Research Articles

Spatio-temporal differences in cloud cover of Landsat-8 OLI observations across China during 2013-2016

  • XIAO Chiwei 1, 2 ,
  • Li Peng , 1, 2, 3 ,
  • FENG Zhiming 1, 2 ,
  • WU Xingyuan 3
Expand

Author: Xiao Chiwei (1991-), PhD Candidate, specialized in resource utilization and remote sensing of land resource. E-mail: xiaocw@igsnrr.ac.cn

*Corresponding author: Li Peng (1984-), PhD and Associate Professor, E-mail: . ORCID ID (https://orcid.org/0000-0002-0849-5955)

Received date: 2017-06-15

  Accepted date: 2017-11-16

  Online published: 2013-02-11

Supported by

National Natural Science Foundation of China, No.41430861

National Key Research and Development Program of China, No.2016YFC0503500

Opening Fund of Key Laboratory of Poyang Lake Wetland and Watershed Research, Ministry of Education, Jiangxi Normal University, No.PK2016004

Copyright

Journal of Geographical Sciences, All Rights Reserved

Abstract

Currently, the historical archive images of Landsat family sensors are probably the most effective data products for tracking global longitudinal changes since the 1970s. However, the issue of the degree and extent of cloud coverage is always a challenge and varies distinctively worldwide. So far, acquisition probability (AP) analyses of cloud cover (CC) of Landsat observations have been conducted with different sensors at regional scale. To our knowledge, CC probability analysis for the newly-launched Landsat-8 Operational Land Imager (OLI) across China is not reported. In this paper, monthly, seasonal, and annual APs for Landsat OLI (44,228 in total) images over China acquired from April 2013 to October 2016 with various CC thresholds were analyzed. The results showed that: first, the cumulative average APs of all OLI data over China at the CC thresholds ≤30% was about 49.6% which illustrated the availability of OLI imagery across China. Second, the spatial patterns of 10%, 20%, and 30% CC thresholds of OLI observations, coincided well with the precipitation distributions separated by the respective 200 mm, 400 mm, and 800 mm isohyetal lines. Third, the APs of images with the 30% CC threshold are the highest in autumn and winter especially in October of 58.7%, while the corresponding lowest probability occurred in June of 41.0%. Finally, the spatial differences in APs of targeted images with ≤30% CC thresholds were quite significant. At regional scales, the arid and semi-arid areas, Inland River and Songliao River basins, and northwestern side of the Hu Huanyong population line had the larger probabilities of obtaining high-quality images. Our study suggested that OLI imagery satisfy the data requirements needed for land surface monitoring, although there existed obvious spatio-temporal differences in APs over China at the 30% CC threshold.

Cite this article

XIAO Chiwei , Li Peng , FENG Zhiming , WU Xingyuan . Spatio-temporal differences in cloud cover of Landsat-8 OLI observations across China during 2013-2016[J]. Journal of Geographical Sciences, 2013 , 28(4) : 429 -444 . DOI: 10.1007/s11442-018-1482-0

1 Introduction

China has experienced rapid land use/land cover changes (LUCC) in ecotones, fragile ecological regions (Liu et al., 2016; Ren and Wang, 2017), and metropolitan areas (Liu et al., 2014; Wu et al., 2016) in the past decades. Indeed, the vast territory, various climates and diverse topography in China, are conducive to a variety of LUCC. However, before detecting LUCC, and its characteristics, rates, drivers, and impacts (Hao and Ren, 2009; Liu and Deng, 2010), the fundamental issue of the “observability” of remote sensing data needs to be evaluated (Foley et al., 2005), as well as understanding its applicability and limitations, especially historical Landsat data.
Currently, remote sensing involving optical, microwave radar, and lidar satellites, or combinations thereof has become an effective tool to monitor land surface changes and ecosystem dynamics (Aplin, 2004; Jin et al., 2013b; Teillet et al., 2007). Of these, the freely available Landsat products have become an increasingly widespread imagery source for agricultural, ecological, and environmental monitoring (Jin et al., 2013b; Li et al., 2014; Qin et al., 2007), although they are usually impacted due to contamination of frequent cloud cover (CC) and cloud shadows (Dong et al., 2013; Hansen and Loveland, 2012; Li et al., 2012; Liu and Liu, 2013). Historical 30-m resolution data of Landsat-4/5 Thematic Mapper (TM, 1982-2013), Landsat-7 Enhanced Thematic Mapper Plus (ETM+, 1999-), and Landsat-8 Operational Land Imager (OLI, 2013-) are important and extensive data sources of Earth observation since the late 1980s, covering geo-information, natural vegetation, agricultural crops, and land cover types (Hansen and Loveland, 2012; Jin et al., 2013a; Steven et al., 2003). However, the most obvious challenge for their application is CC and cloud shadows which are widely reported in Landsat-derived remote sensing analyses (Asner, 2001; Ju and Roy, 2008; Hansen and Loveland, 2012; Liu et al., 2016; Whitcraft et al., 2015), particularly in the subtropics and tropics (Asner, 2001; Li et al., 2016). Consequently, some studies have attempted to address the issue of CC, such as the use of cloud removal algorithms to exclude CC areas or replace them with other images (Ju and Roy, 2008; Zhu and Woodcock, 2012). Nevertheless, these methods tend to result in some errors due to the variability and complexity in cloud types (Kovalskyy and Roy, 2015), and usually require time- and labor-intensive work.
Recently, a few studies have indicated that Landsat images are vulnerable to CC and cloud shadows during the dry season in many regions of tropical, sub-tropical, and temperate monsoon areas (Dong et al., 2013; Li et al., 2016). However, one awkward but unanswered question is what is the acquisition probability (AP) of usable images for the past nearly 40 years with CC less than 30% over China from the Landsat-like optical data, as China has undergone noticeable land use/cover dynamic changes since the 1980s (Liu et al., 2014). Unfortunately, the problem of the ‘observability’ of Landsat family sensors (i.e., TM, ETM+, and OLI) still unaddressed globally (Goward et al., 2006; Sano et al., 2007). Thus, quantifying CC of Landsat images is a paramount prerequisite for monitoring land surface change at various regional scales (Asner, 2001; Li et al., 2017), especially in the monsoon regions (e.g., China). Currently, there are only a few published studies concerning the probability analysis of CC in different Landsat sensors throughout the Amazon Basin (Asner, 2001; Sano et al., 2007), the availability of cloud-free images covering the conterminous United States and globally (Ju and Roy, 2008; Kovalskyy and Roy, 2013; Kovalskyy and Roy, 2015), and the probability differences of CC in Mainland Southeast Asia (MSEA) (Laborde et al., 2017; Li et al., 2017). Unlike cloud avoidance strategy of Landsat-5/7 Long-term Acquisition Plan (LTAP) which does not collect all images outside the US (Arvidson et al., 2001; Asner, 2001), Landsat-8 provides global coverage of observation since 2013. Currently, APs analysis of CC of Landsat-8 observation is not fully investigated worldwide (Laborde et al., 2017).
To our knowledge, a comprehensive analysis of cloud coverage for the newly launched Landsat-8 OLI sensor (February 11, 2013) data over China has not been published to date. In this study, we used the CC information of all OLI historical metadata [a total of 44,228 scenes] acquired over China from January 2013 to October 2016 to evaluate the monthly, seasonal, and annual APs, and the limitation and suitability for such as terrestrial surface studies. The objectives of this study are twofold: (1) to understand and evaluate the average APs of Landsat OLI with different CC thresholds in China; and (2) to analyze the spatial differences in CC of Landsat OLI images among major river basins or grain production regions, dry and wet regions, and the two regions divided by the Hu Huanyong population line (Hu Line) in China. This study contributes to providing necessary guidance in the aspect of Landsat-8 OLI data source for monitoring land surface changes.

2 Materials and methods

2.1 Study area

China has a latitude span of 50 degrees from north to south and a longitude span of approximately 62 degrees from west to east (Figure 1) (Ge et al., 2016). Large spans of latitude and longitude generally lead to significant differences in precipitation, resulting in dry and wet regions (Figure 1), respectively. China has many different types of climate, such as subtropical, temperate monsoon, and temperate continental climate. There are four distinct seasons, namely spring from March to May, summer from June to August, autumn from September to November, and winter from December to February of the following year. In winter, northerly winds from high-latitude regions are cold and dry, while southerly winds from coastal areas at lower latitudes are warm and moist in summer. Among them, the cold-dry season is characterized by low CC with less precipitation per month, which facilitates obtaining high-quality images, or cloud-free or little cloudy images.
Landforms in China include large mountains and plateaus in the west and north, and low hills and alluvial plains in the southern and eastern areas. Northwestern China has very little precipitation because it is located deep inland far from the oceans. However, southwestern China has a distinction between the rainy season (from May to October) and the dry season (from November to April) due to the seasonal changes of Indian Ocean monsoon (Li et al., 2017). Agriculture in northwestern China is greatly determined by the accessibility of irrigation. China’s northern plateau region is dominated by grasslands. However, ranchers who seek for short-term profit have caused overgrazing of the grassland. China’s eastern and southern lowlands (i.e., low hills and alluvial plains) comprise the major agricultural zones and river basins (e.g., Yangtze River and Yellow River) (Jin et al., 2016). This region is also characterized by agriculture for major rice-producing area (Ding, 1961) with a larger population density.
Figure 1 Map showing the location of the dry and wet areas, and the two sides divided by the Hu Line in China, and the 576 Landsat OLI footprints (path/row) over China
With rapid economic development, LUCC is currently a very common phenomenon, especially in southern and eastern China (Liu et al., 2014). Several factors are driving this trend, including an enhancement of the impact of the long-standing Hu Line (Figure 1), which marks the striking difference in the distribution of China’s population. The population and economy flow into the southeastern part of the Hu Line due to better biophysical and social conditions. Meanwhile, series of serious ecological problems have occurred in northern and northwestern China due to overgrazing and other unsuitable farming practice in recent decades, which is a primary form of LUCC in this region.

2.2 Landsat OLI CC data and pre-processing

There are 576 path and row (PR) coverage frames (or footprints, e.g., 121/040) of Landsat-8 OLI satellite over China according to the Worldwide Ref System (http://landsat. usgs.gov/tools_wrs-2_shapefile.php). In particular, all Landsat OLI footprints are required to cover China’s islands (Figure 1). The PR information and all Landsat-8 OLI metadata, including cloud coverage (0-100%) and its levels (0-9), acquisition and processing date, and geographic location (e.g., scene center’s latitude and longitude) were collected from the USGS Landsat inventories provided by the Landsat Bulk Metadata Service (http://landsat. usgs.gov//consumer.php). With regard to the cloud coverage, we only used the CC percentage average value in the whole Landsat OLI scene in this study. The percentage of cloud coverage of Landsat 8 cloud cover assessment (CCA) uses multiple algorithms to detect clouds in scene data (USGS, 2016), including automatic CCA (ACCA), See-5 CCA, Cirrus CCA, and Artificial Thermal-A CCA.
A total of 44,228 Landsat OLI scenes were acquired over China from April 2013 to October 2016. There were 8121, 12,671, 13,086, and 10,350 scenes in 2013, 2014, 2015, and 2016, respectively (Table 1). It should be noted that the Landsat-8 OLI satellite has acquired all the images in China, including the Diaoyu island and other surrounding islands in South China Sea (Figure 1). The thresholds of cloud coverage range from 0 to 100% in these Landsat metadata files, which were divided into ten levels (0-9) at 10% intervals by location (PR) (Asner, 2001). Note that level 0 represents the CC of all scenes equal to 10% or less (or 0% ≤ CC ≤ 10%), and level 1 represents the range from 10% to 20% (including 20%) (or 10% < CC ≤ 20%). Following this categorization, the last level 9 represents greater than 90%, but less than or equal to 100% (or 90% < CC ≤ 100%). CC thresholds are referred as 0%, 10% to 100% or level 0, level 1 to level 9 in this study. In this study, the percentage information of CC in Landsat OLI sensor data was analyzed for all path and row combinations for April 2013 to October 2016 over China.
Table 1 Annual Landsat OLI scenes acquisition statistics in China including cloud cover (CC) levels (0-9) in each footprint (path/row)
Year Number of
scenes
Proportion
(%)
Number of CC level
0 1 2 3 4 5 6 7 8 9
2013 8121 18.4 2546 948 705 634 602 631 608 654 571 222
2014 12671 28.6 3712 1561 1185 957 868 884 915 988 1023 578
2015 13086 29.6 3655 1570 1168 962 951 906 1016 1153 1152 553
2016 10350 23.4 2712 1228 945 808 816 774 843 952 878 394

2.3 AP calculation of different CC thresholds

The monthly, seasonal and annual APs of CC thresholds ranging from level 0 (or 0% ≤ CC ≤ 10%) to level 9 (or 90% < CC ≤ 100%) were computed over China at the spatial scale of each Landsat footprint (Path/Row, 576 in total) (Figure 1). Firstly, cumulative probabilities of different CC thresholds were calculated to show the appropriate CC threshold. Secondly, the seasonal cumulative probabilities of increasing CC (i.e., 10%, 20% and 30%) were calculated to display the regional and intermonth variations. Then, we further delineated monthly average probability of CC for Landsat sensors with ≤ 30%. Lastly, the average APs of regional were calculated based on the above results. The probability of CC for a successful acquisition was calculated using formula (1) with the Microsoft Visual Basic application (Asner, 2001; Li et al., 2017). The metadata archive of cloud coverage from the 44,228 Landsat OLI images and statistical results were spatially displayed along the path and row (e.g., 120/040) using the ArcGIS platform (version 10.1).
(1)
where S represents the probability of a certain acquisition of different CC for each year for each scene, m and y denotes each month (m = January, …) and year (i.e., between 2013 and 2016), t represents the different CC thresholds (t = 0, 10%, …, 100% or level, level 1,…, level 9, as described above), and N denotes the total number of observation scenes in month (m) in the entire Landsat archive.

3 Results and analysis

3.1 Appropriate CC threshold determination for AP analysis

In order to understand the effects of increasing CC thresholds (i.e., from 0, 10% to 100%), the average probability of all Landsat OLI observation was first evaluated. Figure 2 shows that the patterns of the average APs at the monthly (Figure 2a), seasonal, and annual scales (Figure 2b) with increasing CC thresholds for all Landsat OLI images. The results indicated that the curves of average AP with increasing CC thresholds show a “crescent” shape, which suggests that the APs do not increase significantly over a whole year. There was a clear and rapid increase beyond level 0 CC threshold. Overall, in autumn and winter, Landsat OLI data acquisition over China had higher probabilities, while the corresponding APs remained comparatively low in spring and summer. The highest AP during the autumn and winter periods occurred in October, followed by December, January, February, November, and September, especially at the 30% CC threshold (0% ≤ CC ≤ 30%). Similarly, in spring and summer, June had the lowest probability, followed by any other month, August, July, May, April, and March. It should be noted that the values of average AP in March and September play a transition role for CC in response to the transformation between southeast and northwest monsoons.
Figure 2 The cumulative average acquisition probabilities (APs) of increasing CC thresholds range from 0-100% for all OLI footprints in China: monthly (a) and annual and seasonal (b)
In general, a lower CC threshold makes an image more suitable for monitoring land surface changes. Some studies have found that the CC threshold is consistently less than or equal to 30% (Asner, 2001; Goward et al., 2006; Li et al., 2017). This CC threshold (i.e., less than or equal to 30%) represents the greatest extent possible value for land surface monitoring. Thus, the ≤ 30% CC threshold was also selected in this study. The corresponding cumulative annual AP averages of CC for Landsat OLI images were 28.6%, 42.5%, and 49.6% for level 0, level 1 and level 2, respectively. The annual AP results with the ≤ 30% CC in China are much larger than those of MSEA (41%) (Li et al., 2017) and global estimates (37%) (Goward et al., 2006). This also demonstrates that the CC level 2 (i.e., less than or equal to 30%) is more appropriate for analysis of usable Landsat images (Ju and Roy, 2008; Asner, 2001), especially in China.
Next, the monthly average APs were computed for all Landsat OLI and are presented in Figure 2a, where the ≤ 30% CC threshold is shown with a solid vertical line. It showed evident variability in the monthly average APs at the 30% CC threshold. The APs of CC at the 30% were greater than 50% over China for over six months from October to March next year. Figure 2a also shows that Landsat OLI had a relatively higher probability of successfully obtaining high-quality images from October to March during 2013-2016, with an average CC of 55.1%. Among them, the related AP reached a peak value (58.7%) in October, followed by 56.8% in December, 55.0% in January, 54.1% in February, 53.1% in November, and 52.9% in March. It can be explained that the weakening subtropical high pressure and the beginning of cold air from Siberia in October usually leads to dry weather (“crisp autumn day”) in China. This allows the acquisition of cloudless or little-cloud images in this month. On the contrary, the average monthly probability value was 45.3% between April and September. Particularly, the smallest probability for OLI acquiring targeted images occurred in June, merely 41.0%. The average monthly APs of other months are listed as follows, August (42.5%), July (44.2%), May (46.0%), April (49.3%) and September (48.9%). It can also be explained that the major river (e.g. Yangtze river and Yellow river) basins generally enter into flood season because of concentrated rainfall in this month (June), or earlier in April and May and end in September. The rainy weather hinders the observation of high-quality images.
Then, an analogous statistical analysis of seasonal AP showed that the seasonal variance with CC thresholds of ≤ 30% was distinct for the periods of autumn to winter and spring to summer. The average seasonal probability of acquiring a successful Landsat OLI image in winter was 55.3%, followed by 53.4% in autumn, 49.1% in spring, and 42.6% in summer (Figure 2b). In summary, analyses of the amount of useful Landsat OLI images at the CC thresholds ≤ 30% have great importance for monitoring land surface changes in China.

3.2 Spatio-temporal patterns of AP at the 30% CC threshold

3.2.1 Spatial comparison analysis of AP at the 10%, 20%, and 30% CC thresholds
Three thresholds of annual CC (i.e., 10%, 20% and 30%) were further selected to show the spatial differences in AP of OLI images for the period of 2013-2016 (Figure 3). Comparative results indicated that the distributions of OLI scene AP with the three CC thresholds are closely correlated with the isohyetal lines in China (Figure 3).
Specifically, the spatial patterns of AP for the three CC thresholds coincided with the 200 mm (Figure 3a), 400 mm (Figure 3b), and 800 mm (Figure 3c) isohyetal lines, respectively. Particularly, the spatial pattern of AP at 30% CC threshold coincides with the spatial distribution of the northwest and southeast, typically divided by the Qinling Mountains-Huaihe River Line (or the 800 mm isohyetal line), an important natural boundary between southern and northern China, which indicates transitions in geography, climate, and ecology. This correlation provides important insight in the influence of clouds and precipitation on optical satellite acquisition in the monsoon region. Figure 3 also indicates that the proportions of high-quality images data on the northwestern side of the line were much larger than those of the southeastern. In the following sections, the selection of the 30% CC threshold in this study was applied to further confirm the differences of AP at varied spatial and temporal scales. It should be pointed out that our analysis derived from the scene-based metadata were only for the whole Landsat OLI footprints, but did not take the overlay parts of any two adjacent footprints (or path/row) into consideration.
3.2.2 Temporal comparison analysis with the 30% CC threshold
China is situated in the transitional zone between the ocean and continent in longitude, covering the temperate, subtropical, and tropical zones. Therefore, regional differences in climate types or vegetation types are distinct. Explicit analysis of AP at the 30% CC threshold for the OLI archive products can contribute to evaluating the usefulness and applicability of OLI imagery at both spatial and temporal scales. Regional features for the 2013-2016 period have a unique spatial distribution and change in clouds as indicated by the Landsat OLI data at this threshold (Figure 4). On the whole, the probabilities of usable OLI image acquisition in autumn and winter were larger than those of the spring and summer seasons.
Figure 3 Annual AP with ≤ 30% CC for Landsat OLI sensor images (44,228 scenes) between 2013 and 2016 in China. (a), (b) and (c) represent the annual CC thresholds of 10%, 20%, and 30%, respectively.
The monthly AP results of the historical OLI scenes showed clear regional differences (Figure 4). Among these differences, those between the northwest and the southeast regions were distinct, especially from October to March. Historical Landsat OLI observations with a threshold of AP at the 30% CC have higher chances of obtaining cloudless or little-cloud images, in most parts of northern China. This also suggests that the highest AP is in October of each year in China, while the corresponding lowest probability occurred in June. For example, in the middle reach of the Yangtze River, the observation probability remains highest in October, at the beginning of the winter season and the ending of the autumn season (Figure 4). Overall, northern and western China had AP values >90% even during the wetter spell of the year. The maximum variation in CC occurred in the northeast (especially the Sanjiang Plain), the Tibetan Plateau, and Hengduan Mountains. Among these areas, the entire Tibetan Plateau had 90%-100% chance of successful imaging from October to March. In addition, seasonal variance comparisons showed that tropical region of China (e.g., Hainan Island) had the larger APs for cloudless or little-cloud OLI images in spring and summer and lower APs in autumn and winter (Figure 4), while this relationship was slightly different in MSEA (Li et al., 2017). The temporal changes of monsoonal wind systems may well explain the differences between them. These results confirmed the validity of the monthly result at the 30% CC threshold. In total, there is a relatively high probability of acquiring cloudless or little-cloud OLI images in autumn and winter, including the transition periods in March and September (Figure 4). These results also provide useful information for the selection of high-quality Landsat OLI imagery within proper critical window for remote sensing monitoring.
Figure 4 Monthly and seasonal AP for historical Landsat OLI scenes (42,229 scenes) with ≤ 30% CC in China for 2013-2016

3.3 Area and distribution of AP in China based on the 30% CC threshold

The spatial distribution of probability in obtaining OLI scenes at CC level 2 (or ≤ 30% CC) provides a chance for extracting useful surface feature information; for instance, combining image quality in different months for remote sensing monitoring. The relationships between the AP of cloudless or little-cloud images and major geographical regions in China were further analyzed. AP values were calculated for all Landsat OLI images at the regional scale, and categorized into dry and wet regions (climate, Figure 5), river basins or grain production regions (water or land, Figure 6), and two sides of the Hu Line (population, Figure 7).
Figure 5 Regional differences in annual and seasonal (a) and monthly (b) average AP at 30% or less CC for all Landsat data. The four arid and humid areas in China are typically separated by the 200 mm, 400 mm, and 800 mm isohyetal lines.
3.3.1 Spatial differences of AP in dry and wet regions
China has the arid and semi-arid, semi-arid and sub-humid, sub-humid and humid areas, respectively (Cui et al., 2016). Because of rainfall differences in China, the arid area (62.7%) and semi-arid (58.8%) areas had larger annual average AP, compared to those of sub-humid (44.8%) and humid areas (34.7%).
Seasonal variance comparisons (Figure 5a) again showed that the arid area had an AP of about 70% in the autumn season, and 63.0%, 60.4%, and 59.2% for winter, spring, and summer seasons, respectively. In the semi-arid area, the AP was 72.2% in the winter season, followed by 63.3% in autumn, 57.1% in spring, and 46.8% in summer. The average probability of obtaining high-quality images in the sub-humid area in spring was 45.2%, the related probability declined in summer (33.7%), and increased in autumn (46.9%) and winter (56.9%). Only 31.7% of the images had a relatively high AP of a successful observation in summer in the humid area, as well as 36.4% in autumn, 34.5% in winter, and 36.8% in autumn. The variations in seasonal probability for the arid and humid areas were generally slight (less than 15%) in the whole year. However, there were large differences (greater than 50%) in the semi-arid and sub-humid areas between autumn to winter season and spring to summer season. Thus, the AP variations for the semi-arid and sub-humid areas were larger than those for the arid and humid areas. In addition, the chance for successful Landsat observations of arid area was about twice than that of the humid area, equal to the sub-humid area in autumn and winter, and 1.5 times than that of the sub-humid area in spring and summer. The relationship between the semi-arid area and (sub-) humid area was similar. Furthermore, in all (semi-) arid and (semi-) humid areas, the autumn and winter seasons were superior to the spring and summer seasons in obtaining cloudless or little-cloud images.
Finally, regional variations in Figure 5b demonstrate the “best” month in acquiring high-quality images was October for the arid (75.5%) and humid (45.7%) areas. Likewise, the semi-arid (76.9%) and sub-humid areas (58.2%) had higher probabilities of acquiring Landsat OLI cloudless or little-cloud images in December. As a result, Landsat OLI imagery has proven to be more applicable for monitoring land surface changes in the arid (especially in October), semi-arid areas (particularly in December) than the sub-humid (particularly in December), and humid areas (especially in October). The results also demonstrated that CC changed due to the shift between southeast and southwest monsoonal wind systems. This distinct monthly variance of AP values will facilitate studying land surface changes comprehensively and systematically.
3.3.2 Spatial differences of AP in major river basins (or grain production regions)
At present, Landsat imagery is important and successful for evaluating and discriminating the quality of soil and water (Liu et al., 2016). In China, the spatial distribution of grain production region is closely related to nine major river basins: the Songliao River (SLR), Inland River (InR), Luanhai River (LHR), Yellow River (YeR), Pearl River (PR), Yangtze River (YtR), Southwest Rivers (SWRs), Huaihe River (HR), and Southeast Rivers (SERs) basins (Jin et al., 2016; Liu et al., 2007). Spatially, regional differences in the nine river basins varied dramatically in the AP of cloudless or little-cloud OLI images (Figure 6). Among the nine major river basins, the SLR, InR, LHR, and YeR basins had annual AP greater than 50% for obtaining high-quality images, compared with the AP values for the PR (23.4%), YtR (28.7%), SWRs (32.9%), HR (36.6%), and SERs (45.8%) basins. Comparative analyses indicated that there were similar regional differences in CC over the seasonal and monthly variance. Figure 6 summarizes the average seasonal (left panel, Figure 6a) and monthly (right column, Figure 6b) AP values of obtaining cloudless or little-cloud images within a given year in the nine major river basins.
Figure 6 River basin differences in annual and seasonal (a) and monthly (b) average AP at ≤ 30% CC for all Landsat data in China. YtR: Yangtze River; YeR: Yellow River; HR: Huaihe River; LHR: Luanhai River; SLR: Songliao River; PR: Pearl River; SERs: Southeast Rivers; SWRs: Southwest Rivers; InR: Inland River
As shown in Figure 6a, of the nine major river basins, the largest difference of AP happens in the spring season (4.6 times). The LHR basins had the largest probabilities in the spring season (62.9%) whle the corresponding lowest probabilities occurred in the PR basins (13.6%). The ratio of APs between the SERs and PR basins was about 3.1. The SERs basins had the largest probabilities in winter (up to 69.0%) and the corresponding lowest probabilities in the PR basins (22.0%). The difference of statistical results was 2.1 in summer, with 54.8% in InR basins and 25.5% in SERs basins. A similar ratio occurred in the autumn season, the larger probabilities (63.5%) in the InR basins and the corresponding lower probabilities (13.6%). Next, the monthly average APs of Landsat OLI images showed distinct characteristics (Figure 6b). Monthly variance comparisons at a regional scale showed that the SLR basin had the highest AP in January (70.3%, Table 2), and lowest monthly average 48.7% in June. Similar analyses were applied to other basins for the highest and lowest AP values in Table 2. These results also demonstrated that Landsat OLI has higher probabilities of acquiring high-quality iamges in the autumn and winter seasons.
Table 2 Differences in monthly average AP at ≤30% CC threshold for OLI data in the nine major river basins of China
Major River Basin AP (%)
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Yangtze River (YtR) 34.3 28.8 24.6 26.6 25.6 20.4 31.1 27.8 19.6 40.5 31.0 38.0
Yellow River (YR) 57.4 46.5 52.2 61.4 47.5 42.2 45.5 46.5 46.5 53.2 56.2 66.3
Huaihe River (HR) 26.1 29.6 42.5 41.4 42.9 25.9 32.4 36.4 37.5 53.6 31.1 38.9
Luanhai River (LHR) 53.4 49.5 61.7 60.8 65.6 35.3 46.6 55.2 54.7 54.9 59.6 55.7
Songliao River (SLR) 70.3 67.9 64.2 59.1 53.1 48.7 52.1 50.3 59.2 61.2 56.5 68.7
Pearl River (PR) 25.4 16.0 10.0 15.3 14.9 26.9 28.1 24.7 31.2 41.7 17.2 24.0
Southeast Rivers (SERs) 60.9 64.1 52.0 41.8 40.3 27.9 24.4 24.4 37.0 58.1 73.5 70.5
Southwest Rivers (SWRs) 31.0 21.8 25.0 25.6 22.1 29.4 45.1 41.2 37.1 42.2 26.0 42.7
Inland River (InR) 59.5 61.2 58.5 52.9 51.5 52.6 58.5 53.1 61.3 69.6 59.5 62.1
3.3.3 Spatial differences of AP on the two sides of the Hu Line
The Hu Line is considered as one of the most important geographical discoveries in China (Qi et al., 2016). The spatial patterns of CC distribution in the two sides of the Hu Line were obviously different according to the average AP evaluation. Statistical results indicated that the northwestern side of the Hu Line (NWHL) had 56.5% annual AP for obtaining a cloudless or little-cloud Landsat 8 OLI imagery, while the southeastern side of the Hu Line (SEHL) had a lower AP of 39.9%. Generally, CC in the SEHL was much higher than that of the NWHL in any given month or season. This can be explained that SEHL has plenty of rain with humid weather, which easily leads to high CC during the revisit of Landsat sensors. However, the phenomenon was just the opposite in the NWHL. Figure 7 presents the spatial patterns of the AP of cloudless or little-cloud OLI images on both parts of the Hu Line. The results showed that the average monthly or seasonal probabilities of high-quality Landsat OLI images acquisition in the NWHL were larger than those of the SEHL throughout the year.
Seasonally, there were larger APs for cloudless or little-cloud OLI images in winter (65.4%) and autumn (61.5%) seasons in the NWHL, and the corresponding average values declined in the spring (54.7%) and summer (47.4%) seasons. The SEHL showed similar variations in the seasonal average probabilities at the threshold of level 2, with 42.5% and 41.3% in the winter and autumn seasons, respectively, and 42.4% in the spring season and 34.6% in the summer season. Monthly variances showed that the largest AP was 67.8% in December for the NWHL and 47.8% in October for the SEHL, with the difference approximately 20%. In general, the NWHL had higher (over 45%) monthly average AP (including 65.8% in October, 65.0% in November, 64.7% in January, 63.6% in February, 60.3% in March, 55.3% in September, 54.2% in April, 50.7% in May, 49.7% in July, 46.4% in August, and 45.9 in June), when compared to those of the SEHL (about 47.8%, 38.1%, 42.9%, 41.7%, 43.6%, 38.0%, 44.3%, 39.8%, 35.3%, 36.3%, and 32.0% in the corresponding month). Our CC analysis confirmed that the Hu Line acts as a geographical boundary between oceanic climate and continental climate. Therefore, differences in AP are useful for monitoring human settlements with Landsat data on both sides of the Hu Line.
Figure 7 Regional differences in annual, seasonal and monthly average AP at ≤ 30% CC for Landsat OLI data in China on both sides of the Hu Line

4 Conclusions and discussion

4.1 Conclusions

Recently, freely available historical Landsat products have been widely used in land surface monitoring from regional to global level. Cloud cover (CC) analysis of historical Landsat data is a prerequisite for remote sensing monitoring. The acquisition probability (AP) analysis of CC helps to understand the temporal and spatial practicality of obtaining a usable OLI image. In this study, we explicitly investigated the spatio-temporal differences in AP for varied CC thresholds using 44,228 Landsat OLI images over China from April 2013 to October 2016. We then analyzed the differences in AP at the 30% or less CC threshold in the dry and wet regions, river basins or grain production regions, and two sides of the Hu Line (Hu Huanyong population line). Some main conclusions were drawn as follows:
(1) Cumulative frequency showed larger probabilities at lower CC thresholds, in contrast, smaller percentage at higher CC thresholds. The cumulative average AP of all OLI data over China at the CC thresholds ≤ 30% was about 49.6%. This higher AP illustrated the availability of OLI imagery for monitoring land surface changes across China, in spite of obvious national differences in APs.
(2) The spatial patterns of the AP of Landsat OLI at varied CC thresholds and the major isohyetal lines of China coincided well. The spatial patterns of lower-threshold CC of OLI observations, namely 10%, 20%, and 30%, coincided well with the precipitation distributions of China divided by the isohyetal lines of 200 mm, 400 mm, and 800 mm, respectively.
(3) Temporal differences in AP for obtaining cloudless or little-cloud images were clear. China has higher probabilities of acquiring high-quality OLI images in autumn and winter especially in October of 58.7%, while the corresponding lowest probability in June merely 41.0%.
(4) Similarly, the spatial differences in APs of targeted images with ≤ 30% CC thresholds were quite significant. At regional scales, the arid and semi-arid areas, Inland River and Songliao River basins, and northwestern side of the Hu Line had the larger probabilities of obtaining high-quality images for monitoring single locations.

4.2 Discussion

In this study, we delineated the statistical features of CC of Landsat OLI imagery in China and explored the relationships between high-quality imagery and the spatial differences in water, land, and climate. The distinct monthly or seasonal AP variations in different regions at the 30% CC threshold in OLI imagery will contribute to facilitating land surface surveying. This is consistent with the observation from the Asner’s study (Asner, 2001; Li et al., 2017). Our study suggested that the newly launched Landsat 8 OLI imagery satisfy the data requirements needed for land surface monitoring, although there existed the distinct spatio-temporal variances. To our knowledge, this study may be the first systematic analysis of CC in OLI imagery over China. The results and conclusions are of practical guidance for selecting cloudless or little-cloud Landsat OLI data for land surface changes and ecosystem dynamics research. Furthermore, the study provides useful information for future research on the availability and analysis of CC in Landsat images for a specific region, although our study mainly focused on the average value of CC percentage in the whole Landsat OLI footprints. Thus, the area of the Landsat OLI adjacent path/rows in WRS would influence the spatial analysis results in each intra-scenes scale to some extent. Following this, we will continue to analyze the CC percentage differences at the pixel level over China-Southeast Asia in the near future, especially with Landsat family sensors (TM, ETM+ and OLI).
As noted, the results show that the OLI sensor has the generally higher AP for high-quality images in China compared to reported some estimates in other studies (Goward et al., 2006; Li et al., 2017). The scene-based metadata analysis of spatio-temporal differences of AP is a prerequisite for regional remote sensing monitoring, especially for crop mapping such as paddy rice (Li et al., 2016) and wheat (Feng et al., 2014). It is worthwhile to mention that our study focused only on the OLI data CC, which is a global archive. Further analysis of other Landsat data (i.e., TM and ETM+) would help to comprehensively understand suitability and limitation of the historical Landsat data for land surface studies in the future.

The authors have declared that no competing interests exist.

[1]
Aplin P, 2004. Remote sensing: Land cover. Progress in Physical Geography, 28(2): 283-293.

[2]
Arvidson T, Gasch J, Goward S N, 2001. Landsat 7’s long-term acquisition plan: An innovative approach to building a global imagery archive. Remote Sensing of Environment, 78(1/2): 13-26.The Landsat-7 Long-Term Acquisition Plan (LTAP) automates the selection of Landsat scenes to periodically refresh a global archive of sunlit, substantially cloud-free land images. Its automatic nature reduces the workload on both operations and science office management, and makes the best use of limited resources both on board the satellite and in the ground-processing systems. The core of the LTAP is the definition of seasonality for each scene of interest, where seasonality is defined as the occurrence of change over time. When significant change is occurring, acquisition frequencies are high. Conversely, if little change is occurring, acquisition frequencies are low. This definition of seasonality was augmented with suggested acquisition frequencies from the various science community niches that have historically relied on Landsat data to support their research goals. Cloud climatology for each scene was determined, as monthly historical averages, so it could be compared during scheduling with the daily cloud-cover predictions to discriminate between scenes cloudier than usual (less worthy of acquisition) and scenes clearer than normal (more worthy of acquisition). A scheduler system executes the LTAP and builds the daily schedule by evaluating many factors for each scene, including seasonality, acquisition history, cloud cover, and availability of resources. The result of this evaluation is a list of the 鈥渂est鈥 250 scenes to be acquired that day. An important tool in the development of the scheduling system and in the evaluation of the LTAP performance is a modeling capability that allows 鈥渨hat if鈥 analyses to be made and proposed changes to the LTAP or the scheduling software to be evaluated. This modeling capability has proven key to understanding the impacts of potential changes in strategy or software. By considering cloud climatology and current cloud-cover predictions, Landsat 7 has acquired imagery with less cloud contamination than previous Landsat data. After 16 months of operations, a relatively cloud-free global archive exists that captures land-cover change where and when it occurs. Additional user requests have been low, suggesting users are finding suitable data already in the archive. Short-term science campaigns and special requests can be accommodated in the LTAP without perturbing the seasonality of the archive or exceeding the available resources.

DOI

[3]
Asner, G P, 2001. Cloud cover in Landsat observations of the Brazilian Amazon. International Journal of Remote Sensing, 22(18): 3855-3862.High spatial resolution Landsat imagery is employed in efforts to understand the impact of human activities on ecological, biogeochemical and atmospheric processes in the Amazon basin. The utility of Landsat multi-spectral data depends both on the degree to which surface properties can be estimated from the radiometric measurements and on the ability to observe the surface through the atmosphere. Clouds are a major obstacle to optical remote sensing of humid tropical regions, therefore cloud cover probability analysis is a fundamental prerequisite to land-cover change and Earth system process studies in these regions. This paper reports the results of a spatially explicit analysis of cloud cover in the Landsat archive of Brazilian Amazonia from 1984 to 1997. Monthly observations of any part of the basin are highly improbable using Landsat-like optical imagers. Annual observations are possible for most of the basin, but are improbable in northern parts of the region. These results quantify the limitations imposed by cloud cover to current Amazon land-cover change assessments using Landsat data. They emphasize the need for improved radar and alternative optical data fusion techniques to provide time-series analyses of biogeophysical properties for regional modelling efforts.

DOI

[4]
Cui Y P, Ning X J, Qin Y Cet al., 2016. Spatio-temporal change of agricultural hydrothermal conditions in China from 1951 to 2010. Journal of Geographical Sciences, 26(6): 643-657.Based on the daily observation data of 824 meteorological stations during 1951-2010 released by the National Meteorological Information Center, this paper evaluated the changes in the heat and moisture conditions of crop growth. An average value of ten years was used to analyze the spatio-temporal variation in the agricultural hydrothermal conditions within a 1 km2 grid. Next, the inter-annual changing trend was simulated by regression analysis of the agricultural hydrothermal conditions. The results showed that the contour lines for temperature and accumulated temperatures(the daily mean temperature 鈮0掳C) increased significantly in most parts of China, and that the temperature contour lines had all moved northwards over the past 60 years. At the same time, the annual precipitation showed a decreasing trend, though more than half of the meteorological stations did not pass the significance test. However, the mean temperatures in the hottest month and the coldest month exhibited a decreasing trend from 1951 to 2010. In addition, the 0掳C contour line gradually moved from the Qinling Mountains and Huaihe River Basin to the Yellow River Basin. All these changes would have a significant impact on the distribution of crops and farming systems. Although the mechanisms influencing the interactive temperature and precipitation changes on crops were complex and hard to distinguish, the fact remained that these changes would directly cause corresponding changes in crop characteristics.

DOI

[5]
Ding Y, 1961. Cultivation of Rice in China. Beijing: China Agriculture Press. (in Chinese)

[6]
Dong J W, Xiao X M, Chen B Qet al., 2013. Mapping deciduous rubber plantations through integration of PALSAR and multi-temporal Landsat imagery. Remote Sensing of Environment, 134: 392-402.61Phenology of deciduous rubber trees is captured in Landsat time series data.61Rubber plantation delineation by integrating PALSAR and Landsat imagery.61Geo-Referenced Field Photo Library supported algorithm training and validation.

DOI

[7]
Feng Z M, Yang L, Yang Y Z, 2014. Temporal and spatial distribution patterns of grain crops in the West Liaohe River Basin. Journal of Resources and Ecology, 5(3): 244-252.Here, we quantitatively determine temporal and spatial distribution characteristics of main grain crops in the West Liaohe River basin, Inner Mongolia, China, from 2000 to 2010 based on MODIS remote sensing data and NDVI time series information for the years 2000, 2005 and 2010. Phenological calendars and a decision-making tree extraction model were also used to obtain spatial distribution information of spring maize, spring wheat and soybean. We found that in 2010, the sown area of the main grain crops in the West Liaohe River basin was 11 965.08km, of which, the sown area for spring maize accounted for 92.28%and was concentrated in the lower reaches of the region. Spring wheat accounted for 3.14% and was mainly in the middle reaches. Soybean accounted for 4.58% and was predominantly in the upper reaches. From 2000 to 2005, the sown area of these grain crops in the West Liaohe River basin grew by 29.77%, mainly in the lower reaches: spring maize grew by 38.99%, spring wheat by 39.04% and soybean by 21.27%. From 2005 to 2010, growth in the sown area of these crops was slow (5.18% growth) and mainly in the lower reaches of the basin. The sown area of spring maize increased, but decreased for both spring wheat and soybean.

DOI

[8]
Foley J A, DeFries R, Asner G Pet al., 2005. Global consequences of land use. Science, 309: 570-574.Land use has generally been considered a local environmental issue, but it is becoming a force of global importance. Worldwide changes to forests, farmlands, waterways, and air are being driven by the need to provide food, fiber, water, and shelter to more than six billion people. Global croplands, pastures, plantations, and urban areas have expanded in recent decades, accompanied by large increases in energy, water, and fertilizer consumption, along with considerable losses of biodiversity. Such changes in land use have enabled humans to appropriate an increasing share of the planet9s resources, but they also potentially undermine the capacity of ecosystems to sustain food production, maintain freshwater and forest resources, regulate climate and air quality, and ameliorate infectious diseases. We face the challenge of managing trade-offs between immediate human needs and maintaining the capacity of the biosphere to provide goods and services in the long term.

DOI PMID

[9]
Ge Q S, Zheng J, Hao Z Xet al., 2016. Recent advances on reconstruction of climate and extreme events in China for the past 2000 years. Journal of Geographical Sciences, 26(7): 827-854.中国被突出的 monsoonal 气候在这个国家的东方区分,一个大陆人在西北的干旱气候和 Qinghai 西藏高原上的高地寒冷气候。因为中国文明的长历史,为在象许多自然档案一样的国家的全部上的气候变化有丰富、标明日期得好的记录片记录(例如,树戒指,冰核心,石笋, varved 湖沉积和珊瑚) 启用高分辨率的 paleoclimatic 重建。在这份报纸,我们在中国在最后 2000 年在气候和极端事件的重建考察最近的进展。在最后 10 年里,基于有宽空间范围的多代理,许多新重建在中国被出版了。这些重建使我们能象地区性的重建的无常一样在国内理解气候变化的特征。综合重建的温度结果证明在最后 2000 年的温暖的间隔发生在广告 1-200,广告 551-760,广告 951-1320,并且在广告 1921 以后,并且也证明冷间隔在广告 201-350,广告 441-530,广告 781-950,和广告 1321-1920。极端冷冬季,在 1500 和 1900 之间看,是比在 1950 以后的那些更经常的。在最近的全球温暖的上下文,地区性的热浪的紧张不能事实上超过在最后 2000 年看见的自然气候可变性。在中国的东方 monsoonal 区域,十、多十、一百年的摆动在降雨可变性被看见。当时为干旱 / 洪水的整体平均数越过所有冷时期表演的空间模式南方的分布, 25 向南关于干旱有一个 tri 杆模式

DOI

[10]
Goward S, Arvidson T, Williams Det al., 2006. Historical record of Landsat global coverage: Mission operations, NSLRSDA, and international cooperator stations. Photogrammetric Engineering & Remote Sensing, 72(10): 1155-1169.The long-term, 34 year record of global Landsat remote sensing data is a critical resource to study the Earth system and human impacts on this system. The National Satellite Land Remote Sensing Data Archive (NSLRSDA) is charged by public law to: “maintain a permanent, comprehensive Government archive of global Landsat and other land remote sensing data for long-term monitoring and study of the changing global environment” (U.S. Congress, 1992). The advisory committee for NSLRSDA requested a detailed analysis of observation coverage within the U.S. Landsat holdings, as well as that acquired and held by International Cooperator (IC) stations. Our analyses, to date, have found gaps of varying magnitude in U.S. holdings of Landsat global coverage data, which appear to reflect technical or administrative variations in mission operations. In many cases it may be possible to partially fill these gaps in U.S. holdings through observations that were acquired and are now being held at International Cooperator stations.

DOI

[11]
Hansen M C, Loveland T R, 2012. A review of large area monitoring of land cover change using Landsat data. Remote Sensing of Environment, 122: 66-74.Landsat data constitute the longest record of global-scale medium spatial resolution earth observation data. As a result, the current methods for large area monitoring of land cover change using medium spatial resolution imagery (10鈥50m) typically employ Landsat data. Most large area products quantify forest cover change. Forests are a comparatively easy cover type to map as well as a current focus of environmental monitoring concerning the global carbon cycle and biodiversity loss. Among existing change products, supervised or knowledge-based characterization methods predominate. Radiometric correction methods vary significantly, largely as a function of geographic/algorithmic scale. For instance, products created by mosaicking per scene characterizations do not require radiometric normalization. On the other hand, methods that employ a single index or classification model over an entire study area do require radiometric normalization. Temporal updating of cover change varies between existing products as a function of regional acquisition frequency, cloud cover and seasonality. With the Landsat archive opened for free access to terrain-corrected data, future product generation will be more data intensive. Per scene, interactive analyses will no longer be viable. Coupling free and open access to large data volumes with improved processing power will result in automated image pre-processing and land cover characterization methods. Such methods will need to leverage high-performance computing capabilities in advancing the land cover monitoring discipline. Robust validation efforts will be required to quantify product accuracies in determining the optimal change characterization methodologies.

DOI

[12]
Hao H M, Ren Z Y, 2009. Land Use/Land Cover Change (LUCC) and eco-environment response to LUCC in farming-pastoral zone, China. Journal of Integrative Agriculture, 8(1): 91-97.In order to understand land use/land cover changes(LUCC) and the eco-environment response to LUCC in farming-pastoral zone of the northern China during the recent twenty years,Baotou prefecture was selected as a case study area for investigation and quantitative evaluation.Technologies of remote sensing(RS),global positioning system(GPS),geographic information system(GIS),and other statistical methods were employed to implement.Results showed that:(1) During the recent twenty years,the areas of forest lands,grasslands and water were reduced,whereas the areas of other types were enlarged.Parts of forest lands,grasslands,and waters had become farmlands,and about 31.5% of the changed grasslands transferred into unused lands.The newly increased farmlands mainly came from grasslands and unused lands.And the newly increased construction lands mainly came from grasslands and farmlands.(2) Regional eco-environmental quality decreased by 12.6%,for which the land degradation(especially the meadow degeneration) and the developing of the cultivated land were mainly responsible,and their contributions to the regional eco-environment changes were 51.84 and 23.63% respectively.(3) The tendency of LUCC and the eco-environment response to LUCC displayed spatial heterogeneity.It can be concluded that the present agricultural production mode was not sustainable in farming-pastoral zone of northern China.Land degradation,especially meadow degradation induced by over-trampling and overgrazing,and developing of cultivated land were mainly responsible for regional eco-environment deterioration.Changing the cultivated land to forest or grass,however,can relieve deterioration of local eco-environment to some extents.And in the farming-pastoral zone in the northern China,evaluating regional eco-environment responses to LUCC was very necessary due to its fragile eco-environments.

DOI

[13]
Jin S M, Homer C G, Yang L Met al., 2013a. Automated cloud and shadow detection and filling using two-date Landsat imagery in the USA. International Journal of Remote Sensing, 34(5): 1540-1560.A simple, efficient, and practical approach for detecting cloud and shadow areas in satellite imagery and restoring them with clean pixel values has been developed. Cloud and shadow areas are detected using spectral information from the blue, shortwave infrared, and thermal infrared bands of Landsat Thematic Mapper or Enhanced Thematic Mapper Plus imagery from two dates (a target image and a reference image). These detected cloud and shadow areas are further refined using an integration process and a false shadow removal process according to the geometric relationship between cloud and shadow. Cloud and shadow filling is based on the concept of the Spectral Similarity Group (SSG), which uses the reference image to find similar alternative pixels in the target image to serve as replacement values for restored areas. Pixels are considered to belong to one SSG if the pixel values from Landsat bands 3, 4, and 5 in the reference image are within the same spectral ranges. This new approach was applied to five Landsat path/rows across different landscapes and seasons with various types of cloud patterns. Results show that almost all of the clouds were captured with minimal commission errors, and shadows were detected reasonably well. Among five test scenes, the lowest producer's accuracy of cloud detection was 93.9% and the lowest user's accuracy was 89%. The overall cloud and shadow detection accuracy ranged from 83.6% to 99.3%. The pixel-filling approach resulted in a new cloud-free image that appears seamless and spatially continuous despite differences in phenology between the target and reference images. Our methods offer a straightforward and robust approach for preparing images for the new 2011 National Land Cover Database production.

DOI

[14]
Jin S M, Yang L M, Danielson Pet al., 2013b. A comprehensive change detection method for updating the National Land Cover Database to circa 2011. Remote Sensing of Environment, 132: 159-175.The importance of characterizing, quantifying, and monitoring land cover, land use, and their changes has been widely recognized by global and environmental change studies. Since the early 1990s, three U.S. National Land Cover Database (NLCD) products (circa 1992, 2001, and 2006) have been released as free downloads for users. The NLCD 2006 also provides land cover change products between 2001 and 2006. To continue providing updated national land cover and change datasets, a new initiative in developing NLCD 2011 is currently underway. We present a new Comprehensive Change Detection Method (CCDM) designed as a key component for the development of NLCD 2011 and the research results from two exemplar studies. The CCDM integrates spectral-based change detection algorithms including a Multi-Index Integrated Change Analysis (MIICA) model and a novel change model called Zone, which extracts change information from two Landsat image pairs. The MIICA model is the core module of the change detection strategy and uses four spectral indices (CV, RCVMAX, dNBR, and dNDVI) to obtain the changes that occurred between two image dates. The CCDM also includes a knowledge-based system, which uses critical information on historical and current land cover conditions and trends and the likelihood of land cover change, to combine the changes from MIICA and Zone. For NLCD 2011, the improved and enhanced change products obtained from the CCDM provide critical information on location, magnitude, and direction of potential change areas and serve as a basis for further characterizing land cover changes for the nation. An accuracy assessment from the two study areas show 100% agreement between CCDM mapped no-change class with reference dataset, and 18% and 82% disagreement for the change class for WRS path/row p22r39 and p33r33, respectively. The strength of the CCDM is that the method is simple, easy to operate, widely applicable, and capable of capturing a variety of natural and anthropogenic disturbances potentially associated with land cover changes on different landscapes.

DOI

[15]
Jin T, Qing X Y, Huang L Y, 2016. Changes in grain production and the optimal spatial allocation of water resources in China. Journal of Resources and Ecology, 7(1): 28-35.Changes in grain production are decomposed and compared among nine major Chinese river basins for the sake of optimal water allocation. The results show that water-deficient northern China, especially the Songliao River Basin and Huai River Basin, contributed the greatest share of the total grain increment from 1995 to 2010. The Songliao River Basin achieved increased grain output largely by expanding multiple cropping, while the Huai River Basin achieved it mainly by improving the yield per unit area. With increased reliance on expanding irrigation and multi-cropping, most northern basins have high levels of agricultural water consumption, despite the rising share of corn, a lower water intensive crop. In contrast, over the same period the warm and humid south, traditionally a major rice-growing area, mostly experienced a sharp decline in rice cropping area and the Southeast Rivers Basin even reduced multiple cropping indexes, contributing to decreased agricultural water consumption. Implications of our findings and the need for tackling the imbalance of agricultural water use in grain production are discussed.

DOI

[16]
Ju J, Roy D P, 2008. The availability of cloud-free Landsat ETM plus data over the conterminous United States and globally. Remote Sensing of Environment, 112(3): 1196-1211.The U.S. Landsat satellite series provide the longest dedicated land remote sensing data record with a balance between requirements for localized high spatial resolution studies and global monitoring. As with any other optical wavelength satellite sensor, cloud contamination greatly compromises image usability for land surface studies. Additionally, selective scene acquisition due to payload, ground station and mission cost constraints further reduces Landsat image availability. Since the 1999 launch of the Landsat Enhanced Thematic Mapper Plus (ETM+) a Long-term Acquisition Plan (LTAP) has been used to anticipate user requests with the goal of annually refreshing a global daytime archive of cloud-free ETM+ data. This research evaluates the availability of cloud-free Landsat ETM+ data over the conterminous U.S. and globally using 3years of ETM+ cloud fraction metadata archived by the U.S. Landsat project. Landsat application requirements including obtaining at least one cloud-free observation in a year, a season, and two different seasons, or at least a pair of cloud-free observations occurring no more than 16, 32, 48, 64, and 80days apart within a year and season are considered. Probabilistic analyses indicate that over the conterminous U.S., land applications requiring at least one cloud-free observation in a year, a season, two different seasons, or at least two cloud-free observations occurring within any period of the year, are on average largely unaffected by cloud cover, except for certain Winter applications and cloudy scenes near the U.S.鈥揅anada border and the Great Lakes. Cloud becomes a constraint when at least two cloud-free observations are required from the same season over the conterminous U.S., especially when the separation between observations is restricted to short time intervals. Global applications requiring at least one cloud-free observation in a season, in two different seasons, and applications requiring at least two cloud-free observations in a year, are all severely affected by cloud and data availability constraints; and globally it is generally not practical to consider land applications that require at least two cloud-free observations in any season. Globally, only land applications requiring at least one cloud-free observation per year are largely unaffected by cloud cover and the reduced global ETM+ data availability. These results are specific only to the U.S. Landsat ETM+ archive; they suggest the need for an increased global Landsat acquisition rate for the current and future Landsat missions and/or the development of new approaches to mitigating cloud contamination in the U.S. global Landsat ETM+ archive.

DOI

[17]
Kovalskyy V, Roy D P, 2013. The global availability of Landsat 5 TM and Landsat 7 ETM+ land surface observations and implications for Global 30 m landsat data product generation. Remote Sensing of Environment, 130: 280-293.78 The impact of clouds and SLC_OFF on availability of clear Landsat land observations. 78 Combing Landsat 5 TM and Landsat 7 ETM+ data streams is advantages. 78 Compare to ETM+ alone both sensors provide up to 14.4% higher monthly land coverage. 78 36 months of combined Landsat 5 and 7 data can support 30m global land cover mapping.

DOI

[18]
Kovalskyy V, Roy D P, 2015. A one year Landsat 8 conterminous United States study of cirrus and non-cirrus clouds. Remote Sensing, 7(1): 564-578.The first year of available Landsat 8 data over the conterminous United States (CONUS), composed of 11,296 acquisitions sensed over more than 11 thousand million 30 m pixel locations, was analyzed comparing the spatial and temporal incidence of 30 m cloud and cirrus states available in the standard Landsat 8 Level 1 product suite. This comprehensive data analysis revealed that on average over a year of CONUS observations (i) 35.9% were detected with high confidence cloud, with spatio-temporal patterns similar to those observed by previous Landsat 5 and 7 cloud analyses; (ii) 28.2% were high confidence cirrus; (iii) 20.1% were both high confidence cloud and high confidence cirrus; and (iv) 6.9% were detected as high confidence cirrus but low confidence cloud. The results illustrate the potential of the 30 m cloud and cirrus states available in the standard Landsat 8 Level 1 product suite but imply that the historical CONUS Landsat archive has about 7% of undetected cirrus contaminated pixels. Systematic cloud detection commission errors over a minority of highly reflective exposed soil/sand surfaces were found and it is recommended that caution be taken when using the currently available Landsat 8 cloud data over similar surfaces.

DOI

[19]
Laborde H, Douzal V, Ruiz Piña, H Aet al., 2017. Landsat-8 cloud-free observations in wet tropical areas: A case study in South East Asia. Remote Sensing Letters, 8(6): 537-546.The recent launch of the Landsat-8 satellite, in parallel to the opening of 40 years of archives issued from the Landsat program, has created new opportunities for the design of large-scale initiatives for environmental monitoring at moderate spatial resolution. However, the availability of cloud-free observations stays a major limiting factor, especially in wet tropical areas like South East Asia (SEA). Based on the analysis of one year of Landsat-8 data acquired for the 69 path/row combinations that intersect SEA with the Fmask algorithm, we studied the temporal granularity of cloud-free observations. Our results show that the annual production of two quasi-complete composited maps for SEA seems achievable when 090004greenness090005 and 090004brownness090005 are close to their maximum, offering relevant opportunities for scientists and stakeholders involved in environment, biodiversity or health issues.

DOI

[20]
Li P, Feng Z M, Jiang L Get al., 2012. Changes in rice cropping systems in the Poyang Lake Region, China during 2004-2010. Journal of Geographical Sciences, 22(4): 653-668.Rice cropping systems not only characterize comprehensive utilization intensity of agricultural resources but also serve as the basis to enhance the provision services of agro-ecosystems.Yet,it is always affected by external factors,like agricultural policies.Since 2004,seven consecutive No.1 Central Documents issued by the Central Government have focused on agricultural development in China.So far,few studies have investigated the ef-fects of these policies on the rice cropping systems.In this study,based upon the long-term field survey information on paddy rice fields,we proposed a method to discriminate the rice cropping systems with Landsat data and quantified the spatial variations of rice cropping systems in the Poyang Lake Region(PLR),China.The results revealed that:(1) from 2004 to 2010,the decrement of paddy rice field was 46.76 km2 due to the land use change.(2) The temporal dynamics of NDVI derived from Landsat historical images could well characterize the temporal development of paddy rice fields.NDVI curves of single cropping rice fields showed one peak,while NDVI curves of double cropping rice fields displayed two peaks an-nually.NDVI of fallow field fluctuated between 0.15 and 0.40.NDVI of the flooded field during the transplanting period was relatively low,about 0.20卤0.05,while NDVI during the period of panicle initiation to heading reached the highest level(above 0.80).Then,several temporal windows were determined based upon the NDVI variations of different rice cropping systems.(3) With the spatial pattern of paddy rice field and the NDVI threshold within optimum tem-poral windows,the spatial variation of rice cropping systems was very obvious,with an in-creased multiple cropping index of rice about 20.2% from 2004 to 2010.The result indicates that agricultural policies have greatly enhanced the food provision services in the PLR,China.

DOI

[21]
Li P, Feng Z M, Xiao C W, 2017. Acquisition probability differences in cloud coverage of the available Landsat observations over mainland Southeast Asia from 1986 to 2015. International Journal of Digital Earth, 10: 1-14. doi: 10.1080/17538947.2017.1327619.Digital Earth has seen great progress during the last 19 years. When it entered into the era of big data, Digital Earth developed into a new stage, namely one characterized by 090004Big Earth Data090005, confronting new challenges and opportunities. In this paper we give an overview of the development of Digital Earth by summarizing research achievements and marking the milestones of Digital Earth090005s development. Then, the opportunities and challenges that Big Earth Data faces are discussed. As a data-intensive scientific research approach, Big Earth Data provides a new vision and methodology to Earth sciences, and the paper identifies the advantages of Big Earth Data to scientific research, especially in knowledge discovery and global change research. We believe that Big Earth Data will advance and promote the development of Digital Earth.

DOI

[22]
Li P, Jiang L G, Feng Z Met al., 2016. Mapping rice cropping systems using Landsat-derived Renormalized Index of Normalized Difference Vegetation Index (RNDVI) in the Poyang Lake Region, China. Frontiers of Earth Science, 10(2): 303-314.印射在多重收割区域与光形象收割系统的米饭由于云污染和数据可获得性是挑战性的;有一个减少的数据要求的一个基于物候学的算法的开发是必要的。在这研究,规范的差别植被索引(RNDVI ) 的导出 Landsat 的重新使正常化的索引基于 NDVI 单身者并且早在珍视的二个时间的窗口被建议(或迟了) 米饭显示器逆变化,然后适用区别收割系统的米饭。波伊昂·莱克区域(PLR ) ,由收割米饭(SCR,或单个米饭) 和两倍收割米饭(DCR,包括的早米饭和迟了的米饭) 的单身者的一个典型收割系统描绘了,作为一个严峻的区域被选择。结果显示出数据在八点从 Landsat 时间系列导出到十六天俘获的那 NDVI 稻米饭的时间的发展。在 SCR 和 DCR 的 NDVI 价值相反地在变化的重叠生长时期期间有二个关键 phenological 阶段,也就是,早米饭的成熟阶段和象成熟一样的单个米饭的成长阶段近来单个米饭和成长阶段上演米饭。NDVI 在二扇时间的窗户,明确地早的 8 月和早 10 月中源于场面,被用来为区别在 PLR 的新辟的低地区域收割系统的米饭构造 RNDVI,中国。有地面真相数据的比较显示高分类精确性。RNDVI 途径由于在二个时间的窗口之间的米饭生长的差别加亮 NDVI 值的反的变化。当它仅仅需要区分候选人米饭类型是否在生长的时期,这使收割系统的米饭的辨别直接(RNDVI 0 ) 。

DOI

[23]
Liu J Y, Deng X Z, 2010. Progress of the research methodologies on the temporal and spatial process of LUCC. Chinese Science Bulletin, 55(14): 1354-1362.

DOI

[24]
Liu J Y, Kuang W H, Zhang Z Xet al., 2014. Spatiotemporal characteristics, patterns, and causes of land-use changes in China since the late 1980s. Journal of Geographical Sciences, 24(2): 195-210.Land-use/land-cover 变化(LUCC ) 有连接到人和自然相互作用。瓷器 Land-Use/cover 数据集(CLUD ) 从 1980 年代末在 5 年的间隔定期被更新到 2010,与基于 Landsat TMETM+ 图象的标准过程。陆地使用动态区域化方法被建议分析主要陆地使用变换。在国家规模的陆地使用变化的空间与时间的特征,差别,和原因然后被检验。主要调查结果如下被总结。越过中国的陆地使用变化(LUC ) 在最后 20 年(19902010 ) 里在空间、时间的特征显示了一个重要变化。农田变化的区域在南方减少了并且在北方,而是仍然是的全部的区域增加了几乎未改变。回收农田从东北被转移到西北。布满建筑物陆地很快膨胀了,主要在东方被散布,并且逐渐地展开到中央、西方的中国。树林首先减少了,然后增加但是荒芜的区域是反面。草地继续减少。在中国的 LUC 的不同空间模式被发现在之间迟了第 20 世纪并且早第 21 世纪。原版 13 个 LUC 地区在一些地区被边界的变化由 15 个单位代替。包括的这些变化(1 ) 的主要空间特征加速的扩大布满建筑物在 Huang-Huai-Hai 区域,东南的沿海的区域,长江的中流区域,和四川盆登陆;(2 ) 从东北中国和东方内部蒙古在北方转移了陆地开垦到绿洲在西北中国的农业区域;(3 ) 从在到稻的东北中国的喂雨的农田的连续转变回答;并且(4 ) 为在内部蒙古,黄土高原,和西南的多山的区域的南部的农业牧剧的交错群落的格林工程的谷物的有效性。在最后二十年,尽管在北方的气候变化在农田影响了变化,政策规定和经济驱动力仍然是越过中国的 LUC 的主要原因。在第 21 世纪的第一十年期间,在陆地使用模式驾驶了变化的人为的因素从单程的陆地开发转移了强调到开发和保存。动态区域化方法被用来在单位的 zoning 边界,地区的内部特征,和生长和减少的空间17

DOI

[25]
Liu J Y, Shao Q Q, Yan X D.et al., 2016. The climatic impacts of land use and land cover change compared among countries. Journal of Geographical Sciences, 26(7): 889-903.Land use and land cover change (LULCC) strongly influence regional and global climate by combining both biochemical and biophysical processes. However, the biophysical process was often ignored, which may offset the biogeochemical effects, so measures to address climate change could not reach the target. Thus, the biophysical influence of LULCC is critical for understanding observed climate changes in the past and potential scenarios in the future. Therefore, it is necessary to identify the mechanisms and effects of large-scale LULCC on climate change through changing the underlying surface, and thus the energy balance. The key scientific issues on understanding the impacts of human activities on global climate that must be addressed including: (1) what are the basic scientific facts of spatial and temporal variations of LULCC in China and comparative countries? (2) How to understand the coupling driving mechanisms of human activities and climate change on the LULCC and then to forecasting the future scenarios? (3) What are the scientific mechanisms of LULCC impacts on biophysical processes of land surface, and then the climate? (4) How to estimate the contributions of LULCC to climate change by affecting biophysical processes of land surface? By international comparison, the impacts of LULCC on climate change at the local, regional and global scales were revealed and evaluated. It can provide theoretical basis for the global change, and have great significance to mitigate and adapt to global climate changes.

DOI

[26]
Liu R G, Liu Y, 2013. Generation of new cloud masks from MODIS land surface reflectance products. Remote Sensing of Environment, 133: 21-37.78 An inflexion-based approach to generate cloud mask from MOD09 products 78 An approach to evaluating cloud masks with blue reflectance and NDVI 78 The new cloud masks are valuable for production of downstream land products.

DOI

[27]
Liu X N, Feng Z M, Jiang L Get al., 2013. Rubber plantation and its relationship with topographical factors in the border region of China, Laos and Myanmar. Journal of Geographical Sciences, 23(6): 1019-1040.Rubber plantation is the major land use type in Southeast Asia. Monitoring the spatial-temporal pattern of rubber plantation is significant for regional land resource development, eco-environmental...

DOI

[28]
Liu Y J, Yang Y Z, Feng Z M, 2007. The change of the main regions for China’s food grain production and its implications. Resources Science, 29(2): 8-14. (in Chinese)

[29]
Qi W, Liu S H, Zhao M Fet al., 2016. China’s different spatial patterns of population growth based on the “Hu Line”. Journal of Geographical Sciences, 26(11): 1611-1625.

DOI

[30]
Qin Z, Karnieli A, Berliner P, 2001. A mono-window algorithm for retrieving land surface temperature from Landsat TM data and its application to the Israel-Egypt border region. International Journal of Remote Sensing, 22(18): 3719-3746.

DOI

[31]
Ren Z Y, Wang L X, 2007. Spatio-temporal differentiation of landscape ecological niche in western ecological frangible region: A case study of Yan’an region in northwestern China. Journal of Geographical Sciences, 17(4): 479-486.In this study, we attempt to put forward a conception of landscape ecological niche, enlightened by international scholars on extending the ecological niche theory from spatial niche to functional niche. That is helpful for comprehensively appraising landscape spatial patterns and ecological functions, also, presents a new method for analyzing landscape features from multidimensional aspects. The practice process is demonstrated by taking Yan鈥檃n region in northwestern China as a case. Firstly, the indices system including spatial attribute and functional attribute is established for assessing landscape ecological niche. Additionally, two-dimensional figures are drawn for comparing the spatio-temporal features of landscape ecological niche in 1987 and 2000 among the 13 administrative counties. The results show that from 1987 to 2000, towards Yan鈥檃n region, spatial attribute value of landscape ecological niche changes from 1.000 to 1.178 with an obvious increment, and functional attribute value changes from 0.989 to 1.069 with a little increment, both of which enhance the regional landscape ecological niche. Towards each county, spatial attribute value of landscape ecological niche increases to different extent while functional attribute value changes dissimilarly with an increment or a decrement.

DOI

[32]
Sano E E, Ferreira L G, Asner G Pet al., 2007. Spatial and temporal probabilities of obtaining cloud-free Landsat images over the Brazilian tropical savanna. International Journal of Remote Sensing, 28(12): 2739-2752.Remotely sensed data are the best and perhaps the only possible way for monitoring large‐scale, human‐induced land occupation and biosphere‐atmosphere processes in regions such as the Brazilian tropical savanna (Cerrado). Landsat imagery has been intensively employed for these studies because of their long‐term data coverage (>30 years), suitable spatial and temporal resolutions, and ability to discriminate different land‐use and land‐cover classes. However, cloud cover is the most obvious constraint for obtaining optical remote sensing data in tropical regions, and cloud cover analysis of remotely sensed data is a requisite step needed for any optical remote sensing studies. This study addresses the extent to which cloudiness can restrict the monitoring of the Brazilian Cerrado from Landsat‐like sensors. Percent cloud cover from more than 35 500 Landsat quick‐looks were estimated by the K‐means unsupervised classification technique. The data were examined by month, season, and El Ni09o Southern Oscillation event. Monthly observations of any part of the biome are highly unlikely during the wet season (October–March), but very possible during the dry season, especially in July and August. Research involving seasonality is feasible in some parts of the Cerrado at the temporal satellite sampling frequency of Landsat sensors. There are several limitations at the northern limit of the Cerrado, especially in the transitional area with the Amazon. During the 1997 El Ni09o event, the cloudiness over the Cerrado decreased to a measurable but small degree (5% less, on average). These results set the framework and limitations of future studies of land use/land cover and ecological dynamics using Landsat‐like satellite sensors.

DOI

[33]
Steven M D, Malthus T J, Baret Fet al., 2003. Intercalibration of vegetation indices from different sensor systems. Remote Sensing of Environment, 88(4): 412-422.Spectroradiometric measurements were made over a range of crop canopy densities, soil backgrounds and foliage colour. The reflected spectral radiances were convoluted with the spectral response functions of a range of satellite instruments to simulate their responses. When Normalised Difference Vegetation Indices (NDVI) from the different instruments were compared, they varied by a few percent, but the values were strongly linearly related, allowing vegetation indices from one instrument to be intercalibrated against another. A table of conversion coefficents is presented for AVHRR, ATSR-2, Landsat MSS, TM and ETM+, SPOT-2 and SPOT-4 HRV, IRS, IKONOS, SEAWIFS, MISR, MODIS, POLDER, Quickbird and MERIS (see Appendix A for glossary of acronyms). The same set of coefficients was found to apply, within the margin of error of the analysis, for the Soil Adjusted Vegetation Index SAVI. The relationships for SPOT vs. TM and for ATSR-2 vs. AVHRR were directly validated by comparison of atmospherically corrected image data. The results indicate that vegetation indices can be interconverted to a precision of 1鈥2%. This result offers improved opportunities for monitoring crops through the growing season and the prospects of better continuity of long-term monitoring of vegetation responses to environmental change.

DOI

[34]
Teillet P M, Fedosejevs G, Thome K J Eet al., 2007 Impacts of spectral band difference effects on radiometric cross-calibration between satellite sensors in the solar-reflective spectral domain. Remote Sensing of Environment, 110(3): 393-409.In order for quantitative applications to make full use of the ever-increasing number of Earth observation satellite systems, data from the various imaging sensors involved must be on a consistent radiometric scale. This paper reports on an investigation of radiometric calibration errors due to differences in spectral response functions between satellite sensors when attempting cross-calibration based on near-simultaneous imaging of common ground targets in analogous spectral bands, a commonly used post-launch calibration methodology. Twenty Earth observation imaging sensors (including coarser and higher spatial resolution sensors) were considered, using the Landsat solar reflective spectral domain as a framework. Scene content was simulated using spectra for four ground target types (Railroad Valley Playa, snow, sand and rangeland), together with various combinations of atmospheric states and illumination geometries. Results were obtained as a function of ground target type, satellite sensor comparison, spectral region, and scene content. Overall, if spectral band difference effects (SBDEs) are not taken into account, the Railroad Valley Playa site is a “good” ground target for cross calibration between most but not all satellite sensors in most but not all spectral regions investigated. “Good” is defined as SBDEs within ± 3%. The other three ground target types considered (snow, sand and rangeland) proved to be more sensitive to uncorrected SBDEs than the RVPN site overall. The spectral characteristics of the scene content (solar irradiance, surface reflectance and atmosphere) are examined in detail to clarify why spectral difference effects arise and why they can be significant when comparing different imaging sensor systems. Atmospheric gas absorption features are identified as being the main source of spectral variability in most spectral regions. The paper concludes with recommendations on spectral data and tools that would facilitate cross-calibration between multiple satellite sensors.

DOI

[35]
U.S. Geological Survey (USGS), 2016. Landsat 8 (L8) Data Users Handbook. Available online: https://landsat.usgs.gov/landsat-8-l8-data-users-handbook (accessed on 18 April 2017).

[36]
Wu Y Y, Li S Y, Yu S X, 2016. Monitoring urban expansion and its effects on land use and land cover changes in Guangzhou city, China. Environmental Monitoring Assessment, 188(541), 609-621.Phthalates are endocrine-disrupting chemicals which affect endocrine system by bio-accumulation in aquatic organisms and produce adverse health effects in aquatic organisms as well as human beings, wh

DOI PMID

[37]
Whitcraft A K, Vemiote E F, Becker-Reshef Iet al., 2015. Cloud cover throughout the agricultural growing season: Impacts on passive optical earth observations. Remote Sensing of Environment, 156: 438-447.61MODIS cloud QA is analyzed over croplands during different phenological periods.61Cloud cover presence & amount vary throughout agricultural growing season (AGS).61Early-to-mid AGS – an important period for monitoring – has most pervasive clouds.61Many areas with extensive cultivation coincide with pervasive cloud cover.61This analysis aids in data acquisition planning for global agriculture monitoring.

DOI

[38]
Zhu Z, Woodcock C E, 2012. Object-based cloud and cloud shadow detection in Landsat imagery. Remote Sensing of Environment, 118: 83-94.78 A new method for automated cloud and cloud shadow detection in Landsat images. 78 It is the result of combining past approaches and a new object-based approach. 78 It is an improvement over the traditional ACCA cloud algorithm. 78 The average Fmask cloud overall accuracy is 96.4%.

DOI

Outlines

/