Research Articles

Spatial and temporal variation of the urban impervious surface and its driving forces in the central city of Harbin

  • LI Miao , 1, * ,
  • ZANG Shuying , 1 ,
  • WU Changshan 2 ,
  • NA Xiaodong 1
  • 1. School of Geographical Sciences, Harbin Normal University, Harbin 150025, China
  • 2. Department of Geography, University of Wisconsin-Milwaukee, PO Box 413, Milwaukee, WI 53201-0413, USA

Author: Li Miao (1984-), PhD, specialized in land use/cover change and urban remote sensing. E-mail:

*Corresponding author: Zang Shuying (1963-), Professor, E-mail:

Received date: 2017-01-31

  Accepted date: 2017-04-01

  Online published: 2018-03-10

Supported by

Natural Science Foundation of Heilongjiang Province, No.QC2016050

National Natural Science Foundation of China, No.41571199, No.41601382, No.41771195


Journal of Geographical Sciences, All Rights Reserved


Associated with the rapid economic development of China, the level of urbanization is becoming a serious concern. Harbin, the capital city of Heilongjiang Province, China and one of the political, economic, cultural, and transportation centers of the northeastern region of China, has experienced rapid urbanization recently. To examine the spatial patterns of long-term urbanization and explore its driving forces, we employed the impervious surface fraction derived from remote sensing image as a primary indicator. Specifically, urban impervious surface information for the central city of Harbin in 1984, 1993, 2002, and 2010 was extracted from Landsat Thematic Mapper image using a Linear Spectral Mixture Analysis (LMSA). Then, the spatial and temporal variation characteristics and the driving factors of percent impervious surface area (ISA) changes were analyzed throughout this 26-year period (1984 to 2010). Analysis of results suggests that: (1) ISAs in the central city of Harbin constantly increased, particularly from 1993 to 2010, a rapid urbanization period; (2) the gravity center of impervious surface area in the central city was located in Nangang District in 1984, moving southeast from 1984 to 1993, northwest from 1993 to 2002, and continuing toward the southeast from 2002 to 2010; and (3) the urban growth of the central city can be characterized as edge-type growth.

Cite this article

LI Miao , ZANG Shuying , WU Changshan , NA Xiaodong . Spatial and temporal variation of the urban impervious surface and its driving forces in the central city of Harbin[J]. Journal of Geographical Sciences, 2018 , 28(3) : 323 -336 . DOI: 10.1007/s11442-018-1475-z

1 Introduction

Urban expansion, one of the basic characteristics of urbanization, is considered an important factor influencing the natural urban ecosystem. Urban expansion inevitably causes land cover changes, shown by the rapid conversion from natural cover to artificial cover (Kuang, 2012). Therefore, precise spatial and temporal dynamic information regarding urban land cover is required to understand the dynamics of the mechanism of the effect of urbanization within the regional ecosystem evolution. Land-Use and Land-Cover Change (LUCC) data is used for most research on urban expansion. However, the land cover changes brought about by urban expansions are generally shown by gradient features, rendering it impossible to use Land-Use and Land-Cover Change to effectively identify inner heterogeneous characteristics of the same land cover type. Changes in the impervious surface are the main indicators of urban expansion. Urban impervious surfaces refer to the artificial materials that water cannot penetrate to soil. For impervious surface, a continuous value of 0-1 is used to represent the percentage of impervious surface in each pixel. Impervious surfaces, a typical land cover constituent, can effectively describe the spatial gradient features of land cover changes (Ridd 1995; Li and Wu, 2016). It is a significant indicative factor for urban environmental quality and urban ecosystems, and its growth is closely related to driving factors such as urban development strategic targets and overall urban planning, significantly effecting the sound sustainable development of a city (Wang, 2013). Impervious surface produces a direct effect on regional vertical radiation balance through changes in surface albedo, emissivity, and surface roughness. This is caused by changes within the urban surface structure, thereby aggravating the surface sensible heat flux and heat island intensity, changing the regional climate, and affecting urban ecosystem service functions, particularly the thermal regulation function (Haashemi et al., 2016). Concurrently, urban impervious surfaces exhibit poor water storage capacity and obstruct air current transmission, resulting in substantial eco-environmental element effects, such as the urban land surface hydrological cycle, non-point source pollution, and biodiversity, thus becoming an important cause of urban eco-environmental changes. Yang and Liu (2005) and Yang (2006) have utilized the urban impervious surface to analyze the speed and spatial features of urban growth and have suggested that urban impervious surface information indicates urban expansion. Weng (2004) and Hao (2016) demonstrated that distribution of the urban impervious surface has an important relationship with the urban heat island effect. In addition to being used to measure the natural environment and ecological health, the urban impervious surface also reflects the urban inner structure, which is closely linked with urban social and humanistic conditions (Weng et al., 2009; Yuan and Bauer, 2007). Wu and Murray (2005, 2007) have utilized urban impervious surface information to estimate the detailed distribution information of residential population of a city. Yu and Wu (2004) demonstrated that urban impervious surface information can reflect quality of living and have used it to study urban population isolation. Research by Yu and Wu (2006) found that the urban impervious surface has certain effects on housing prices. Therefore, the extraction of urban impervious surface information has become a hot topic, with research on dynamic changes in the urban impervious surface being of great practical significance.
We chose the central city of Harbin as the study area to analyze the spatial and temporal patterns of impervious surface and the associated driving forces during a 26-year period (1984 to 2010). Further, these spatial and temporal patterns of impervious surface were examined through employing a boosted regression tree method with eight selected driving forces, including slope, aspect, DEM, distance to rivers, distance to expressways, distance to railways, distance to main roads, and distance to the city center. With the relationship between driving forces and impervious surface dynamics, this research may provide implications for urban planning policies.

2 Study area and data

2.1 Study area

Harbin city, located in the southwest of Heilongjiang Province, China, was selected as the study area. It is surrounded by the middle reaches of the Songhua River, between Xiao Hinggan Mountains and Zhangguangcai Ridge (see Figure 1). Harbin’s geographical coordinates are approximately 125°42'-130°10'E and 44°04'-46°40'N, with a total geographic area of 53,100 km2. It has an average annual temperature of 3.4℃, average annual evaporation of 1326 mm, average annual frost-free period of 130 days, and average annual rainfall of 500 mm. Harbin is the capital city of Heilongjiang Province and one of the political, economic, and cultural centers, and transport hubs in Northeast China. Harbin has been gradually developed into a synthetic city (Song and Gao, 2008), with the largest geographical area and second in terms of residential population within all provincial cities in China.
Figure 1 Location of the study area

2.2 Dataset

For this research, Landsat Thematic Mapper (TM) images and Digital Elevation Model (DEM) data were obtained from the United States Geological Survey. These imageries were re-projected to the Universal Transverse Mercator (UTM) coordinate system with a datum of the World Geodetic System 84 (WGS84). Statistical data were extracted from China City Statistical Yearbook and Harbin City Statistical Yearbook covering 1984, 1993, 2002 and 2010. The TM images on September 14, 1984; September 7, 1993; September 16, 2002; and September 22, 2010 were selected. The 2002 image was with a cloud coverage of 3%, but after cropping, the cloud coverage in the research area was 0%. For further analyses, bands 1-5 and 7 of the TM data with a spatial resolution of 30 m were selected.

2.3 Remote sensing data preprocessing

2.3.1 Geometric correction
For this research, the 2010 TM image, which has gone through precise geometric correction, was selected as the base map to correct other images. For georeferencing, the polynomial geometric correction module embedded in ERDAS Imagine 9.2 was utilized, with UTM as the coordinate system and WGS 84 as the datum to ensure the exact matching of sampling point positioning coordinates and remote sensing image projection coordinates (Mei, 2001; Markham, 1986). After the determination of projection parameters, 60 ground control points (GCPs), including road intersections, building corners, and other well-defined objects, were selected through field work and careful examination of Google Earth high resolution aerial photos. With the georeference, the total root mean squared error (RMSE) was controlled within 0.5 pixel, indicating a satisfactory geometric positioning accuracy.
2.3.2 Radiometric correction
Radiometric calibration is a process in which the digital numbers (DNs) recorded by the sensor are converted to absolute radiance values (Liang, 2009; Liu et al., 2005) or to relative values relating to surface reflectance or surface temperature. In this research, a Landsat Calibration special module provided by ENVI 4.7 was employed for radiometric calibration, according to the following formula:
${{L}_{\lambda }}=LMI{{N}_{\lambda }}+\left( \frac{LMA{{X}_{\lambda }}-LMI{{N}_{\lambda }}}{QCALMAX-QCALMIN} \right)\left( QCAL-QCALMIN \right)$ (1)
where QCAL is the DN of the original quantization; LMINXλ is the radiance pixel value in the case of QCAL = 0; and LMAXλ is the radiance pixel value in the case of QCAL = QCALMAX.

3 Methods

3.1 Linear Spectral Mixture Analysis (LSMA)

The pixels in remote sensing images are rarely pure pixels consisting of a single uniform land cover, but rather mixed pixels comprising several kinds of land covers. In spite of different spectral characteristics within different land covers, the pixels involved in sensing records only have a single spectral characteristic, i.e., the characteristics of several types of land covers are mixed. Therefore, the spectral characteristics of pixels in the image are not the spectral characteristics of a single land cover but are a mixed reflection of the spectral characteristics of several types of land covers (Tang and Xu, 2014). If a mixed pixel can be decomposed, and the percentage that its land cover type constituent occupies in the pixel can be obtained, misclassification problems caused by the attribution of the mixed pixel can be solved. This process is generally referred to as spectral mixture analysis (SMA). SMA can refine remote sensing classification from the pixel level to the sub-pixel level, thus providing a superior solution to the mixed pixel problem. SMA comes in a linear type and a nonlinear type, with many researchers proposing different impervious surface estimate methods based on SMA models. Linear SMA (LSMA) is widely used for the extraction of impervious surface data from remote sensing images with medium and coarse resolution. A linear spectral mixture model is defined as the reflectivity of a pixel at a certain band being a linear combination of the reflectivity of several different endmembers, with the percentage they occupy in the pixel area as a weight coefficient (Wu et al., 2006), as shown in the following formula:
${{R}_{i}}=\sum\limits_{k=1}^{n}{{{f}_{k}}{{R}_{ik}}+}{{\varepsilon }_{i}}\left( \sum\limits_{k=1}^{n}{{{f}_{k}}=1,0\le }{{f}_{k}}\le 1 \right)$ (2)
where Ri is the reflectivity of Band i of a mixed pixel, containing one or more endmember components; k represents a particular endmember, and n is the number of end members; fk is the percentage of endmember k inside the pixel; Rik is the reflectivity of endmember k at band i; and εi is the fitting error of the model at band i.
3.1.1 Endmember selection
A scatter diagram shows the distribution of the vector in the reflectivity space that is made up of the reflectivity values in the same pixel position of different bands. For this research, endmembers were selected through analyzing the N-dimensional feature space interactions of the first three principal components obtained through principal component analysis (PCA) transformation. Figure 2 depicts the space scatter diagram of the first three component endmembers after PCA transformation of four images. Endmembers are selected through pure pixels of each category and are generally distributed at the top point of a triangular feature space; the closer to the edge, the higher the purity. Endmember include high albedo, low albedo, soil, and vegetation.
Figure 2 Feature space representation of the first three PCA components
3.1.2 Accuracy assessment
Mean absolute error (MAE) and root-mean-square error (RMSE) were used to evaluate the accuracy. MAE reflects the practical situation of the predicted value error as an absolute value employed for dispersion, avoiding the offset of positive and negative values, and the RMSE is capable of evaluating the deviation between the observed value and true value.
$MAE=\frac{1}{n}\sum\limits_{i=1}^{n}{\left| {{f}_{i}}-{{y}_{i}} \right|}$ (3)
$EMSE=\sqrt{\frac{1}{n}\sum\limits_{i=1}^{n}{{{({{f}_{i}}-{{y}_{i}})}^{2}}}}$ (4)
where fi is the estimated percentage of endmember i within a sample; yi is the measured percentage of endmember i for the sample, and n is the total number of samples.
For the selection of samples, we used a random sampling strategy to choose 200 samples from the original image. A sample size of 3×3 pixel, which could reduce the effect of geometric error, was employed. The actual proportion of impervious surface in the 3×3 pixel was calculated by visual interpretation and digitization. The proportion of the impervious surface derived from LSMA was then calculated. Finally, RMSE and MAE were calculated to be 0.19 and 0.14, respectively, thus indicating the applicability of present result.

3.2 Boosted regression tree

The boosted regression tree(BRT)is an ensemble learning method based on the Classification and Regression Tree (CART) algorithm. This method produces a multiple regression tree through random selection and a self-learning method that is able to improve the model stability and prediction accuracy (Li et al., 2014; Elith et al., 2008). The BRT algorithm is an optimization technology involving the minimized loss function’s prediction of the residual value of the previous tree through repeated fitting of the new tree. BRT involves random selection of a certain quantity of data a given number of times in the operation process to analyze the effect of independent variables on dependent variables, and the remaining data is used to test the fitting result. The use of BRT within urban expansion research can help obtain not only the relative effect of each driving factor but also the mechanism of the relationship between relative effect and the changes in each driving factor, thereby obtaining accurate and intuitive results (Freund and Schapire, 1997).

4 Results and discussion

4.1 Impervious surface expansion analysis

The distributions of impervious surfaces from 1984 to 2010 (see Figure 3) illustrate obvious spatial variations over this period. In 1984, the impervious surface in Harbin was mainly concentrated within the city center. In spite of a small geographical area, there were a number of impervious surface areas with high percent coverage. In 1993, the area of impervious surface expanded. In 2002, the area of impervious surface did not expand significantly when compared with the area in 1993, but the scope of impervious surface expanded a lot in terms of scope, with impervious surface distributed north of the Songhua River. In 2010, the impervious surface significantly expanded to the north of the Songhua River, with the distribution of impervious surface patches with high coverage found.
Figure 3 Distribution of impervious surfaces in Harbin central city (1984-2010) (a. 1984; b. 1993; c. 2002; d. 2010)
From 1984 to 1993, the area of the impervious surface increased from 48.51 to 57.1 km2, with an increased rate of 0.95 km2 per year. From 1993 to 2002, the area of the urban impervious surface increased from 57.1 to 72.85 km2, with an increased rate of 1.75 km2 per year. In 2010, the area of impervious surface increased to 125.04 km2. The increased area of impervious surface during the 8-year period from 2002 to 2010 was more than three times that during the period from 1993 to 2002. The increased area of impervious surface during the 8-year period from 2002 to 2010 doubled with respect to that during the period from 1984 to 2002, indicating rapid development within the city during the period 2002-2010.

4.2 Shift of the gravity center of the impervious surface in the central city

In this study, the calculation of the gravity center shift of the urban impervious surface in the city center is not based on a simple gravity center calculation but on dividing the percentage of impervious surface in the pixel into 10 grades from 0.1 to 1 at 0.1 intervals. Then, each grade is given a corresponding weight, based on which the calculation is conducted, yielding the gravity center shift of the urban impervious surface in the Harbin city center from 1984 to 2010. From 1984 to 2010, the gravity center of the urban impervious surface in the city center was distributed in the Nangang District, but moved 346.26 m towards the southeast during 1984-1993, which is consistent with the results of the impervious surface pattern analysis above, because the impervious surface mainly spread towards the south from 1984 to 1993. From 1993 to 2002, the gravity center of the urban impervious surface in the central city moved 685.68 m towards the northwest, with the movement direction being the opposite of that from 1984 to 1993 and the movement distance being nearly twice that from 1984 to 1993. From 2002 to 2010, the impervious surface in the central city continued to move 877.51 m towards the northwest.

4.3 Analysis of the driving force for impervious surface changes

The spatial characteristics of land cover can be effectively described by impervious surface. The growth level of the impervious surface area is closely related to the urban population, urban economic development, overall urban planning, and other driving factors. In this study, population, economic, and planning factors were selected as the external drivers of variation in the urban impervious surface area. Internal driving factors included slope, aspect, DEM, distance from the river, distance from the highway, distance from the railway, distance from the main road, and distance from the urban area. These external and internal driving factors were selected for analysis of the driving force behind variation in the urban impervious surface.
4.3.1 External factors affecting impervious surface area variation
The non-agricultural population is an important factor reflecting the urbanization level of a region. Because variation in the non-agricultural population inevitably leads to expansion of the urban area, the non-agricultural population was selected as the demographic factor index in this study. With the development of the socialist market economy, the regulating function of the market economic system plays a very important role in the exploitation and utilization of land resources. Gross Domestic Product (GDP) is often recognized as the best indicator of the state of the country’s economy. Harbin is the former northeastern industrial base, and the proportion of its secondary industry in GDP is also an important factor in urban land expansion. In recent years, Harbin has undergone industrial restructuring, and the proportion of its tertiary industry in GDP has also become a significant factor. Meanwhile, the development of the urban construction industry is directly reflected by its increased gross output. During this period of rapid economic growth, personal income was also on the rise. Brueckner (2000) suggested that the increase of personal income is one of the main driving factors of urban spatial expansion. Because of their effects on urban expansion, population, GDP, GDP of the secondary and tertiary industries, average wage, and GDP of the construction industry were selected to analyze the driving forces of the variation in the urban impervious surface area (see Table 1).
Table 1 Driving factors and indices
Driving factors Indices
Demographic factor X1 non-agricultural population
Economic factor X2 GDP
X3 total output of the secondary and
tertiary industries
X4 the average wage
X4 GDP of the construction industry
(1) The population factor
In 1984, the non-agricultural population of Harbin was 2.59 million, with the urban population increasing to 4.72 million in 2010, a factor of 1.82 over 26 years. In this study, regression analysis was carried out for the non-agricultural population and the impervious surface area, as shown in Figure 4. The x-axis represents the impervious area, and the y-axis represents the non-agricultural population. The goodness of fit of a second order trend line is 0.967, indicating a high correlation of population growth to impervious surface area. Because of the population’s production and living needs, a large amount of construction land is required to meet the needs of the spatial expansion. An increasing population requires additional houses, roads, factories, schools, and shopping malls, as well as supporting facilities. Furthermore, with improved living standards, people in pursuit of material goods will also pursue spiritual improvement, such that the number of tourism and entertainment will also increase, leading to additional development and a corresponding increase of the impervious surface area.
Figure 4 Non-agricultural population as a function of impervious surface area
(2) The economic factors
In this study, regression analysis was carried out for the impervious surface area, GDP, GDP of the second and tertiary industries, average wage, and GDP of the construction industry, as shown in Figure 5. The x-axis represents the impervious area and the y-axis in each of the four graphs represents GDP, GDP of the second and tertiary industries, average wage, and GDP of the construction industry, respectively. The correlations measured 0.998, 0.998, 0.999, and 1.0, respectively. During the 15 years spanning the 9th-11th Five-Year Plans, the economic level of Harbin improved significantly. In 1984, the GDP of Harbin was 6.88 billion yuan, and increased to 366.99 billion yuan in 2010, a factor of 53.27 over 26 years. Similarly, the GDP of the secondary and tertiary industries increased from 5.26 billion yuan in 1984 to 325.2 billion yuan in 2010, and the proportion of the secondary and tertiary industries increased from 76.45% to 88.7%, respectively, over the same time period. The development of the secondary industry not only increased the industrial land area but also led to a large number of people moving into the city, causing the urban housing land area to increase. Furthermore, the development of the tertiary industry also caused the urban impervious surface area to increase. In Harbin, the average wage in 1984 was 996 yuan and 32,397 yuan in 2010, an increase by a factor of 32.53. With improvement in living standards, more people buy houses, expanding the real estate industry and subsequently causing the urban impervious surface area to increase.
Figure 5 Function of impervious surface area (a. GDP; b. total output of the secondary and tertiary industries; c. average wage; d. gross product of construction)
(3) The planning factors
The ‘Decision on Economic Restructuring’ adopted in 1984 suggested that the reformation focus was shifted from rural areas to urban areas. During the 10th Five-Year Plan (2001-2005), the process of urban planning approval in Harbin was reformed. During these five years, 5623 construction projects of various types were approved, adding floor area of 41.27 km2. These projects included more than 80 shantytown renovation projects and more than 1500 urban infrastructure construction projects. During the 10th Five-Year Plan (2001-2005), the developed area increased from 211 km2 to 318 km2. During the 11th Five-Year Plan (2006-2010), the municipal government proposed an increase in the residential land area, as well as public and municipal facilities, which led to an increase in the urban impervious surface area.
4.3.2 Internal factors affecting impervious surface area variation
The dependent variable in this study was the increase in the impervious surface in the city center from 1984-2010. Through the different operations based on the urban impervious surface within the city center during 1984 and the urban impervious surface diagram in 1984, the magnitude of change in the urban impervious surface in the city center from 1984 to 2010 is obtained. Zero means that in 1984 and 2010 no urban impervious surfaces were involved, indicating no new buildings in this area. A number greater than 0 indicates no impervious surface in 1984 and impervious surface in 2010, or that the percentage of impervious surfaces in 2010 is greater than that in 1984, showing an increase in the number of buildings in this area. A one is assigned to anything greater than 0, and this change value depicts a dependent variable. Eight factors are selected: slope, aspect, DEM, distance from the river, distance from the highway, distance from the railway, distance from the main road, and distance from the urban area in 1984. The eight driving factors are as shown in Figure 6. DEM is between 98 m and 199 m, with the northern DEM lower than the southern DEM in general. The distance from the main urban area in 1984 is 0-11,433.9 m, with a uniform distance from the central city as a whole. The northwestern section is a bit longer than that in other parts. Main roads and railways are distributed in various directions in the central city owing to the prevalence of highways; the distances are distributed in a strip shape. The eight factors are obtained through the Spatial Analyst Tool module in ArcGIS, with 10,000 sample points randomly generated. Then the attribute values within the points for each figure layer correspondence are exported, and finally the data is imported into the software R for the BRT analysis. The Elith program is utilized for the BRT analysis, with the learning rate set to 0.005. Fifty percent of the data is extracted for analysis each time, and the remaining 50% of the data was used as training, with cross validation conducted five times. The BRT operation is conducted in R. The relative operating characteristic (ROC) value is 0.923, indicating the results is significant.
Figure 6 Driving factors (a. DEM; b. slope; c. aspect; d. distance from city center of 1984; e. distance from main road; f. distance from river; g. distance from railway; h. distance from highway)
Through the BRT analysis, the effect of each factor on the changes in the impervious surface in the Harbin city center is analyzed; the factor that has the greatest effect is the distance from the urban area, with its contribution rate being the highest, 29%. The contribution rates of the other driving factors in order from the largest to the smallest are as follows: DEM (13.4%), distance from the highway (12.4%), distance from the railway (12.2%), distance from the river (10.3%), distance from the main road (9.9%), slope (6.8%), and aspect (6%).
The analysis shows that neighborhood factors have a great effect on the urban impervious surface, with the relative effect of five neighborhood factors on the urban impervious surface ranking 1st, 3rd, 4th, 5th, and 6th among all factors, with their total effect reaching up to 73.8%. DEM also plays a significant role in the increase of the impervious surface of the city center, with a relative effect of 13.4%, ranking second among the eight factors.
Figure 7 shows the effect change curves of the relative effects of driving factors, and these curves show the changes with the values taken for the driving factors and the changes in their effects on the urban impervious surface. When the relative effect value is greater than 0, the effect of the driving factors on the increase in the urban impervious surface is positive. When the relative effect value is less than 0, the effect of the driving factors on the urban impervious surface is negative. When the value is 0, there is no relationship. The distance to the urban area reflects the importance of geographical location. The Harbin city center provides goods to the surrounding areas, and the influence of the city center decreases with increasing distance. The relative effect of the distance from the urban area in 1984 demonstrates that the effect decreases with an increase in distance from the city center. When the distance from the urban area is within 4300 m, the effect of the distance to the urban area on the urban impervious surface shows a positive correlation: the closer to the urban area, the greater the effect. When the distance is larger than 4300 m, the effect is negative, such that the distance has a restrictive effect on the urban impervious surface. When the distance is greater than 10,000 m, it is beyond the urban radiation effect scope, and the distance to the urban area has almost no effect on the urban impervious surface. Harbin is a flat terrain plain area, with a low altitude. When the DEM is between 120 m and 165 m, the relative effect is positive, which indicates that the altitude in this area is suitable for urban development. Because of the presence of many rivers in Harbin, the altitude has a restrictive effect on urban expansion. Highways, railways, and main roads are the main transport carriers of traffic flows and logistics in modern cities, which is of great significance for urban spatial expansion, having a direct effect on the urban expansion direction. When there is a distance of 5600 m from the railway and 1000 m from the main road, the relative effect is positive, indicating social and economic development can be created in this area, thereby contributing to the urban expansion along the railway or the road.
Figure 7 Relative influence on the change of impervious surface in the central city

(a. distance from the city center of 1984; b. DEM; c. distance from highway; d. distance from railway; e. distance from river; f. distance from main road)

5 Conclusions and limitations

By using the Landsat TM in 1984, 1993, 2002, and 2010 as a reference for the central city of Harbin, the urban impervious surface was extracted using the linear spectral mixture analysis method. The spatial and temporal variation characteristics of the urban impervious surface in the city center were analyzed, and the factors influencing the expansion of the urban impervious surface in the city center during a 26-year period (1984 to 2010) were further analyzed, with specific conclusions as follows:
(1) From 1984 to 2010, the impervious surface coverage of Harbin city continued to increase, particularly from 2002 to 2010 during which rapid increase occurred. The city center has constantly spread.
(2) From 1984 to 2010, the gravity center of the urban impervious surface in Harbin city center was distributed throughout Nangang District but moved to the southeast from 1984 to 1993, to the northwest from 1993 to 2002, and continued to move towards the southeast from 2002 to 2010. This is due to the rapid development of Songbei District under the development strategy “Southern Leap, Northern Expansion, Central Revitalization, and County Strengthening,” showing a close link between urban expansion and policies of development planning.
(3) The driving force of the urban expansion was analyzed, and regression analysis was carried out in order to quantify the correlation between population and economic changes and impervious surface area changes. Additionally, the influence of planning policies on the increase of the urban impervious surface was analyzed. It can be seen that population, economic factors, and planning policies all play a positive role in increasing the impervious surface area.
(4) The relative effect of each driving factor on the expansion of the city center indicates that the expansion is mainly shown by edge-type growth. In other words, the closer to a city edge an area is, the more easily the area is developed into a city. The time scale in this study was very short, only 26 years, and over such a short time scale, natural factors have little effect on urban expansion.
(5) It has been difficult to account for the spatialization of socio-economic factors, which affects the analysis of driving factors to a certain degree. This study involves qualitative analysis of economic factors, with no good solutions to this problem presented. Efforts will be increased to make improvements in future research.

The authors have declared that no competing interests exist.

China City Statistical Yearbook. . Accessed on 2013-03-01.

Elith J, Leathwick J R, Hastie T, 2008. A working guide to boosted regression trees.Journal of Animal Ecology, 77(4): 802-813.Summary 1 Ecologists use statistical models for both explanation and prediction, and need techniques that are flexible enough to express typical features of their data, such as nonlinearities and interactions. 2 This study provides a working guide to boosted regression trees (BRT), an ensemble method for fitting statistical models that differs fundamentally from conventional techniques that aim to fit a single parsimonious model. Boosted regression trees combine the strengths of two algorithms: regression trees (models that relate a response to their predictors by recursive binary splits) and boosting (an adaptive method for combining many simple models to give improved predictive performance). The final BRT model can be understood as an additive regression model in which individual terms are simple trees, fitted in a forward, stagewise fashion. 3 Boosted regression trees incorporate important advantages of tree-based methods, handling different types of predictor variables and accommodating missing data. They have no need for prior data transformation or elimination of outliers, can fit complex nonlinear relationships, and automatically handle interaction effects between predictors. Fitting multiple trees in BRT overcomes the biggest drawback of single tree models: their relatively poor predictive performance. Although BRT models are complex, they can be summarized in ways that give powerful ecological insight, and their predictive performance is superior to most traditional modelling methods. 4 The unique features of BRT raise a number of practical issues in model fitting. We demonstrate the practicalities and advantages of using BRT through a distributional analysis of the short-finned eel ( Anguilla australis Richardson), a native freshwater fish of New Zealand. We use a data set of over 13000 sites to illustrate effects of several settings, and then fit and interpret a model using a subset of the data. We provide code and a tutorial to enable the wider use of BRT by ecologists.


Feng D J, Li Yongshu, Deng F, 2004. Detecting change information automatically by method of wavelet coefficient difference. Remote Sensing Information, (2): 13-15.Tow dimensional wavelet transform is widely used in various fields of image processing.It is utilized in landuse remote sensing dynamic monitoring to detect chang information automatically in the paper. By using PCA transform on the multi spectrum images of two different periods, the first three principal components were chosen to perform wavelet transform, and the two temporal wavelet coefficients of corresponding principal components were subtracted. It is detected that value difference of land using areas with no change was approximately zero, while the absolute value difference of the changing areas was great and with high energy, so the image performed wavelet counter transform can display land using changing areas with high brightness. Using image data to emulate and field validate the monitor result, the precision using the method mentioned here can meet the demand of practical application.


Freund Y, Schapire R E, 1995. A decision-theoretic generalization of on-line learning and an application to boosting. In: European Conference on Computational Learning Theory. Springer-Verlag: 119-139.We consider the problem of dynamically apportioning resources among a set of options in a worst-case on-line framework. The model we study can be interpreted as a broad, abstract extension of the well-studied on-line prediction model to a general decision-theoretic setting. We show that the multiplicative weight-update rule of Littlestone and Warmuth [10] can be adapted to this mode yielding bounds that are slightly weaker in some cases, but applicable to a considerably more general class of learning problems. We show how the resulting learning algorithm can be applied to a variety of problems, including gambling, multiple-outcome prediction, repeated games and prediction of points in n . We also show how the weight-update rule can be used to derive a new boosting algorithm which does not require prior knowledge about the performance of the weak learning algorithm.


Hao P, Niu Z, Zhan Yet al., 2015. Spatiotemporal changes of urban impervious surface area and land surface temperature in Beijing from 1990 to 2014.Giscience & Remote Sensing, 53(1): 1-22.This study examined changes in urban expansion and land surface temperature in Beijing between 1990 and 2014 using multitemporal TM, ETM+, and OLI images, and evaluated the relationship between percent impervious surface area (%ISA) and relative mean annual surface temperature (RMAST). From 1990 to 2001, both internal land transformation and outward expansion were observed. In the central urban area, the high-density urban areas decreased by almost 7 km2, while the moderate- and high-density urban land areas increased by 250 and 90 km2, respectively, outside of the third ring road. From 2001 to 2014, high-density urban areas between the fifth and sixth ring roads experienced the greatest increase by more than 210 km2, and RMAST generally increased with %ISA. During 19900900092001 and 20010900092014, RMAST increased by more than 1.5 K between the south third and fifth ring roads, and %ISA increased by more than 50% outside of the fifth ring road. These trends in urban expansion and RMAST over the last two decades in Beijing can provide useful information for urban planning decisions.


Harbin City Statistical Yearbook. . Accessed on 2013-03-01.

Kuan W H, 2012. Evaluating impervious surface growth and its impacts on water environment in Beijing-Tianjin-Tangshan Metropolitan Area.Journal of Geographical Sciences, 22(3): 153-165.The impervious surface area (ISA) at the regional scale is one of the important environmental factors for examining the interaction and mechanism of Land Use/Cover Change (LUCC)-ecosystem processes-climate change under the interactions of urbanization and global environmental change. Timely and accurate extraction of ISA from remotely sensed data at the regional scale is challenging. This study explored the ISA extraction based on MODIS and DMSP-OLS data and the incorporation of China land use/cover data. ISA datasets in Beijing-Tianjin-Tangshan Metropolitan Area (BTTMA) in 2000 and 2008 at a spatial resolution of 250 m were developed, their spatiotemporal changes were analyzed, and their impacts on water quality were then evaluated. The results indicated that ISA in BTTMA increased rapidly along urban fringe, transportation corridors and coastal belt both in intensity and extents from 2000 to 2008. Three cities (Tangshan, Langfang and Qinhuangdao) in Hebei Province had higher ISA growth rates than Beijing due to the pressure of population-resources-environments in the city resulting in increasingly transferring industries to the nearby areas. The dense ISA distribution in BTTMA has serious impacts on water quality in the Haihe River watershed. Meanwhile, the proportion of ISA in sub-watersheds has significantly linear relationships with the densities of river COD and NH3-N.


Li C L, Liu M, Hu Y Met al., 2014. Driving forces analysis of urban expansion based on boosted regression trees and logistic regression.Acta Ecologica Sinica, 34(3): 727-737.(in Chinese)The rapid relentless urban area expansion has led to a series of problems in China. Many researches focused on this issue in recent years. Driving forces are the core topic in urban expansion,as well as the basic component of modeling and predicting. It is very useful and meaningful to analyze the driving force of urban expansion, which may provide us with a scientific basis to rationally utilize land resources, determining the law of urban development, researching the evolution process, predicting the urban expansion trends, and also providing guidance for the development of rational control policies. The Shenyang city was chosen as study area. Eight categories of land use types were extracted from remote sensing images (1997 and 2010) with ArcGIS software. Ten driving forces were chosen, including three natural factors, three distance factors, four social and economic factors. which were calculated based on the land use maps, DEM, topographic maps, zoning maps and the statistical yearbooks. The dependent variable was the change of built-up area of Shenyang from 1997 to 2010. Boosted regression trees (BRT) is an ensemble method and is a combination of techniques between statistical and machine learning traditions that has shown to be effective to identify relationships between results and influencing factors. Logistic regression is a method to discover the empirical relationships between a binary dependent and several independent categorical and continuous variables. Boosted Regression Trees and Logistic regression were used to analyze the main driving force of urban expansion synthetically. The result illustrated the relative influence of driving factors was followed by distance from urban area of 1997, distance from river, DEM, distance from highway and railway, land use types, development plan, GDP, population density, aspect, and slope based on BRT analysis. According to Logistic regression analysis, the relative influence of important factors was followed by development zone, distance from urban land of 1997, DEM, distance from highway and railway, population density, distance from river, rural residential areas and slope. The most important driving forces affecting the expansion of Shenyang are distance from urban area of 1997, DEM, distance from highway and railway. Meanwhile, they were all located in the top four of the main factors. The results revealed that the distance factors were the most important factors, and the total contribution rate of relative influence was up to 61.4%. It is demonstrated distance factors are the main driving forces of urban expansion. Natural factorswere less important, but the relative influence of DEM was important, and the contribution rate was 12.5%. Development zones and rural settlements are the only two factors have much influence in the socio-economic factors. On the whole, location factors, which refer to the distance from urban land in this study, were the leading factors of urban expansion. Natural factors, such as DEM, rivers and so on, are the basis of urban development, determining the overall urban spatial form. The construction of infrastructures, such as roads and railways, are the frame of the city. The social and economic factors decided the speed of urban expansion. Urban planning and development zone construction provided the direction of urban expansion.


Li W L, Wu C S, 2016. A geostatistical temporal mixture analysis approach to address endmember variability for estimating regional impervious surface distributions. Giscience & Remote Sensing, 53(1): 102-121.Spectral mixture analysis (SMA) is a major approach for estimating fractional land covers through modeling the relationship between the spectral signatures of a mixed remote sensing pixel and those of the comprised pure land covers (also termed as endmembers). When SMA is implemented, endmember variability has proven to have significant impact on the accuracy of land cover fraction estimates. To address the endmember variability problem, this article developed a geostatistical temporal mixture analysis (GTMA) technique, with which spatially varying per-pixel endmember sets were estimated using an ordinary kriging interpolation technique. The method was applied to time-series moderate-resolution imaging spectroradiometer normalized difference vegetation index imagery in Wisconsin and North Carolina, United States to estimate regional impervious surface distributions. Analysis of results suggests that GTMA has achieved a promising accuracy. Detailed analysis indicates that a better performance has been achieved in less-developed areas than developed areas, and slight underestimation and slight overestimation have been detected in developed areas and less-developed areas, respectively. Moreover, while the performance of GTMA is comparable to those of phenology-based TMA and phenology-based multiple endmember TMA over the entire study area and in less-developed areas, a much better performance has been achieved in developed areas. Finally, this article argues that endmember variability may be more essential in developed areas when compared to less-developed areas.


Liang S L, 2009. Quantitative Remote Sensing. Beijing: Science Press, 14-16. (in Chinese)

Liu Z, Wang Y, Peng Jet al., 2011. Using ISA to analyze the spatial pattern of urban land cover change: A case study in Shenzhen.Acta Geographica Sinica, 66(7): 961-971.(in Chinese)Based on land cover classification data,there are two key problems that has been often overlooked when we use indices method to analyze the landscape pattern of land cover in rapid urbanization areas.One is mixed pixel involved in the classification process,which has impact on classification accuracy and the final conclusion of landscape pattern analysis.The other is that it is difficult for landscape pattern indices method to detect the change in a pixel and local urban areas,which can only explain the macro regional patterns of urbanization.To solve these problems,based on continuous data,Linear Spectral Method Analysis(LSMA) is used to acquire the index of Impervious Surface Area(ISA) in this case study,considering impervious surface component as the main landscape in urban areas.Thus,we can effectively analyze the spatial pattern and expansion processes of urbanization.Taking Shenzhen as a study area,spatial autocorrelation,semi-variance function and other geo-statistical methods are used to reveal the macro spatial-temporal patterns of a continuous landscape change,and fractal dimension and profile methods are also used to analyze urban landscape along the change direction of location.The results indicated that the continuous landscape metrics and geostatistical methods can help us to understand the spatial and temporal changes of urbanization at regional and local levels,since land cover change,especially in rapid urbanization areas,has a significant gradient characteristic and spatial continuity.


Markham B L, 1986. Landsat MSS and TM Post-Calibration Dynamic Ranges, Exoatmospheric Reflectances and At-Satellite Temperatures. Landsat Technical Notes, 3-8.react-text: 543 Energy and water balance parameters were measured in two commercial vineyards in the semiarid region of the S o Francisco river basin, Brazil. Actual evapotranspiration (ET) was acquired with the Bowen ratio surface energy balance method. The ratio of the latent heat flux to the available energy, or evaporative fraction (EF), was 81% on average for two growing cycles in wine grape and 88% for... /react-text react-text: 544 /react-text [Show full abstract]

Mei X A, Peng W L, Qing Q M et al., 2001. Introduction to Remote Sensing. Beijing: Higher Education Press. (in Chinese)

Song G, Gao N, 2008. Economic benefit analysis of urban land utilization based on DEA method: A case of Harbin City.Scientia Geographica Sinica, 28(2): 185-188. (in Chinese)In recent years,the adjustment of administrative divisions has changed in the city land scale,the land utilization structure and its using way.Adopting scientific and effective methods in appraisal economic benefit of urban land utilization economic benefit become more important.This article takes landuse in Harbin built-up district as the object of research,combining Harbin development as well as the present situation of landuse and construct evaluating indicator system.It takes DEA(Data Envelopement Analysis) as method and operates with Matlab to appraise Harbin land utilization economic benefit from 2001 to 2005.According to the analysis result,it may be seen that Harbin City land utilization economic benefit level is ordinary and its land investment has been sufficient.Accordingly,the article proposes three facets to improve Harbin City land utilization economic benefit,such as strengthening the use of existing urban lands,as well as paying more attention to the investment ratio of different land types and the adjustment of the industrial structure.


Tang F, Xu Q, 2014. Comparison of performances in retrieving impervious surface between hyperspectral (Hyperion) and multispectral (TM/ETM+) images. Spectroscopy and Spectral Analysis, 4: 1075-1080.Abstract The retrieval of impervious surface is a hot topic in the remote sensing field in the past decade. Nevertheless, studies on retrieving impervious surface from hyperspectral image and the comparison of the performances in retrieving impervious surface between hyperspectral and multispectral images are rarely reported. Therefore, The present paper focuses on the characteristics of hyperspectral (EO-1 Hyperion) and multispectral (Landsat TM/ETM+) images and implements a complementary study on the comparison based on the retrieved impervious surface information between Hyperion and TM/ETM+ data. For up to 242 bands of Hyperion image, a further study was carried out to select feature bands for impervious surface retrieving using stepwise discriminant analysis. As a result, 11 feature bands were selected and a new image named Hyperion' was thus composed. The new Hyperion' image was used to investigate whether this band-reduced image could obtain higher accuracy in retrieving impervious surface. The three test regions were selected from Fuzhou, Guangzhou and Hangzhou of China, with date-coincident or nearly coincident image pairs of the used sensors. The linear spectral mixture analysis (LSMA) was employed to retrieve impervious surface and the results were accessed for their accuracy. The comparison shows that the Hyperion image has higher accuracy than TM/ETM+, and the Hyperion' composed of the selected 11 feature bands has the highest accuracy. The advantages of Hyperion in spectral and radiometric resolutions over TM/ETM+ are believed to be the main factors contributing to the higher accuracy. The high spectral and radiometric resolutions of Hyperion image allow the sensor to have higher sensitivity in distinguishing subtle spectral changes of ground objects. While, the highest accuracy the 11-band Hyperion' image achieved is owing to the significant reduction of the band dimension of the image and thus the band redundancy.


Wang D, 2013. Study on urban growth based on information extraction of impervious surface from remote sensing imagery: A case study on urban built-up area of Lanzhou City [D]. Lanzhou: Lanzhou University. (in Chinese)

Weeks J R, 2005. Measuring temporal compositions of urban morphology through spectral mixture analysis: Toward a soft approach to change analysis in crowded cities. International Journal of Remote Sensing, 26(4): 699-718.This paper reports on preliminary results from a study applying the technique of spectral mixture analysis (SMA) to the measurement of temporal changes in the composition of urban morphology in the metropolitan area of Greater Cairo, Egypt, between 1987 and 1998. Although several remote sensing techniques have been used successfully for urban change analysis, most of these focus on change ‘between’ classes measured in a discrete, crisp way through which each pixel is assigned to a label indicating either a change or no change in the class to which the pixel originally belonged. In many major cities, such as Cairo, change also occurs within classes (e.g. vertical growth of buildings, increase in housing density, decrease in open spaces) and is reflected by an aggregation of land cover and urban materials. None of these materials may seem important in isolation. Rather, the significance of these urban land covers arises from the way they interweave with each other to structure the morphology of the urban place. In this paper, a ‘soft’ approach is presented to identify and measure the composition of changing morphology from multi‐temporal, multi‐spectral satellite images. SMA is demonstrated to be capable of deriving spatially continuous variables quantified at the sub‐pixel level. These variables represent measures that can be compared across urban places and at different time periods. They can be integrated readily into a wide range of applications and models concerned with physical, economic and/or socio‐demographic phenomena in the city.


Weng Q, Lu D, Schubring J, 2004. Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies.Remote Sensing of Environment, 89(4): 467-483.Remote sensing of urban heat islands (UHIs) has traditionally used the Normalized Difference Vegetation Index (NDVI) as the indicator of vegetation abundance to estimate the land surface temperature (LST) egetation relationship. This study investigates the applicability of vegetation fraction derived from a spectral mixture model as an alternative indicator of vegetation abundance. This is based on examination of a Landsat Enhanced Thematic Mapper Plus (ETM+) image of Indianapolis City, IN, USA, acquired on June 22, 2002. The transformed ETM+ image was unmixed into three fraction images (green vegetation, dry soil, and shade) with a constrained least-square solution. These fraction images were then used for land cover classification based on a hybrid classification procedure that combined maximum likelihood and decision tree algorithms. Results demonstrate that LST possessed a slightly stronger negative correlation with the unmixed vegetation fraction than with NDVI for all land cover types across the spatial resolution (30 to 960 m). Correlations reached their strongest at the 120-m resolution, which is believed to be the operational scale of LST, NDVI, and vegetation fraction images. Fractal analysis of image texture shows that the complexity of these images increased initially with pixel aggregation and peaked around 120 m, but decreased with further aggregation. The spatial variability of texture in LST was positively correlated with those in NDVI and in vegetation fraction. The interplay between thermal and vegetation dynamics in the context of different land cover types leads to the variations in spectral radiance and texture in LST. These variations are also present in the other imagery, and are responsible for the spatial patterns of urban heat islands. It is suggested that the areal measure of vegetation abundance by unmixed vegetation fraction has a more direct correspondence with the radiative, thermal, and moisture properties of the Earth's surface that determine LST.


Weng Q H, Hu X F, Liu H, 2009. Estimating impervious surfaces using linear spectral mixture analysis with multitemporal ASTER images.International Journal of Remote Sensing, 30(18): 4807-4830.Impervious surface is a key indicator of urban environmental quality and degree of urbanization. Therefore, estimation and mapping of impervious surfaces by using remote sensing digital images has attracted increasing attention recently. For mid-latitude cities, seasonal vegetation phenology has a significant effect on the spectral response of terrestrial features, and image analysis must take into account this environmental characteristic. In this paper, three Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images, acquired on 3 October 2000, 16 June 2001 and 5 April 2004, respectively, were used to test the seasonal sensitivity of impervious surface estimation. The study area was the city of Indianapolis (Marion County), Indiana, USA. Linear spectral mixture analysis (LSMA) was applied to generate high-albedo, low-albedo, vegetation and soil fraction images (endmembers), and impervious surfaces were then estimated by adding high- and low-albedo fraction images. In addition, land use/land cover (LULC) and land surface temperature (LST) maps were generated and used to create image masks to remove non-impervious pixels. The accuracy of the impervious surface maps was checked against Digital Orthophoto Quarter Quadrangle (DOQQ) images. Three accuracy indicators, the root mean square error (RMSE), mean average error (MAE) and correlation coefficient (R 2), were calculated and compared to analyse the seasonal sensitivity of impervious surface estimation. Our results indicate that vegetation phenology has a fundamental impact on impervious surface estimation. The summer (June) image was better for impervious surface estimation than the spring (April) and autumn (October) images. The LULC and LST image masks can significantly increase the accuracy of impervious surface estimation. The mean LST was found appropriate to be set as the threshold for the various image masks. A summer image was most appropriate because there was full growth of vegetation, and mapping of impervious surfaces was more effective with a contrasting spectral response from green vegetation. The mixing space, based on the four endmembers, was perfectly three-dimensional. By contrast, there was significant amount of bare soil and ground and non-photosynthetic vegetation in the spring and autumn images. Plant phenology caused changes in the variance partitioning and impacted the mixing space characterization, leading to a less accurate estimation of the impervious surfaces.


Wu C, Murray A T, 2005. A cokriging method for estimating population density in urban areas.Computers Environment & Urban Systems, 29(5): 558-579.Population information is typically available for analysis in aggregate socioeconomic reporting zones, such as census blocks in the United States and enumeration districts in the United Kingdom. However, such data mask underlying individual population distributions and may be incompatible with other information sources (e.g. school districts, transportation analysis zones, metropolitan statistical areas, etc.). Moreover, it is well known that there are potential significance issues associated with scale and reporting units, the modifiable areal unit problem (MAUP), when such data are used in analysis. This may lead to biased results in spatial modeling approaches. In this study, impervious surface fraction derived from Thematic Mapper (TM) imagery was applied to derive the underlying population of an urban region. A cokriging method was developed to interpolate population density by modeling the spatial correlation and cross-correlation of population and impervious surface fraction. Results suggest that population density can be accurately estimated using cokriging applied to impervious surface fraction. In particular, the relative population estimation error is 0.3% for the entire study area and 10 15% at block group and tract levels. Moreover, unlike other interpolation methods, cokriging gives estimation variance at the TM pixel level.


Wu C, Murray A T, 2007. Population estimation using landsat enhanced thematic mapper imagery.Geographical Analysis, 39(1): 26-43.Abstract An assessment of two groups of approaches for estimating urban population with remote-sensing information is presented in this article. These approaches, zonal and pixel-based models, are applied to Landsat Enhanced Thematic Mapper images of a portion of Columbus, Ohio , to generate population estimates. The zonal approach uses impervious surface fraction, spectral radiance, and land-use/land-cover classification to derive population estimates. The pixel-based approach uses impervious surface fraction and spectral radiance to estimate the population of residential areas. To assess robustness, these models were applied to Dayton, Ohio . A comparative study indicates that the models generated promising results in estimating regional population counts. However, zonal regression with spectral radiance produced large errors (76%) for census block groups, whereas other models gave significantly better estimation accuracy. Comparing the performance of the indicators, impervious surface fraction is competitive, and slightly but consistently better than land-use classification. In comparison with traditional zonal approaches, pixel-based models give somewhat better estimation accuracy.


Wu C Q, Wang Q, Yang Z F, 2006. Cloud-moving of water RS image based on mixed pixel model.Journal of Remote Sensing, 10(2): 176-183.There are always some uneven aerosol or thin cloud effects in the remote sensing image of large-scale inland water.These uneven effects bring great difficulty to atmospheric radiant correction of remote sensing images in such regions.Furthermore,because the water is the object with low reflectivity,these uneven effects bring large errors to the water information inversion by remote sensing technology.For some reasons,it is some times impossible for us to use traditional atmospheric radiant correction algorithms(such as MODTRAN or 6S software) to reduce such effects.With some remote sensing images and ground collecting data in Taihu Lake in China,an important inland water research place of Chinese remote sensing scholars,the authors use a new method to resolve this problem.This method considers the optical characteristics of water-atmosphere environment carefully,assumes each pixel's spectrum is the mixed result of water,pollutant and aerosol or thin cloud,for the optical property of Taihu(Lake's) remote sensing images is determined by water,pollutant and aerosol or thin cloud.Based on Mixed Pixel Model,this method reduces the effects of aerosol and thin cloud effectively.After the process of this method,we can get more veracious water quality information from remote sensing images.In(author's) test,to the same remote sensing inversion model,this method can increase 5 percent precision.

Yang X, 2006. Estimating landscape imperviousness index from satellite imagery.IEEE Geoscience and Remote Sensing Letters, 3: 6-9.This letter presents a practical method for landscape imperviousness estimation through the synergistic use of Landsat Enhanced Thematic Mapper Plus (ETM+) and high-resolution imagery. A 1-m resolution color-infrared digital orthophoto was used to calibrate a stepwise multivariate statistical model for continuous landscape imperviousness estimation from medium-resolution ETM+ data. A variety of predictive variables were initially considered, but only brightness and greenness images were retained because they were account for most of the imperviousness variation measured from the calibration data. The performance of this method was assessed, both visually and statistically. Operationally, this method is promising because it does not involve any more sophisticated algorithms, such as classification tree or neural networks, but offers comparable mapping accuracy. Further improvements are also discussed.


Yang X, Liu Z, 2005. Use of satellite-derived landscape imperviousness index to characterize urban spatial growth.Computers, Environment and Urban System, 29: 524-540.Urban change analysis has traditionally been supported through land use/cover classification and map-to-map comparison. In this research, we investigate the usefulness of satellite-derived imperviousness index as an alternative for urban spatial growth characterization. The study area, Pensacola, FL, has witnessed considerable growth in population and regional economies during the past decade. The research consists of a number of procedures. First, we identify a method for landscape imperviousness estimation by synergistic use of medium-resolution satellite imagery and high-resolution color orthophoto through multivariate statistical analysis. We apply this method to map landscape imperviousness index for the years of 1989 and 2002, respectively. We assess the maps’ accuracy with the imperviousness estimation from high-resolution DOQQ imagery as the reference. The overall error is estimated to be less than 10%. Then, we analyze the spatio-temporal changing trend of landscape imperviousness index with the emphasis upon some ‘hot’ spots of development areas. We find that this trend is compatible with the urban land use/cover changing trend detected through image interpretation. We conclude that satellite-derived landscape imperviousness index is able to serve as an invaluable alternative for quick and objective assessment of urban spatial growth, particularly over large areas.


Yu D L, Wu C S, 2004. Understanding population segregation from Landsat ETM+ imagery: A geographically weighted regression approach.Giscience & Remote Sensing, 41(3): 187-206.This study attempts to understand population segregation issues in Milwaukee County, Wisconsin utilizing remote sensing and regression technologies. Population segregation was measured with a local segregation index Di based on the theory of the index of dissimilarity. Remote sensing information was extracted from a Landsat ETM+ image through spectral mixture analysis, unsupervised classification, and texture analysis. Global ordinary least squares (OLS) regression and geographically weighted regression (GWR) analyses were applied to explore the relationships between population segregation and remote sensing variables. Results indicate that remote sensing information has the potential to increase our understanding of socio-cultural issues such as population segregation.


Yu D L, Wu C S, 2006. Incorporating remote sensing information in modeling house values.Photogrammetric Engineering & Remote Sensing, 72(2): 129-138.This paper explores the possibility of incorporating remote sensing information in modeling house values in the City of Milwaukee, Wisconsin, U.S.A. In particular, a Landsat ETM+ image was utilized to derive environmental characteristics, including the fractions of vegetation, impervious surface, and soil, with a linear spectral mixture analysis approach. These environmental characteristics, together with house structural attributes, were integrated to house value models. Two modeling techniques, a global OLS regression and a regression tree approach, were employed to build the relationship between house values and house structural and environmental characteristics. Analysis of results indicates that environmental characteristics generated from remote sensing technologies have strong influences on house values, and the addition of them improves house value modeling performance significantly. Moreover, the regression tree model proves as a better alternative to the OLS regression models in terms of predicting accuracy. In particular, based on the testing dataset, the mean average error (MAE) and relative error (RE) dropped from 0.202 and 0.434 for the OLS model to 0.134 and 0.280 for the regression tree model, while the correlation coefficient between the predicted and observed values increased from 0.903 to 0.960. Further, as a nonparametric and local model, the regression tree method alleviates the problems with the OLS techniques and provides a means in delineating urban housing submarkets.


Yuan F, Bauer M E, 2007. Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery.Remote Sensing of Environment, 106(3): 375-386.This paper compares the normalized difference vegetation index (NDVI) and percent impervious surface as indicators of surface urban heat island effects in Landsat imagery by investigating the relationships between the land surface temperature (LST), percent impervious surface area (%ISA), and the NDVI. Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) data were used to estimate the LST from four different seasons for the Twin Cities, Minnesota, metropolitan area. A map of percent impervious surface with a standard error of 7.95% was generated using a normalized spectral mixture analysis of July 2002 Landsat TM imagery. Our analysis indicates there is a strong linear relationship between LST and percent impervious surface for all seasons, whereas the relationship between LST and NDVI is much less strong and varies by season. This result suggests percent impervious surface provides a complementary metric to the traditionally applied NDVI for analyzing LST quantitatively over the seasons for surface urban heat island studies using thermal infrared remote sensing in an urbanized environment.