, we review the development of global scale data sets of land use and land cover change. Furthermore, we assess the differences between two recently developed global data sets of historical land cover change due to land use. Based on historical statistical inventories (e.g. census data, tax records, land surveys, historical geography estimates, etc) and applying different spatial analysis techniques, changes in agricultural land cover (croplands, pastures) were reconstructed for the last 300 years. The two data sets indicate that cropland areas expanded from 3-4 million km in 1700 to 15-18 million km in 1990 (mostly at the expense of forests), while grazing land area expanded from 5 million km in 1700 to 31 million km in 1990 (mostly at the expense of natural grasslands). The data sets disagree most over Latin America and Oceania, and agree best over North America. Major differences in the two data sets can be explained by the use of a fractional versus Boolean approach, different modelling assumptions, and inventory data sets.

DOI

[33]
Hadush G, 2012. Modeling of Hydrological Process in the Geba Riv er Basin in the northern Ethiopia [D]. Brussels, Belgium: Vrije Universiteit.

[34]
Haregeweyn N, Fikadu G, Tsunekawa Aet al., 2012. The dynamics of urban expansion and its impacts on land use/land cover change and small-scale farmers living near the urban fringe: A case study of Bahir Dar, Ethiopia.Landscape and Urban Planning, 106(2): 149-157.78 We evaluate urban expansion and its impacts on farmers living near the urban fringe. 78 The urban area expanded by about 32% annually where built-up areas accounting for 75%. 78 The farmers face land expropriation leading to social unrest. 78 Government land ownership system influences implementation of fair money compensation.

DOI

[35]
Hoffer R, 1975. Natural resource mapping in mountainous terrain by computer analysis of ERTS-1 satellite data. Purdue University. LARS Research Bulletin, 919: 124 pp.

[36]
HTS, 1976. Tigrai Rural Development Study, Annex 1. Land and Vegetation Resources. Hunting Technical Services Ltd: Hemel Hempstead.

[37]
Hu Q, Wu W, Xia Tet al., 2013. Exploring the use of Google Earth imagery and object-based methods in land use/cover mapping.Remote Sensing, 5(11): 6026-6042. doi: 10.3390/rs5116026.

[38]
Hurni H, Tato K, Zeleke G, 2005. The implications of changes in population, land use, and land management for surface runoff in the upper Nile basin area of Ethiopia.Mountain Research and Development, 25: 147-154.Much concern has been raised about population increase in the highlands of Ethiopia and its potential to decrease runoff from the upper Nile Basin to the lowland countries of Sudan and Egypt. The present article examines long-term data on population, land use, land management, rainfall, and surface runoff rates from small test plots ($30\ m^{2}$) and micro-catchments (73-673 ha) in the highlands of Ethiopia and Eritrea. Although the data were generated only on small areas, the results of the analyses can nevertheless be used to draw some conclusions relevant to the highland-lowland water controversies that have persisted in this particular region for many decades. The data indicate that there have been no significant trends over the long term in total annual rainfall in the highlands over the past 30-50 years. Nevertheless, test plot surface runoff rates are clearly influenced by land use and soil degradation, and hence by population density and duration of agriculture. In effect there is 5-30 times more surface runoff from cultivated or degraded test plots than from forested test plots. Analysis and interpretation of data support the hypothesis that surface runoff and sediment yield from the Ethiopian and Eritrean highlands into the upper Nile Basin have most probably increased in the long term due to intensified land use and land degradation induced by population increase, when seen in a historical perspective. Rates of base flow, in turn, must have decreased during the same period, but to a much lesser extent, although conclusive empirical evidence cannot be gained from this experimental setting. One can assume that soil and water conservation measures aiming to ensure long-term livelihoods in the humid to sub-humid highlands will, on the one hand, barely affect overall catchment runoff to the downstream areas, though they will considerably reduce surface runoff and soil loss on slopes as well as river sedimentation rates. On the other hand, in a semiarid catchment where intensive soil and water conservation was carried out, reduction in runoff rates was more pronounced. It can be concluded that population increase in the Ethiopian highlands increased overall runoff rates to lowland areas in earlier times, while recent efforts to conserve watersheds might affect total runoff rates in catchments only in semiarid parts, and much less in humid parts of the Ethiopian highlands.

DOI PMID

[39]
Hurni H, Wiesmann U, 2010. Global change and sustainable development: A synthesis of regional experiences from research partnerships. University of Bern. Switzerland. Perspectives of the Swiss National Centre of Competence in Research (NCCR) North-South, University of Bern,Vol. 5. Bern, Switzerland: Geographica Bernensia, 578 pp.

[40]
IGM. Instituto Geografico Militare, Ente Cartografico dello Stato, 2012. www.igmi.org. Accessed 10 January 2016.

[41]
Jacob M, Frankl A, Beeckman Het al., 2015. North Ethiopian Afro-alpine tree line dynamics and forest cover change since the early 20th century.Land Degrad. Dev., 26: 654-664. doi: 10.1002/ldr.2320.Abstract High-altitude forests are very important for local livelihood in the vulnerable environment of the densely populated tropical highlands. Humans need the ecosystem services of the forest and directly impact the forest through livestock herding, fire, and wood harvesting. Nevertheless, temperature-sensitive tree lines in the tropics are scarcely investigated in comparison with higher northern latitudes. In this study, the Erica arborea L. tree line is studied in a tropical mountain in the North Ethiopian highlands: Lib Amba of the Abune Yosef Mountain range (12°04′N, 39°22′E, 399365m asl). The present tree line and forest cover was recorded by high-resolution satellite imagery from Google Maps and field data (2010–2013), while historical forest cover was studied from aerial photographs (1965–1982) and repeat photography (1917–2013). The aerial and satellite images were orthorectified and classified in forest/non-forest binary maps. The binary forest layers were used to detect forest-cover change and tree line dynamics by image differencing between the three time layers (1965–1982–2010). These maps and a terrestrial photograph indicate two periods of deforestation (1917–1965 and 1982–2013), whereas the forest cover was stable between 1965 and 1982. Deforestation was especially severe (with 63%) between 1982 and 2010, associated with a population increase from 77 to 153 inhabitants per square km. There is significant evidence that the elevation of the E. arborea L. tree line increased from 7 to 15 vertical meters between 1965 and 2010, in an area with decreasing anthropozoogenic pressure. Copyright 08 2014 John Wiley & Sons, Ltd.

DOI

[42]
Jacob M, Frankl A, Hurni Het al., 2017. Land cover dynamics in the Simien Mountains (Ethiopia), half a century after establishment of the National Park.Reg. Environ. Change, 17: 777-787. doi: 10.1007/s10113-016- 1070-8.The Simien Mountains house several endangered and endemic wildlife species and provide important ecosystem services. Despite its regional environmental importance, the Simien Mountains are listed as World Heritage in Danger since 1997. This raised the need for an evaluation of landscape changes from before the establishment of the Simien Mountain National Park (SMNP) in 1969. For this purpose, historical terrestrial photographs (1966-2009) were re-analyzed from 2014 repeats, using an expert rating system with eight correspondents. An increase in forest was observed in the eastern and western edge of the SMNP at Sankaber and Imet Gogo (20-40%). In contrast, centrally in the SMNP (around Gich), the area covered with dense forest decreased with an estimated rate of -1.4% per decade. There is no significant effect (p > 0.05) of the boundary of the SMNP on woody vegetation change, because of continued anthropogenic pressure (especially wood cutting and livestock grazing) inside the SMNP. Also elevation and distance to scout camps do not affect rates of change, and however, the density of houses within 2.2 km (a proxy of population pressure) is able to explain 32% of the spatial distribution of woody vegetation decrease (p < 0.05). A subset of six repeated photographs, indicated an uplift of the treeline by more than 1 m year(-1), in areas with low anthropogenic pressure. This is potentially related to increasing (average annual) temperature warming of up to 1.5 degrees C over the past 50 years. Overall, further reduction in anthropogenic pressure is urgent and crucial for recovery of the afro-alpine vegetation and the interrelated endangered wildlife in the Simien Mountains.

DOI

[43]
Kidane G, Dejene A, Malo M, 2010. Agricultural based Livelihood Systems in Drylands in the Context of Climate Change: Inventory of Adaptation Practices and Technologies of Ethiopia. Environment and Natural Resource Working Paper 38, FAO, Rome, 57pp.

[44]
Lambin E F, Geist H J, Lepers E, 2003. Dynamics of land-use and land-cover change in tropical regions.Annu. Rev. Environ. Resour., 28(1): 205-241. doi: 10.1146/annurev.energy.28.050302.105459.

[45]
Lambin E F, Turner B L, Geist H Jet al., 2001. The causes of land-use and land-cover change: Moving beyond the myths.Glob. Environ. Chang. 11(4): 261-269. doi: 10.1016/S0959-3780(01)00007-3.Common understanding of the causes of land-use and land-cover change is dominated by simplifications which, in turn, underlie many environment-development policies. This article tracks some of the major myths on driving forces of land-cover change and proposes alternative pathways of change that are better supported by case study evidence. Cases reviewed support the conclusion that neither population nor poverty alone constitute the sole and major underlying causes of land-cover change worldwide. Rather, peoples responses to economic opportunities, as mediated by institutional factors, drive land-cover changes. Opportunities and constraints for new land uses are created by local as well as national markets and policies. Global forces become the main determinants of land-use change, as they amplify or attenuate local factors.

DOI

[46]
Lanckriet S, Derudder B, Naudts Jet al., 2015. A political ecology perspective of land degradation in the north Ethiopian Highlands.Land Degrad. Dev., 26: 521-530. doi: 10.1002/ldr.2278.Abstract Severe environmental degradation in the north Ethiopian Highlands is among others the result of mismanagement, overpopulation and droughts. However, here, we investigate the linkages of land degradation with the historical dynamics of the political–ecological system and regional land policies. We performed semi-structured interviews with 93 farmers in eight villages in the Tigray region (north Ethiopia) and conceptualised a political–ecological model of land tenure and degradation changes for the region. Results show that different land policies caused and still cause land degradation in several ways. Interviews reveal that the unequal character of land rights during feudal times played an important role in 19th and 20th century land degradation. In particular, poor farmers were forced to construct their farms on marginal terrains, such as steep slopes in dry areas and marshes in cold and humid areas, increasing the catchment water runoff and degradation. The interviews further suggest that after the Derg regime (1974–1991), environmental conservation strategies were successfully implemented at larger scales. Overall, feudal, Derg and contemporary land policies have all had impacts on environmental degradation and have left their fingerprints on the physical landscape of northern Ethiopia. Copyright 08 2014 John Wiley & Sons, Ltd.

DOI

[47]
Lepers E, Lambin E F, Janetos A Cet al., 2005. A synthesis of information on rapid land-cover change for the period 1981-2000.BioScience, 55(2): 115-124. doi: 10.1641/0006-3568.This article presents a synthesis of what is known about areas of rapid land-cover change around the world over the past two decades, based on data compiled from remote sensing and censuses, as well as expert opinion. Asia currently has the greatest concentration of areas of rapid land-cover changes, and dryland degradation in particular. The Amazon basin remains a major hotspot of tropical deforestation. Rapid cropland increase, often associated with large-scale deforestation, is prominent in Southeast Asia. Forest degradation in Siberia, mostly related to logging activities, is increasing rapidly. The southeastern United States and eastern China are experiencing rapid cropland decrease. Existing data do not support the claim that the African Sahel is a desertification hotspot. Many of the most populated and rapidly changing cities are found in the tropics.

DOI

[48]
Loelkes G L, Howard G E, Schwertz E Let al., 1983. Land use/land cover and environmental photointerpretatlon keys.U.S. Geological Survey Bulletin, 1600.

[49]
Maitima J M, Mugatha S M, Reid R S et al., 2009. The linkages between land use change, land degradation and biodiversity across East Africa.LUCID Working Paper, No.42. Nairobi (Kenya): ILRI..

[50]
Meire E, Frankl A, De Wulf Aet al., 2013. Land use/cover dynamics in Africa since the nineteenth century: warped terrestrial photographs of North Ethiopia.Reg. Environ. Chang., 13(3): 717-737. doi: 10.1007/s10113- 012-0347-9.Quantitative research on land use and land cover (LUC) in Africa usually addresses the second half of the twentieth century, by using remote sensing data. Terrestrial photographs, which are available since 1868 in Ethiopia, are seldom used in a quantitative way. This paper presents a methodology that allows to produce land use and land cover (LUC) maps on the basis of old terrestrial photographs. Therefore, land use and land cover was investigated on historical and present-day photographs, and these interpretations were warped to the horizontal plane of the map. The resulting maps allow to gain better insights into LUC changes over a period of 140years. The results show that woody vegetation increased strongly, together with an increase in built-up area. This occurred especially at the expense of bushland. The study validates pervious findings and shows that improved land management strategies in one of the world most degraded areas can lead to environmental rehabilitation.

DOI

[51]
Mengistu D A, Waktola D K, Woldetsadik M, 2012. Detection and analysis of land-use and land-cover changes in the Midwest escarpment of the Ethiopian Rift Valley.J. Land Use Sci., 7(3): 239-260. doi: 10.1080/1747423X. 2011.562556.This study detects patterns of land-use and land-cover changes in the last three decades (19722004) and analyses its causative factors in the Upper Dijo River catchment, Midwest escarpment of Ethiopian Rift Valley. Data captured through the synergy of an aerial photo, satellite image and ground-based socio-economic survey were analysed by GIS and SPSS. The results showed a decline in shrub-grassland and riverine trees at 21.5 and 16.3 ha per year, respectively, and increase in plantation trees, annual crops and bare/open grasslands at 2.8, 12.5 and 24.8 ha per year, respectively. The results are interpreted in the light of population dynamics, socio-economic condition and policy framework. The findings warrant that land-use and land-cover change in the region is very rapid, and unless the existing management of land resources is up-scaled, land ownership is secured, and family planning is mainstreamed in regional and local development plans, the region''s fragile ecological balance would collapse irrecoverably.

DOI

[52]
MEA, 2005. Ecosystem and Human Well-being:Synthesis. Millennium Ecosystem Assessment, 2005. Washington DC: : Island Press.

[53]
Miheretu B A, Yimer A A, 2018. Land use/land cover changes and their environmental implications in the Gelana sub-watershed of Northern Highlands of Ethiopia.Environmental Systems Research, 6(1): 7.Soil erosion in the Ethiopian highlands is considered to be one of the major problems threatening agricultural development and food security in the country. However, knowledge about the forces driving

DOI

[54]
Mitiku H, Herweg K, Stillhardt B, 2006. Sustainable land management: A new approach to soil and water conservation in Ethiopia. Mekelle University, Mekelle, Ethiopia, University of Berne, Berne, Switzerland.

[55]
Munro R N, Deckers J, Haile Met al., 2008. Soil landscapes, land cover change and erosion features of the Central Plateau region of Tigrai, Ethiopia: Photo-monitoring with an interval of 30 years.Catena, 75(1): 55-64. doi: 10.1016/j.catena.2008.04.009.Human land use of the Tigray landscape (north Ethiopia) can be traced back for at least 3000years and is recognizably very complex, but in the past half-century there have been multiple narratives on environmental change in the Northern Ethiopian Highlands in which statements such as he forest and soil resources in Tigray are dwindling at unprecedented rates are common. In an attempt to provide an objective assessment, we made a semi-quantitative analysis of observed changes in the environment of the central Tigray plateau, between 1975 and 2006, and its impact on soil erosion. The first part of this period saw strong degradation, caused by a combination of drought, impoverishment, poor land husbandry and war; but over the whole period intense rehabilitation activities have been high on the agenda. To study these changes, two sets of 51 landscape photographs have been used. The older photo-set was taken in 1975 by R.N. Munro during the Tigrai Rural Development Study; locations were revisited in 2006 by J. Nyssen and colleagues, when a new set of photographs was made at the same locations and with the same aspect. Based on longstanding experience in soil erosion and landscape analysis worldwide and in Ethiopia, the time-lapsed photographs were rated for visible erosion, land cover and protective measures. We present a quantitative evaluation of the change of soil loss by sheet and rill erosion, involving the Universal Soil Loss Equation (USLE) and particularly the changes in the C (cover) and the P (management) factors. This allowed assessing soil loss in 2006 as a percentage of the 1975 situation. Both the landscape and land unit analysis show that the situation for natural resources has improved (and locally strongly improved) since 1974. The rehabilitation is due both to improved vegetation cover and to physical conservation structures. The USLE application indicates that in terms of a whole landscape the current average soil loss would be at around 68% of its 1975 rate. Exceptionally, degradation is still ongoing around Desa'a forest and some other remnant forests, and conservation should be strongly implemented too in these forests. On average, gullies have expanded slightly since 1975, but these incisions appear to have originated in the drought years of the 1980s. This photo-monitoring analysis invalidates hypotheses on (a) irreversibility of land degradation in Tigray; and (b) futility of Soil and Water Conservation (SWC) programmes. The study demonstrates that (a) land management has become an inherent part of the farming system in Tigray, and (b) that the authorities and NGOs are on the right track when promoting SWC.

DOI

[56]
Nyssen J, Frankl A, Mitiku Haile Hurni Het al., 2015. Environmental conditions and human drivers for changes to north Ethiopian mountain landscapes over 145 years.Sci. Total Environ., 485/486: 164-179. doi: 10.1016/j.scitotenv.2014.03.052.61We re-photographed 361 landscapes that appear on historical photographs (1868–1994).61Visible evidence of environmental changes was analyzed through expert rating.61More trees and conservation structures occur where there is high population density.61Direct human impacts on the environment override the effects of climate change.61The northern Ethiopian highlands are greener than at any time in the last 145years.

DOI PMID

[57]
Nyssen J, Haile M, Naudts Jet al., 2009. Desertification? Northern Ethiopia re-photographed after 140 years.Sci. Total Environ., 407(8): 2749-2755. doi: 10.1016/j.scitotenv.2008.12.016.A collection of sepia photographs, taken during Great Britain's military expedition to Abyssinia in 1868, are the oldest landscape photographs from northern Ethiopia, and have been used to compare the status of vegetation and land management 140 years ago with that of contemporary times. Thirteen repeat landscape photographs, taken during the dry seasons of 1868 and 2008, were analyzed for various environmental indicators and show a significant improvement of vegetation cover. New eucalypt woodlands, introduced since the 1950s are visible and have provided a valuable alternative for house construction and fuel-wood, but more importantly there has also been locally important natural regeneration of indigenous trees and shrubs. The situation in respect to soil and water conservation measures in farmlands has also improved. According to both historical information and measured climatic data, rainfall conditions around 1868 and in the late 19th century were similar to those of the late 20th/early 21st century. Furthermore, despite a ten-fold increase in population density, land rehabilitation has been accomplished over extensive areas by large-scale implementation of reforestation and terracing activities, especially in the last two decades. In some cases repeat photography shows however that riparian vegetation has been washed away. This is related to river widening in recent degradation periods, particularly in the 1970s 1980s. More recently, riverbeds have become stabilized, and indicate a decreased runoff response. Environmental recovery programmes could not heal all scars, but this study shows that overall there has been a remarkable recovery of vegetation and also improved soil protection over the last 140years, thereby invalidating hypotheses of the irreversibility of land degradation in semi-arid areas. In a highly degraded environment with high pressure on the land, rural communities were left with no alternative but to improve land husbandry: in northern Ethiopia such interventions have been demonstrably successful.

DOI PMID

[58]
Nyssen J, Naudts J, De Geyndt Ket al., 2008a. Soils and land use in the Tigray highlands (Northern Ethiopia).Land Degrad. Dev., 19(3): 257-274. doi: 10.1002/ldr.840.Land use in a 208 ha representative catchment in the Tigray Highlands, Dogu'a Tembien district in Northern Ethiopia was studied in relation to soil geography. Typical soils are Vertisols, Vertic Cambisols, Cumulic Regosols, Calcaric Regosols and Phaeozems. Patterns of land use vary greatly within the catchment and results from 2-tests showed strong associations ( p < 0001) between soil type and land use and crop production system. There is a strong association between cropland and colluvium high in basaltic content because the most fertile soils, such as Vertisols and Vertic Cambisols, have developed on this material. Preference is for autochthonous soils on in situ parent material, irrespective of the rock type, to be put under rangeland. Land use by smallholders in Dogu'a Tembien appears to be the result primarily of the interaction between environmental and social factors. Copyright 2007 John Wiley & Sons, Ltd.

DOI

[59]
Nyssen J, Petrie G, Mohamed Set al., 2016. Recovery of the aerial photographs of Ethiopia in the 1930s.J. Cult. Herit., 17: 170-178. doi: 10.1016/j.culher.2015.07.010.The aerial photographs (APs) acquired by the Istituto Geografico Militare (IGM) in the period of the Italian occupation of Ethiopia (1935–1941) have recently been discovered, scanned and organised. Until recently, the oldest APs of the country that were available had been taken in the period 1958–1964. The APs over Ethiopia in 1935–1941 consist of 8281 assemblages on approximately 50cm×20cm hardboard tiles, each holding a label, one nadir-pointing photograph flanked by two low-oblique photographs and one high-oblique photograph. The four APs were exposed simultaneously and were taken across the flight line. The high-oblique photograph is presented alternatively at left and at right. There is approximately 60% overlap between subsequent sets of APs. One of Santoni's glass plate multi-cameras was used, with focal length of 178mm and with a flight height of 4000–4500m a.s.l., which resulted in an approximate scale of 1:11,500 for the central photograph and 1:16,000 to 1:18,000 for the low-oblique APs. The surveyors oriented themselves with maps of Ethiopia at 1:400,000 scale, compiled in 1934. The flights present a dense AP coverage of Northern Ethiopia, where they were acquired in the context of upcoming battles with the Ethiopian army. Several flights preceded the later advance of the Italian army southwards to the capital Addis Ababa. Further flights took place in central Ethiopia for civilian purposes. As of 1936, the APs were used to prepare topographic maps at 1:100,000 and 1:50,000 scales. To re-process the imagery using novel techniques, procedures using digital image-based modelling have been developed. The 1935-1941 APs together with those of 1958–1964, 1994 and recent high-resolution satellite imagery are currently being used in spatio-temporal analysis, including change studies of land cover, land management and geomorphology in Ethiopia over a time span of 80years.

DOI

[60]
Nyssen J, Poesen J, Descheemaeker Ket al., 2008b. Effects of region-wide soil and water conservation in semi-arid areas: The case of northern Ethiopia.Z. Geomorphol., 52(3): 291-315. doi: 10.1127/ 0372-8854/ 2008/0052-0291.Studies on the impacts of environmental rehabilitation in semi-arid areas are often conducted over limited space and time scales, and do typically not include detailed biophysical components. This study makes a multi-scale assessment over a time span of 30 years of environmental rehabilitation in one of the world most degraded areas: the Tigray highlands of Northern Ethiopia. The study shows that in Tigray sheet and rill erosion rates have decreased by approximately 68%, infiltration and spring discharge are enhanced and vegetation cover has improved. These impacts are evidenced and quantified by a comprehensive comparison of the current landscape with a coverage of 30-year old photographs and substantiated by field investigations. The positive changes in ecosystem service supply that result from these conservation activities in the Tigray highlands are an issue of global concern.

DOI

[61]
Nyssen J, Vandenreyken H, Poesen Jet al., 2005. Rainfall erosivity and variability in the Northern Ethiopian Highlands.J. Hydrol., 311(1): 172-187. doi: 10.1016/j.jhydrol.2004.12.016.The Ethiopian Highlands are subjected to important land degradation. Though spatial variability of rain depth is important, even at the catchment scale, this variability has never been studied. In addition, little is known on rain erosivity for this part of the world. The objectives of this study are (a) to assess the spatial variation of rain in a 80 km 2 mountain area (2100–2800 m a.s.l.) in the Northern Tigray region, and how this variation is influenced by topography, geographical position and lithology, (b) to analyse the temporal variations and (c) to quantify rain erosivity and the different factors determining it, such as rain intensity, drop size and kinetic energy. Spatial variation of rain was measured over a 6-y period by installing 16 rain gauges in the study area. Topographical factors, especially general orientation of the valley and slope gradient over longer distances, determine the spatial distribution of annual rain, which is in the order of 700 mm y 611. Precipitation is highest nearby cliffs and other eminent slopes, perpendicular to the main valleys which are preferred flow paths for the air masses. Rain intensity is smaller than expected: 88% falls with an intensity <30 mm h 611. High intensities have a short duration; maximum recorded rain depth over 1 h (32 mm) is only 2 mm less than that over 24 h. Using the blotting paper method 65,100 rain drops were sampled. For all observed rain intensities, the median volume drop diameters ( D 50) are significantly larger than those reported for other regions of the world. A relation between rain intensity ( I) and volume specific kinetic energy (Ek vol) was developed for the Ethiopian Highlands: Ek vol = 36.65 ( 1 61 ( 0.6 / I ) ) ( R 2 = 0.99 , n = 18 ) , ( Ek vol in J m 61 2 mm 61 1 , I in mm h 61 1 ) .Due to the occurrence of large drop sizes, probably linked to the prevailing semi-arid to subhumid mountain climate, this relation yields, within the intensity range [0.6–84 mm h 611], larger values for Ek vol than elsewhere in the world. It is recommended to use this new relationship for calculating Ek vol of rain in the Ethiopian Highlands, as well as for the computation of Universal Soil Loss Equation's rain erosivity factor on yearly basis.

DOI

[62]
Pankhurst R, 1995. The history of deforestation and afforestation in Ethiopia prior to World War I.Northeast African Studies, 2(1): 119-133.Based on historical records, mostly travellers' accounts, this paper examines deforestation in Ethiopia over the centuries. Major deforestation occurred throughout Ethiopian history, though this is more easily documentable for the 19th and 20th centuries. The countryside, much of which is believed to have once been covered with trees, became progressively barer, as forests were steadily cut or burned down. Deforestation, however, took place mainly in areas of extensive settlement, and especially in the vicinity of towns or moving capitals, though Gond盲r, Adwa and most urban centres of the past were the site of many 'wanza' ('Cordia abyssinica') and other local trees, and therefore relatively well-forested. In the late 19th and early 20th centuries, the advent of the eucalyptus tree turned Addis Ababa and other settlements into forest towns, albeit ones still by no means copiously supplied with wood. Although the eucalyptus was not without its disadvantages (it was in particular so thirsty a plant that it dried up rivers, lakes and wells in its vicinity) and there was considerable opposition to the tree, its advantages were too great for it ever to be eradicated, but it was an essentially urban tree and the country at large remained largely deforestated. Note, ref.

DOI

[63]
Quiros E, Felicísimo A M, Cuartero A, 2009. Testing multivariate adaptive regression splines (MARS) as a method of land cover classification of TERRA-ASTER satellite images.Sensors, 9(11): 9011-9028. doi: 10.3390/s91109011.This work proposes a new method to classify multi-spectral satellite images based on multivariate adaptive regression splines (MARS) and compares this classification system with the more common parallelepiped and maximum likelihood (ML) methods. We apply the classification methods to the land cover classification of a test zone located in southwestern Spain. The basis of the MARS method and its associated procedures are explained in detail, and the area under the ROC curve (AUC) is compared for the three methods. The results show that the MARS method provides better results than the parallelepiped method in all cases, and it provides better results than the maximum likelihood method in 13 cases out of 17. These results demonstrate that the MARS method can be used in isolation or in combination with other methods to improve the accuracy of soil cover classification. The improvement is statistically significant according to the Wilcoxon signed rank test.

DOI PMID

[64]
Ramankutty N, Foley J A, 1999. Estimating historical changes in global land cover: Croplands from 1700 to 1992.Glob. Biogeochem. Cycles., 13(4): 997-1027. doi: 10.1029/1999GB900046.Human activities over the last three centuries have significantly transformed the Earth's environment, primarily through the conversion of natural ecosystems to agriculture. This study presents a simple approach to derive geographically explicit changes in global croplands from 1700 to 1992. By calibrating a remotely sensed land cover classification data set against cropland inventory data, we derived a global representation of permanent croplands in 1992, at 5 min spatial resolution [Ramankutty and Foley, 1998]. To reconstruct historical croplands, we first compile an extensive database of historical cropland inventory data, at the national and subnational level, from a variety of sources. Then we use our 1992 cropland data within a simple land cover change model, along with the historical inventory data, to reconstruct global 5 min resolution data on permanent cropland areas from 1992 back to 1700. The reconstructed changes in historical croplands are consistent with the history of human settlement and patterns of economic development. By overlaying our historical cropland data set over a newly derived potential vegetation data set, we analyze our results in terms of the extent to which different natural vegetation types have been converted for agriculture. We further examine the extent to which croplands have been abandoned in different parts of the world. Our data sets could be used within global climate models and global ecosystem models to understand the impacts of land cover change on climate and on the cycling of carbon and water. Such an analysis is a crucial aid to sharpen our thinking about a sustainable future.

DOI

[65]
Ramankutty N, Foley J A, Olejniczak N J, 2002. People on the land: Changes in global population and croplands during the 20th century.Ambio, 31(3): 251-257.This study reviews the major changes in global distribution of croplands during the 20th century. During the 20th century, the cropland base diminished greatly (from 65 0.75 ha$\text{person}^{-1}$in 1900 to 65 0.35 ha$\text{person}^{-1}$in 1990). This loss of croplands was not globally uniform: more than half the world's population, living in developing nations, lost nearly two-thirds of their per capita cropland base. The distribution of croplands has become increasingly skewed -in 1990, 80% of the population lived off less than 0.35 ha$\text{person}^{-1}$. While agricultural yields have generally increased, they have barely kept pace with population growth in developing nations. Overall, the global food production system is becoming increasingly vulnerable to regional disruptions because of our increasing reliance on expensive technological options to increase agricultural production, or on global food trade.

DOI PMID

[66]
Reid R S, Kruska R L, Muthui Net al., 2000. Land-use and land-cover dynamics in response to changes in climatic, biological and socio-political forces: The case of southwestern Ethiopia.Landsc. Ecol., 15(4): 339-355. doi: 10.1023/A:1008177712995.Few studies of land-use/land-cover change provide an integrated assessment of the driving forces and consequences of that change, particularly in Africa. Our objectives were to determine how driving forces at different scales change over time, how these forces affect the dynamics and patterns of land use/land cover, and how land-use/land-cover change affects ecological properties at the landscape scale. To accomplish these objectives, we first developed a way to identify the causes and consequences of change at a landscape scale by integrating tools from ecology and the social sciences and then applied these methods to a case study in Ghibe Valley, southwestern Ethiopia. Maps of land-use/land-cover change were created from aerial photography and Landsat TM imagery for the period, 1957–1993. A method called `ecological time lines' was developed to elicit landscape-scale explanations for changes from long-term residents. Cropland expanded at twice the speed recently (1987–1993) than two decades ago (1957–1973), but also contracted rapidly between 1973–1987. Rapid land-use/land cover change was caused by the combined effects of drought and migration, changes in settlement and land tenure policy, and changes in the severity of the livestock disease, trypanosomosis, which is transmitted by the tsetse fly. The scale of the causes and consequences of land-use/land-cover change varied from local to sub-national (regional) to international and the links between causes and consequences crossed scales. At the landscape scale, each cause affected the location and pattern of land use/land cover differently. The contraction of cropland increased grass biomass and cover, woody plant cover, the frequency and extent of savanna burning, and the abundance of wildlife. With recent control of the tsetse fly, these ecological changes are being reversed. These complex patterns are discussed in the context of scaling issues and current conceptual models of land-use/land-cover change.

DOI

[67]
Rembold F, Carnicelli S, Nori Met al., 2000. Use of aerial photographs, Landsat TM imagery and multidisciplinary field survey for land-cover change analysis in the lakes region (Ethiopia).International Journal of Applied Earth Observation and Geoinformation, 2(3/4): 181-189.Early multidisciplinary surveys in the Lakes region of central/south Ethiopia show a highly variable land cover pattern characterised by complex interactions between environmental parameters and socio-economic dynamics. From an ecological point of view the area is highly sensitive and both food security and soil conservation are becoming serious problems for the rapidly growing population. The intensive land cover changes observed in this area during the last few decades beg accurate analysis. Land-cover change analysis over a long time-span was performed. Interpretation (API) of aerial photographs dated 1972 and classification of a 1994 Landsat TM image were used. Problems due to the heterogeneous nature of the data were overcome with a method for quantifying land cover on aerial photographs, thus producing data comparable to TM classification results. As land cover is linked, through land use, to social dynamics, in ground control use was made of the results of parallel socio-economic investigations. From the analysis, a general trend of increase in cultivated surfaces was noted. Unique strategies of land allocation according to physical settings were observed. A trend in the evolution of badlands was identified: rapid reactivation of previous erosion in newly cropped areas occurred; within a few decades this erosion reached quasi-equilibrium. The methods adopted showed some accuracy limitations, but allowed land-cover change analysis over a 22-year time-span, providing important insight into recent phenomena and present trends.

DOI

[68]
Romanov P, Tarpley D, Gutman Get al., 2003. Mapping and monitoring of the snow cover fraction over North America.J. Geophys. Res., 108(D16): 8619. doi: 10.1029/2002JD003142.1] Automated snow maps over North America have been produced at the National Environmental Satellite Data and Information Service (NESDIS) of the National Oceanic and Atmospheric Administration (NOAA) since 1999. The developed snow-mapping system is based on observations in the visible, middle infrared, infrared, and microwave spectral bands from operational geostationary and polar orbiting meteorological satellites and generates daily maps of snow cover at a spatial resolution of 4 km. Recently, the existing snow-mapping technique was extended to derive the fractional snow cover. To obtain snow fraction, we use measurements of the Imager instrument on board Geostationary Operational Environmental Satellite (GOES). The algorithm treats every cloud-clear image pixel as a 090008mixed scene090009 consisting of a combination of snow-covered and snow-free land surface. To determine the portion of the pixel that is covered with snow, we employ a linear mixture approach, which relies on the Imager measurements in the visible spectral band. The estimated accuracy of subpixel snow fraction retrievals is about 10%. In this paper, we present a description of the snow cover and snow fraction mapping algorithms. Application of the developed algorithms over North America for three winter seasons from 19990900092000 to 20010900092002 has shown that the spatial distribution of the fractional snow cover over areas affected by seasonal snow closely corresponds to the distribution of the forest cover. The fraction of snow in the middle of the winter season generally varied from 100% over croplands, grasslands, and other nonforested areas to 2009000930% over dense boreal forests. The snow fraction over dense boreal forests exhibited a slight intraseason variability; however, no obvious correlation of these changes with snowfalls was noticed. Over areas with no or sparse tree vegetation cover (croplands, grasslands), snow fraction showed a noticeable correlation with snow depth for snow depths up to 3509000940 cm.

DOI

[69]
Saebo H V, 1983. Land Use and Environmental statistics obtained by point sampling. Central Bureau of Statistics of Norway. Artikler 144: 35pp.

[70]
Scanlon B R, Reedy R C, Tachovsky J A, 2007. Semiarid unsaturated zone chloride profiles: Archives of past land use change impacts on water resources in the southern High Plains, United States.Water Resour. Res., 43(W06423): 1-13. doi: 10.1029/2006WR005769.Unsaturated zone chloride profiles in semiarid regions provide a decadal- to century-scale record of past environmental changes, similar to climate change records provided by tree rings and ice cores. Impacts of conversions from natural ecosystems to rain-fed agriculture on water resources are recorded in chloride profiles in semiarid regions, as typified by the southern High Plains (SHP), Texas, southwestern United States. Large chloride accumulations beneath natural grassland and shrubland ecosystems (3 profiles) reflect evapotranspirative enrichment of atmospherically derived chloride during the Holocene, indicating no recharge in interdrainage areas. Conversion to rain-fed agriculture is recorded by downward displacement (9 profiles) or complete flushing (10 profiles) of chloride bulges, indicating increased recharge. Increased recharge associated with cultivation (median 24 mm/yr, 5% of precipitation, 19 profiles) was quantified using chloride mass balance calculations. The timing of land use change was estimated using chloride data, and results (43-89 years) are consistent with aerial photo records and landowner surveys. New equilibrium volumetric recharge rates beneath rain-fed agriculture in the SHP (0.63 km/yr) will require decades to establish and represent one to eight times recharge rates for baseline precultivated conditions that are focused beneath ephemeral lake or playa drainages (0.08-0.83 km/yr). These chloride profiles generally represent decadal-scale monitoring of subsurface response to land use change.

DOI

[71]
Shockey W R, 1969. Point sampling of land use in the Washita Basin, United States Department of Agriculture, Agricultural research service. Archived Documents, 41-149.

[72]
Stehman S V, 1996. Estimating the Kappa Coefficient and its variance under stratified random sampling.Photogrammetric Engineering & Remote Sensing, 62(4): 401-402.

[73]
Tadesse G, Zavaleta E, Shennan Cet al., 2014. Policy and demographic factors shape deforestation patterns and socio-ecological processes in southwest Ethiopian coffee agroecosystems.Applied Geography, 54: 149-159. doi: 10.1016/j.apgeog.2014.08.001.61We used remote sensing and surveys to assess deforestation in southwest Ethiopia.61Geospatial analysis and local perceptions corroborate rapid deforestation.61Deforestation rates varied in space and time.61Temporal variations overlapped with land-tenure and agricultural development policies.61Regional variations correlated with demographic factors and livelihood practices.

DOI

[74]
Taye G, Poesen J, Wesemael B Vet al., 2013. Effects of land use, slope gradient, and soil and water conservation structures on runoff and soil loss in semi-arid Northern Ethiopia.Physical Geography, 34(3): 236-259.Land degradation and recurrent drought are the major threats to rain-fed agriculture in the semi-arid Ethiopian highlands. Water harvesting has become a priority in the Tigray region since 1990. However, the success of water harvesting in reservoirs is limited due to reduced inflow. The aim of this study was to investigate the effects of typical land-use types, slope gradients, and different soil and water conservation (SWC) structures on runoff and soil loss at the runoff-plot scale. Six runoff measuring sites, corresponding to three slope gradients, were established for cropland (cultivated land for annual crop production) and rangeland (heavily grazed land on hillslopes with high rock-fragment cover) at Mayleba catchment in Tigray, Ethiopia. SWC structures tested were stone bunds, trenches, and stone bunds with trenches, in addition to control plots. In total, 21 large runoff plots (with lengths of 60 to 100 m) were monitored daily for runoff production and soil loss during the main rainy season (July090009September) in 2010. The results show that the seasonal runoff coefficient (RCs) representing the fraction of rainfall measured as runoff was much higher for rangeland (0.38 < RCs < 0.50) compared to that for cropland (0.11 < RCS < 0.15). Seasonal soil loss (SLs) values were five to six times larger on rangeland (28.6 < SLs < 50.0 ton ha0908081) compared to that for cropland (4.6 < SLs < 11.4 ton ha0908081). Stone bunds with trenches were the most effective SWC structures in reducing runoff and soil loss. With the same SWC structures installed, RCs and SLs for both rangeland and cropland tend to decrease with increasing slope gradient mainly due to a corresponding increase in rock-fragment cover. The effects of SWC structures on runoff production and soil loss are considerable; hence, it is crucial to consider these effects for optimal design of water-harvesting schemes such as micro-dams that collect and store surface runoff for irrigation development in the Ethiopian highlands.

DOI

[75]
Tayyebi A, Pijanowski B C, 2014. Modeling multiple land use changes using ANN, CART and MARS: Comparing tradeoffs in goodness of fit and explanatory power of data mining tools.Int. J. Appl. Earth Obs. Geoinf., 28: 102-116. doi: 10.1016/j.jag.2013.11.008.Over half of the earth's terrestrial surface has been modified by humans. This modification is called land use change and its pattern is known to occur in a non-linear way. The land use change modeling community can advance its models using data mining tools. Here, we present three data mining land use change models, one based on Artificial Neural Network (ANN), another on Classification And Regression Trees (CART) and another Multivariate Adaptive Regression Splines (MARS). We reconfigured the three data mining models to concurrently simulate multiple land use classes (e.g. agriculture, forest and urban) in South-Eastern Wisconsin (SEWI), USA (time interval 1990 2000) and in Muskegon River Watershed (MRW), Michigan, USA (time interval 1978 1998). We compared the results of the three data mining tools using relative operating characteristic (ROC) and percent correct match (PCM). We found that ANN provided the best accuracy in both areas for three land use classes (e.g. urban, agriculture and forest). In addition, in both regions, CART and MARS both showed that forest gain occurred in areas close to current forests, agriculture patches and away from roads. Urban increased in areas of high urban density, close to roads and in areas with few forests and wetlands. We also found that agriculture gain is more likely for the areas closer to the agriculture and forest patches. Elevation strongly influenced urbanization and forest gain in MRW while it has no effect in SEWI.

DOI

[76]
Teka K, Van Rompaey A, Poesen J, 2013. Assessing the role of policies on land use change and agricultural development since 1960s in northern Ethiopia.Land Use Policy, 30(1): 944-951. doi: 10.1016/j.landusepol. 2012.07.005.Policy has long been considered as one of the major driving forces for land use change and agricultural development. However, a designated and in-depth study on its interaction with land use change and agricultural development is still very limited. The authors, therefore, attempted to address this issue by using five villages representing three agro-ecologies (highland, midland and lowland) for the period between 1965 and 2007. The study indicated that land policies of the imperial and communist regimes contributed largely to the increase of arable land while a decrease in vegetated land. This is, however, reversed in the EPRDF regime. Land productivity/crop harvest (t/ha) and herd size per household have declined. Agricultural policies played active roles in the change of water area and indirectly contributed to the change of construction land.

DOI

[77]
Teka K, Van Rompaey A, Poesen Jet al., 2015. Spatial analysis of land cover changes in eastern Tigray (Ethiopia) from 1965 to 2007: Are there signs of a forest transition?Land Degrad. Dev., 26(7): 680-689. doi: 10.1002/ldr.2275.Abstract This paper examines whether eastern Tigray is still in a phase of land degradation or if a trend of landscape greening has started. Hitherto, the land cover in five representative Tigray villages was mapped for three different periods: 1965, 1994, and 2007. These maps were based on the interpretation of aerial photographs and high-resolution satellite imagery in combination with field validation. The results show a strong decrease of the forest and bush land in favor of arable land and rangeland from 1965 to 1994. This trend is, however, slowed down and even reversed from 1994 onwards whereby some of the arable land and rangeland are abandoned allowing shrubs and bushes to regenerate. Household interviews and census data showed that the rural population number is still increasing. However, the productivity of the farming activities did not show a significant increase. The observed abandonment of marginal farm and rangeland is, therefore, made possible only through food aid and the import of food from other regions. Furthermore, policymakers stimulate land abandonment and landscape greening by establishing exclosures. Copyright 2014 John Wiley & Sons, Ltd.

DOI

[78]
Tesfamichael G, De Smedt F, Miruts Het al., 2010. Large-scale geological mapping of the Geba basin, northern Ethiopia. Tigray Livelihood Paper No 9, VLIR - Mekelle University IUC Program, 46pp. Tesfaye C, Gebretsadik E, 1982. Hydrogeology of Mekelle Aea (ND37-11).Ministry of Mines and Energy, Ethiopian Institute of Geological Survey. The Provisional Military Government of Socialist Ethiopia. Memoir No.2, Addi Ababa, Ethiopia,, pp50.

[79]
Tielens S, 2012. Towards a Soil Map of the Geba Catchment using benchmark soils. Dissertation, K.U. Leuven, Belgium, pp226.

[80]
Tilahun K, 2006. Analysis of rainfall climate and evapo-transpiration in arid and semi-arid regions of Ethiopia using data over the last half a century.J. Arid Environ., 64(3): 474-487. doi: 10.1016/j.jaridenv.2005.06.013.Rainfall and evapo-transpiration are the two major climatic factors affecting agricultural production. While rainfall can be directly measured, evapo-transpiration is estimated from weather data. In this study reference evapo-transpiration ET o was estimated using Penman–Monteith equation, under full data and limited data availability conditions, and Hargraves method. Monthly rainfall and evapotarnspiration were plotted and compared in order to determine moisture deficit periods for several stations in arid and semi-arid parts of the country. Rainfall and reference evapo-transpiration at different probability levels were compared. Temporal variation of annual rainfall was analysed using variability indices: cumulative departure index, standard climate departure index and rainfall anomaly index. It was found that when there is only limited data, it is better to estimate ET o using Penman–Monteith method than using simplified methods such as Hargraves. ET o showed relatively low variation with time while year-to-year rainfall variability was very high. From these plots information on rainfall pattern over the past 50 years, such as drought years was obtained.

DOI

[81]
Tsegaye D, Moe S R, Vedeld Pet al., 2010. Land-use/cover dynamics in Northern Afar rangelands, Ethiopia.Agriculture, Ecosystems & Environment, 139(1): 174-180.This study uses a combination of remote sensing data, field observations and information from local people to analyze the patterns and dynamics of land-use/cover changes for 35 years from 1972 to 2007 in the arid and semi-arid Northern Afar rangelands, Ethiopia. A pixel-based supervised image classification was used to map land-use/cover classes. People's perceptions and ecological time-lines were used to explain the driving forces linked to the changes. A rapid reduction in woodland cover (97%) and grassland cover (88%) took place between 1972 and 2007. Bushland cover increased more than threefold, while the size of cultivated land increased more than eightfold. Bare land increased moderately, whereas bushy grassland and scrubland remained stable. According to accounts from local people, major events that largely explain the changes include: (1) severe droughts in 1973/74 and 1984/85; (2) increase in dry years during the last decade; and (3) immigration and increased sedentarization of pastoralists. If the present land-use/cover change were to continue, coupled with a drier climate, people's livelihoods will be highly affected and the pastoral production system will be under increasing threat.

DOI

[82]
Turner B, Meyer W B, Skole D L, 1994. Global land-use/land-cover change: Towards an integrated study.Ambio Stockholm, 23(1): 91-95.Human actions are altering the terrestrial environment at unprecedented rates, magnitudes, and spatial scales. Land-cover change stemming from human land uses represents a major source and a major element of global environmental change. Not only are the global-level data on land-use and land-cover change relatively poor, but we need a much better understanding of the underlying driving forces for these changes. Many forces have been proposed as significant, but single-factor explanations of land transformation have proved to be inadequate. How the human causes interact, and under what circumstances each is important, are questions needing systematic research. An international and interdisciplinary agenda is currently being developed to address these issues, through several closely-connected foci of study. A division of the world according to common situations of environment, human driving forces, and land-cover dynamics will be followed by detailed study of the processes at work within each situation. The results will form the basis for a concurrent effort to develop a global land model that can offer projections of patterns of land transformation.

DOI

[83]
USGS, 2016. Shuttle Radar Topography Mission (SRTM)(1-arc second) documentation. 2016.

[84]
Van de Wauw J, Baert G, Moeyersons Jet al., 2008. Soil-landscape relationships in the basalt-dominated highlands of Tigray, Ethiopia.Catena, 75(1): 117-127. doi: 10.1016/j.catena.2008.04.006.Though knowledge about the distribution and properties of soils is a key issue to support sustainable land management, existing knowledge of the soils in Tigray (Northern Ethiopian Highlands) is limited to either maps with a small scale or with a small scope. The goal of this study is to establish a model that explains the spatial soil variability found in the May-Leiba catchment, and to open the scope for extrapolating this information to the surrounding basalt-dominated uplands. A semi-detailed (scale: 1/40 000) soil survey was conducted in the catchment. Profile pits were described and subjected to physico-chemical analysis, and augerings were conducted. This information was combined with information from aerial photographs and geological and geomorphologic observations. The main driving factors that define the variability in soil types found were: 1) geology, through soil parent material and the occurrence of harder layers, often acting as aquitards or aquicludes; 2) different types of mass movements that occupy large areas of the catchment; and 3) severe human-induced soil erosion and deposition. These factors lead to “red-black” Skeletic Cambisol–Pellic Vertisol catenas on basalt and Calcaric Regosol–Colluvic Calcaric Cambisols–Calcaric Vertisol catenas on limestone. The driving factors can be derived from aerial photographs. This creates the possibility to extrapolate information and predict the soil distribution in nearby regions with a comparable geology. A model was elaborated, which enables the user to predict soil types, using topography, geomorphology, geology and soil colours, all of which can be derived from aerial photographs. This derived model was later applied to other catchments and validated in the field.

DOI

[85]
Vancampenhout K, Nyssen J, Gebremichael Det al., 2006. Stone bunds for soil conservation in the northern Ethiopian Highlands: Impacts on soil fertility and crop yield.Soil and Tillage Research, 90(1): 1-15.In the Ethiopian highlands, large-scale stone bund building programs are implemented to curb severe soil erosion. Development of soil fertility gradients is often mentioned as the major drawback of stone bund implementation, as it would result in a dramatic lowering of crop yield. Therefore, the objectives of this study are to assess soil fertility gradients on progressive terraces and their influence on crop yield, in order to evaluate the long-term sustainability of stone bunds in the Ethiopian Highlands. The study was performed near Hagere Selam, Tigray and comprises (i) measurement of P av, N tot and C org along the slope on 20 representative plots and (ii) crop response measurement on 143 plots. Results indicate that levels of P av, N tot and C org in the plough layer are highly variable between plots and mainly determined by small-scale soil and environmental features, plot history and management. After correcting for this “plot effect” a significant relationship ( p < 0.01) was found between the position in the plot relative to the stone bund and levels of P av and N tot, which are higher near the lower stone bund, especially on limestone parent material. For C org and on basalt-derived soils in general no significant relationship was found. Although soil fertility gradients are present, they are not problematic and can be compensated by adapted soil management. Only in areas where a Calcaric or Calcic horizon is present at shallow depth, care should be taken. Crop Yields increased by 7% compared to the situation without stone bunds, if a land occupation of 8% by the structures is accounted for. Yield increased from 632 to 683 kg ha 611 for cereals, from 501 to 556 kg ha 611 (11%) for Eragrostis tef and from 335 to 351 kg ha 611 for Cicer arietinum. No negative effects reducing stone-bund sustainability were found in this study. Soil erosion on the other hand, poses a major threat to agricultural productivity. Stone bund implementation therefore is of vital importance in fighting desertification and establishing sustainable agriculture in the Ethiopian highlands.

DOI

[86]
Vanmaercke M, Poesen J, Broeckx Jet al., 2014. Sediment yield in Africa.Earth-Sci. Rev., 136: 350-368. doi: 10.1016/j.earscirev.2014.06.004.

[87]
Virgo K, Munro R, 1978. Soil and erosion features of the Central Plateau region of Tigrai, Ethiopia.Geoderma, 20(2): 131-157. doi: 10.1016/0016-7061(78)90040-X.The results of reconnaissance soil surveys covering 6,000 km 2 are used to describe the Central Plateau region, which lies at elevations of 2,000 to 2,800 m in northern Ethiopia. Landform and soil sequences on calcareous shales, dolerites and sandstones are described, in which the principal soil units are Lithosols, Luvisols, Cambisols, Arenosols and Vertisols. Detailed morphological and analytical data are presented for a profile representative of arable soils in each sequence. Small-scale subsistence cultivation of cereals is the dominant land use; all land which is physically cultivable is at present cultivated. Settlement patterns are closely related to soil type, nucleated settlement occurring on fine textured soils but dispersed settlement on coarser textured and more freely draining soils. Erosion and soil moisture features of the three landforms described were investigated and compared. Empirical methods and suspended sediment measurements indicate high rates of regional soil loss (17–33 t ha 611 yr 611), accounted for by seasonally high rates of rainfall erosivity, steep terrain and poor land use. The recent development of gully erosion is seen to be linked to the disintegration of waterfall tufas. Application of the universal soil loss equation to arable lands indicates potential annual soil losses in the range of 400 t ha 611 on Vertisols to 200 t ha 611 on Cambisols: differences in rates are ascribed principally to differences in crop planting dates, which affect the degree of vegetative protection during periods of high rainfall erosivity. Soil moisture is shown to be in the available range for less than three months in the year. The time at which moisture in the profile enters the available range differed between the three soils monitored and was found to be closely related to the crop planting date, thus indirectly affecting the erosion hazard.

DOI

[88]
Woien H, 1995. Deforestation, information and citations.GeoJournal, 37(4): 501-511. doi: 10.1007/BF00806939.

[89]
WBISPP (Woody Biomass Inventory and Strategic Planning Project), 2003. Tigray Regional State: A strategic plan for the sustainable development, conservation, and management of the woody biomass resources. Final Report, Mekelle, Ethiopia.

[90]
Zeimetz K A, Dillon E, Hardy E Eet al., 1976. Using area point samples and airphotos to estimate land use change.Agricultural Economics Research, 28(2): 65-74.

[91]
Zeleke G, Hurni H, 2001. Implications of land use and land cover dynamics for mountain resource degradation in the Northwestern Ethiopian Highlands.Mountain Research and Development, 21(2): 184-191.Land use and land cover changes that occurred from 1957 to 1995 in the Dembecha area, Gojam, in the Northwestern highlands of Ethiopia, were monitored using a geographic information system (GIS) and a remote sensing approach with field verification. The study area covers 27,100 ha and is representative of Gojam, which is known for its cereal production and export of surplus to major cities of the country. However, given the age-old tradition of clearing increasingly steeper land for cultivation and the lack of appropriate land use policies, productivity is currently heavily threatened by soil degradation. The results show that the natural forest cover declined from 27% in 1957 to 2% in 1982 and 0.3% in 1995. The total natural forest cleared between 1957 and 1995 amounts to 7259 ha. This is 99% of the forest cover that existed in 1957. On the other hand, cultivated land increased from 39% in 1957 to 70% in 1982 and 77% in 1995. The greatest expansion occurred between 1957 and 1982 (about 78%) and slowed down between 1982 and 1995 (only 10%) because almost no land was left for further expansion. Throughout the period covered by the study, cultivation encroached upon the very last marginal areas and steep slopes with gradients >30%. Such a dramatic change in 4 decades and the increasing proportion of completely degraded lands, from virtually nil in 1957 to about 3% in 1995, clearly indicates the prevailing danger of land degradation in the area.

DOI

Options
Outlines

/

〈 〉