EN中文

Journal of Geographical Sciences >

2024 , Vol. 34 >Issue 2: 329 - 354

DOI: https://doi.org/10.1007/s11442-024-2207-1

Research Articles

Land cover change and its response to water level around Tonle Sap Lake in 1988-2020

  • ZHANG Jing , 1, 2 ,
  • MA Kai 1, 2 ,
  • FAN Hui 1, 2 ,
  • HE Daming , 1, 2, *
Expand
  • 1. Yunnan Key Laboratory of International Rivers and Transboundary Eco-Security, Kunming 650091, China
  • 2. Institute of International Rivers and Eco-Security, Yunnan University, Kunming 650091, China
*He Daming (1958-), PhD and Professor, specialized in hydroecology and hydrogeography. E-mail:

Zhang Jing (1986-), PhD, specialized in ecological security and environmental remote sensing. E-mail:

Received date: 2023-03-27

  Accepted date: 2023-11-09

  Online published: 2024-02-06

Supported by

National Key Research and Development Program of China(2016YFA0601600)

Yunnan Scientist Workstation for Daming He International River Research(KXJGZS-2019-005)

Fold

Abstract

The transboundary influence of environmental change is a critical issue in the Lancang-Mekong region. As the largest river-connected lake in the lower Mekong, the ecological change and influence of Tonle Sap Lake have received widespread attention and discussion, especially after 2008, when the hydrological regime of the Lancang-Mekong River mainstream underwent distinct changes. However, the linkage and coupling mechanism between the lake riparian environment and mainstream water level change are still unclear. In this study, the interannual spatiotemporal changes in land cover in the Tonle Sap Lake riparian zone (TSLRZ) and their relationship with mainstream water levels were analysed. The results showed that the expansion of farmland was the most notable change in 1988-2020. After 2008, the land cover changes intensified, manifested as accelerated farmland expansion, intensified woodland fragmentation and significant water body shrinkage. Furthermore, the responses of the water body, degraded land, wasteland and grassland areas to the mainstream water levels weakened after 2008. Evidently, the land cover changes in the TSLRZ in the last 30 years were less related to the mainstream water level change than to local reclamation and logging. These results can offer a new scientific basis for the transboundary influence analysis of hydrological change.

Key words： riparian zone land cover change; mainstream water level change and transboundary influence; Tonle Sap Lake; Lancang-Mekong River

Cite this article

ZHANG Jing , MA Kai , FAN Hui , HE Daming . Land cover change and its response to water level around Tonle Sap Lake in 1988-2020[J]. Journal of Geographical Sciences, 2024 , 34(2) : 329 -354 . DOI: 10.1007/s11442-024-2207-1

1 Introduction

The transboundary influence of environmental change is a critical issue in the context of global change, especially in the Lancang-Mekong River Basin, which flows through six countries, including China, Myanmar, Laos, Thailand, Cambodia and Vietnam (He et al., 2016; Wang et al., 2021; Zhong et al., 2021). Tonle Sap Lake (abbreviated as TSL) in Cambodia, connected to the Lancang-Mekong River mainstream by the Tonle Sap River, is the largest freshwater lake in Southeast Asia and the largest fishery base in Asia, and it is an important habitat for endangered and rare species, with very important ecological and economic value (Lamberts, 2008; Ziv et al., 2012).
In recent decades, with increasingly obvious global climate change and intensified human activities, the ecological environment of the TSL area has deteriorated, showing effects such as lake shrinkage, rare species habitat destruction and biodiversity loss and receiving widespread attention and discussion (Arias, 2013; Lin and Qi, 2017; Pokhrel et al., 2018; Mahood et al., 2020; Morovati et al., 2023). In this area, the ecological changes are directly affected by local climate change and human activities and are also related to the hydrological regime (such as water level and runoff) of the mainstream since the majority of the supply of lake water comes from the mainstream on an annual scale (the supply proportion may fluctuate over time) (Kummu et al., 2014; Ji et al., 2018; Chua et al., 2022; Morovati et al., 2023). Therefore, the attribution of the ecological changes in the TSL area is extremely complex and sensitive.
Land cover changes affect resource availability, environmental conditions, habitat structure and spatial configuration, closely linking human socio-economic activities with natural ecological processes (Lambin et al., 2001; Turner et al., 2007; Sterling et al., 2013; He et al., 2022; Wang et al., 2022). The Tonle Sap Lake riparian zone (abbreviated as TSLRZ) is flat and broad and has various land cover types (JICA, 1999; Eng and Ouch, 2006; Mahood et al., 2020). Compared with the instantaneous change in the lake water body (indicated by the permanent water body and its associated water body areas, abbreviated as LWB), the land cover changes in the TSLRZ indicate a more stable and comprehensive manifestation of various impacts, including local human activities and mainstream water levels, which can more accurately represent the status and quality of the eco-environment. Thus, the analysis of the relationship between the land cover change in the TSLRZ and the mainstream water level change can further reveal their hydroecological linkage and coupling mechanism.
To date, there have been numerous studies on the hydrological regime changes of the Lancang-Mekong River (abbreviated as LMR), especially after the reservoir impoundment of the Xiaowan mega dam in 2008, which was the first mega dam in the mainstream of Lancang River (upper Mekong) (Li et al., 2017; Binh et al., 2020; Yun et al., 2020; Guan and Zheng, 2021; Lu and Chua, 2021; Chua et al., 2022; Dang et al., 2022; Morovati et al., 2023). Most of these studies showed that the hydrological regime of the mainstream experienced distinct changes after 2008, indicating that the water level or flow increased in the dry season and decreased in the wet season, although whether the dominant influencing factor is climate change or mega dam operation remains controversial. Moreover, studies on the changes in the various land cover types in the TSL area have made some progress (Zhan et al., 2002; Trung et al., 2013; Kim et al., 2019; Zhao et al., 2019; Huang et al., 2020). In particular, the studies by Meng et al. (2018), Mahood et al. (2020) and Chen et al. (2022) focused on these changes in the area around the TSLRZ (abbreviated as A-TSLRZ) based on Landsat images or European Space Agency Climate Change Initiative (ESA CCI) land cover products, which have temporal scales of 26 years, 5 years and 9 years, respectively. These previous studies showed that the area of farmland increased and other land cover types generally shrank in the A-TSLRZ from 1990 to 2016 (Meng et al., 2018), from 1993 to 2018 (Mahood et al., 2020) and from 1992 to 2019 (Chen et al., 2022), and they further discussed the influence of local precipitation, reclamation and logging on land cover changes. In addition, since 2008, scholars have paid more attention to the relationship between the land cover change in the TSL area and the hydrological regime of the mainstream and the lake (Arias, 2013; Trung et al., 2013; Ji et al., 2018; Wang et al., 2020; Gu et al., 2021). For instance, Trung et al. (2013) analysed the seasonal change in these land cover types (forest, shrub, grass and agriculture) in the northwestern part of the TSL during the period from 2007 to 2010 based on Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) data and Advanced Land Observation Satellite (ALOS) data and its response to the lake water level. Arias (2013) proposed a spatial modelling framework to link the habitat, flood duration and animal species in the A-TSLRZ. Ji et al. (2018) explored the relationship between the LWB and mainstream runoff during the period from 2000 to 2014 at an annual scale based on MODIS images. Gu et al. (2021) discussed the relationship between the LWB and the mainstream water level during the period from 1987 to 2017 at monthly and annual scales based on Landsat images.
In these studies, the overall characteristics of the land cover changes in the TSLRZ and the relationship between the change in the LWB and the hydrological regime of the mainstream were clarified, providing a great foundation for our study. However, the time scales for the change analyses of the other land cover types excluding water bodies in these studies were relatively broad due to different research objectives, while the water levels of the TSL and the LMR mainstream varied and fluctuated greatly every year. On different time scales, the relationship between the two is difficult to clarify and still unclear. In addition, examining broad time scales makes it difficult to understand the detailed spatiotemporal changes in land cover in the TSLRZ, especially after 2008, when the hydrological regime of the mainstream underwent distinct changes.
In this study, we combined land cover classification maps extracted from Landsat images and hydrologic measurements to study the changes in land cover and its major driving factors in the TSLRZ, with two main objectives: (1) to create an annual scale land cover classification dataset and document interannual spatiotemporal changes in land cover in the TSLRZ during the period from 1988 to 2020, focusing on the changes after 2008; and (2) to clarify the relationship between these land cover changes and the water level of the LMR mainstream.

2 Materials and methods

2.1 Study area

Tonle Sap Lake is located on the alluvial plain of the lower Lancang-Mekong River (Figure 1). The lake displays a distinct dichotomy between dry and wet seasons due to the influence of the tropical monsoon climate and the changes in the recharge of the LMR (Tangdamrongsub et al., 2016; Siev et al., 2018; Morovati et al., 2023). During the wet season, the water level of the LMR is higher than that of the TSL, reversing the flow in the Tonle Sap River and discharging into the lake, which results in a maximum lake area greater than 15,000 km2. During the dry season, the water level of the LMR becomes lower, and the lake water pours into the river when the water level of the river falls below that of the lake, which results in a minimum lake area of less than 2500 km2. The annual mean temperature of the lake area is 28-33℃, and the annual precipitation is between 1300 and 1500 mm (Siev et al., 2018; Daly et al., 2020).
Figure 1 Location and spatial pattern of the study area. The false-colour maps of the Tonle Sap Lake riparian zone (TSLRZ) are composed of the 5, 4 and 3 bands of Landsat images, which were collected in 1988 and 2020. The extent of Tonle Sap Lake (permanent water body) was extracted from Landsat images during 1988-2020.
The TSLRZ (Figure 1) includes the area from the lakeside to the lake road, with a total area of approximately 15,149.72 km2 (Figure 1). There are three Biosphere Reserve core areas belonging to the United Nations Educational, Scientific and Cultural Organization Programme on Man and the Biosphere (UNESCO, 2006) and one Ramsar site defined under the Ramsar Convention (RAMSAR, 2010) in the TSLRZ. The land cover types in the study area mainly include woodland, wasteland, water body, and grassland (JICA, 1999; Eng and Ouch, 2006; Mahood et al., 2020; Chen et al., 2022) and man-made land use, such as rice-growing farmland and fragmented woodland, due to human interference (ADB, 2019; Mahood et al., 2020; Chen et al., 2022). In recent years, the TSL area has been vigorously reclaimed to increase rice production, with the use of high-yielding rice varieties and an expansion of irrigation infrastructure (ADB, 2019; Mahood et al., 2020). In addition, large net forest loss is occurring in this area due to logging and agricultural farming (Chen et al., 2022).

2.2 Data

(1) Landsat dataset
The surface reflectance (SR) products of Landsat images, with a spatial resolution of 30 m, were used for land cover classification in the TSLRZ during the period from 1988 to 2020. These products were generated based on Landsat-5 TM, Landsat-7 ETM+ and Landsat-8 OLI images. All products were subjected to atmospheric correction and geometric fine correction, and the cloud cover of a single-scene image was less than 5%. The 78 SR products were downloaded from the USGS website (https://www.usgs.gov/land-resources/nli/landsat). The clear images mainly appeared in January, February, March and December in the TSLRZ due to cloudy and rainy weather throughout the year. To ensure that the various land cover types were identifiable and that the acquired data had similar hydrological and phenological conditions in each year, land cover classification in our study was performed using images from January to March in each year. Notably, we did not gather clear images covering the entire study area for all the seven years (1990, 1997, 2004, 2008, 2010, 2012 and 2013) because the cloud cover was too high. Three Landsat images covered the entire study area, with paths/rows of 127/051, 126/051, and 126/052. Specific information on the satellite data is listed in Table 1.
Table 1 Specific information on the satellite data collected for land cover classification
Year Image name, Cloud cover (%) Image name, Cloud cover (%) Image name, Cloud cover (%)
1988 LT51270511988044BKT01, 0.02 LT05L1TP12605119880222, 0.99 LT05L1TP12605219880309, 0.35
1989 LT41270511989054XXX02, 0.15 LT05L1TP12605119890224, 1.33 LT05L1TP12605219890224, 0.04
1991 LT51270511991036BKT00, 0.14 LT05L1TP12605119910129, 0.01 LT05L1TP12605219910129, 0
1992 LT51270511992087BKT01, 1.39 LT05L1TP12605119920304, 0.01 LT05L1TP12605219920304, 0.01
1994 LT51270511994028BKT01, 0.01 LT05L1TP12605119940105, 1.31 LT05L1GS12605219940121, 0.46
1995 LT51270511995031BKT00, 0 LT05L1TP12605119950209, 0.03 LT05L1TP12605219950209, 0
1996 LT51270511996002CLT00, 0.03 LT05L1TP12605119960127, 0.39 LT05L1TP12605219960127, 0
1998 LT51270511998007BKT00, 0.04 LT05L1TP12605119980321, 0.59 LT05L1TP12605219980305, 2
1999 LT51270511999026BKT00, 0.4 LT05L1TP12605119990119, 0.88 LT05L1TP12605219990220, 0
2000 LT51270512000013BKT00, 1.08 LT05L1TP12605120000326, 1.74 LT05L1TP12605220000326, 0
2001 LT51270512001047BKT00, 0.07 LT05L1TP12605120010124, 1.88 LT05L1TP12605220010124, 0
2002 LE71270512002010SGS00, 0.02 LE71260512002051BKT00, 0.02 LE07L1TP12605220020220, 0
2003 LE71270512003029SGS00, 0.01 LE71260512003038SGS00, 0.31 LE07L1TP12605220030207, 0
2005 LT51270512005042BKT00, 0.7 LT05L1TP12605120050119, 0.14 LT05L1TP12605220050119, 0.02
2006 LT51270512006045BKT00, 2.02 LT05L1TP12605120060207, 0.14 LT05L1TP12605220060207, 0
2007 LT51270512007032BKT00, 0.01 LT05L1TP12605120070125, 0.13 LT05L1TP12605220070210, 0.07
2009 LT05L1TP12705120090121, 0.98 LT05L1TP12605120090114, 0.12 LT05L1TP12605220090114, 0
2011 LT05L1TP12705120110127, 1.57 LT05L1TP12605120110120, 6.18 LT05L1TP12605220110120, 3.12
2014 LC81270512014035LGN00, 2.16 LC81260512014028LGN00, 0.09 LC08L1TP12605220140128, 0.17
2015 LC81270512015022LGN00, 0.01 LC81260512015015LGN00, 0 LC08L1TP12605220150115, 0.01
2016 LC81270512016089LGN00, 3.67 LC81260512016034LGN00, 0 LC08L1TP12605220160203, 0
2017 LC08L1TP12705120170212, 0.04 LC08L1TP12605120170205, 0.18 LC08L1TP12605220170205, 0.78
2018 LC08L1TP12705120180215, 0.22 LC08L1TP12605120180312, 2.63 LC08L1TP12605220180312, 1.12
2019 LC08L1TP12705120190117, 0.01 LC08L1TP12605120190211, 0.07 LC08L1TP12605220190211, 0.01
2020 LC81270512020020LGN00, 0.04 LC81260512020013LGN00, 0.01 LC81260522020013LGN00, 0.03
(2) Dataset for generating training and validation samples
One hundred and twenty-four Gaofen-1 (GF-1) and GF-2 images (30 in 2015, 28 in 2016, 38 in 2017, and 28 in 2018) were acquired from Suzhou Zhongke Tianqi Remote Sensing Technology Co., Ltd. The GF-1 images included panchromatic images with a 2 m spatial resolution and multispectral images with an 8 m spatial resolution. The GF-2 images included panchromatic images with a 1 m spatial resolution and multispectral images with a 4 m spatial resolution. Sixty-two fused GF images covering the entire study area, with a resolution of at least 2 m, were produced from these images through principal component analysis (PCA) (Pal et al., 2007) and were used to generate training and validation samples for land cover classification.
Two thematic maps of land cover classification (JICA, 1999; Eng and Ouch, 2006; Arias, 2013) created by the Japan International Cooperation Agency in 1996 and the Spatial Analysis and Information Laboratory, Ministry of Agriculture, Forestry, and Fisheries, Cambodia, in 2005 were used to generate the training and validation samples, respectively. The former was created based on SPOT images with a spatial resolution of 10 m and Landsat images with a spatial resolution of 30 m. The latter was created based on aerial images at a 1:5,000 scale. Notably, the two maps were unpublished data and sourced from the work of Arias (2013).
Google Earth historical images, with a spatial resolution of 0.61-10 m, were another useful data source for generating training and validation samples for these years without thematic land cover classification maps and fused GF images. These historical images were produced by integrating aerospace and aerial images (such as QuickBird, IKONOS, SPOT and others). The time sequence of the Google Earth historical images from 1984 to the present and the historical images can be achieved through the Google Earth (GE) Platform.
(3) Digital elevation model (DEM)
The DEM used in this study was the Shuttle Radar Topography Mission (SRTM) DEM product (version 3.0). These data have a spatial resolution of 30 m and were obtained from https://gdex.cr.usgs.gov/gdex/. In the follow-up study, the elevation, slope and aspect of the TSLRZ were extracted from the DEM for land cover classification.
(4) Measured water level data
The water level changes of the Kratie gauging station (Figure 1) were used to represent the water level changes of the LMR mainstream. The selection of the Kratie gauging station mainly hinges on its historically statistical correlation within the TSL’s inundation range prior to the development of upstream mega dams (Wang et al., 2020; Gu et al., 2021), a timeframe that establishes the baseline for this study. In addition, it has continuous and complete water level data and is a nearer station to the TSL among all gauging stations in the LMR mainstream. The daily water level data of the Kratie station during the period from 1987 to 2020 were obtained from http://ffw.mrcmekong.org/. In subsequent analyses (trend analysis and correlation analysis), the daily water level was processed into the average water level during the dry season (abbreviated as WLd), the average water level during the wet season (abbreviated as WLw) and the annual average water level (abbreviated as WLa). Notably, before August 31, 2006, water level measurements per day were taken once, while after this date, two measurements per day were recorded, which may have led to the occasional occurrence of distinct fluctuations within a single day. These fluctuations on the daily scale have been ignored in this study, considering that the time scales of hydrological indicators analysed (WLd, WLw and WLa) are longer than the daily scale.

2.3 Methods

2.3.1 Remote sensing-based mapping of land cover

(1) Land cover classification system in the TSLRZ
The land cover classification system of the study area (Table 2) was mainly determined based on the real-world conditions in the TSLRZ, referring to the two abovementioned thematic maps and the study by Mahood et al. (2020). The land cover types include farmland, woodland, wasteland, degraded land, water body and grassland. Notably, farmland includes rice fields and village crops that are under cultivation or not under cultivation, and wasteland is defined as a land cover type that is suitable for agricultural farming, which is sometimes cultivated and sometimes uncultivated. Woodland refers to natural vegetation areas with vegetation coverage greater than or equal to 60%. Degraded land is fragmented woodland and refers to natural vegetation areas with a vegetation coverage of less than 60%. The sample areas are marked with red wireframes (Table 2). The base maps were composed of the 5, 4 and 3 bands of Landsat images collected in 1988.
Table 2 Land cover classification system in the Tonle Sap Lake riparian zone
Land cover types Sample areas Land cover types Sample areas
Water body Woodland
Wasteland Degraded land
Grassland Farmland
(2) Land cover classification based on random forest (RF)
The RF model is an ensemble learning algorithm that has been widely used for land cover classification (Breiman, 2001; Liaw and Wiener, 2002; Yigez et al., 2023). A random forest is a combination of many trees. The number of trees (ntree) in the forest and the number of predictor variables in the random subset at a node of a tree (mtry) are the only two user-defined parameters needed for an RF model (Breiman, 2001; Liaw and Wiener, 2002). In the present study, b1, b2, b3, b4, b5, b6, longitude, latitude, elevation, slope, aspect and vegetation coverage were input as independent variables, which were extracted from the six bands of the SR products of Landsat images, graticules, DEM and other auxiliary data, respectively. The land cover type was used as the dependent variable. The calculation of vegetation coverage was performed as described in previous studies (Li, 2003; Wang et al., 2023). The RF model was used for land cover classification in this study, as explained below.
First, a point dataset with 1,000,000 points covering the study area was generated, and then the sample dataset with 3000 points was constructed through two consecutive random samplings from the point dataset, which was conducted separately in each year. Second, the land cover type of each point in the sample dataset was assigned using visual interpretation based on the reference images, and its corresponding independent variables were linked to the point by the map coordinates. Third, all independent variables from each year were merged by utilizing 30×30 m grids covering the study area and then converted into a matrix as the predicted dataset. Fourth, the RF model was trained and validated over the sample dataset, and the verified model was applied to the predicted dataset. The value of ntree was set as 1000, and mtry was equal to the square root of the number of independent variables. Fifth, the classification results were postprocessed via a majority filter and visual interpretation. Sixth, the accuracy of the classification results was evaluated using producer’s accuracy and user’s accuracy based on a confusion matrix (Foody, 2002).

2.3.2 Measured metrics of land cover change

(1) Trend analysis
The trend line based on the univariate linear regression equation and the least squares method (Wen et al., 2006) was used to analyse the change trends in the land cover type areas. The formula of the linear slope of the trend line (k) is as follows:
$k=\frac{\sum_{i=1}^n x_i \times i-\frac{1}{n}\left(\sum_{i=1}^n x_i\right)\left(\sum_{i=1}^n i\right)}{\sum_{i=1}^n i^2-\frac{1}{n}\left(\sum_{i=1}^n i\right)^2}$
where a positive k indicates an increasing trend and a negative k indicates a decreasing trend. $x_i$ denotes the area of a land cover type in year i. n denotes the total number of years. The significance and the significance level are denoted by p and α, respectively, and α was set to 0.05.
Considering that the operation schedule of the mega dams in the mainstream of the LMR and the hydrological conditions (water level, runoff) of the mainstream showed distinct changes in 2008 (Ji et al., 2018; Yun et al., 2020), we focused on the changes in land cover in two periods: the prechange period (1988-2007) and the postchange period (2009-2020). Importantly, 2008 was not included in the study periods due to the lack of clear Landsat images available to generate a land cover classification map.
(2) Transition matrix
A transition matrix (Takada et al., 2010) is used to reflect the mutual conversion process between various land cover types from the initial stage to the final stage in an area. The matrix is as follows:
$S_{i j}=\left[\begin{array}{cccc}s_{11} & s_{12} & \ldots & s_{1 n} \\ s_{21} & s_{22} & \ldots & s_{2 n} \\ \ldots & \ldots & \ldots & \ldots \\ s_{n 1} & s_{n 2} & \ldots & s_{n n}\end{array}\right]$
where s is the area and n is the number of land cover types. i and j (i, j=1, 2,..., n) are the land cover types at the initial and final stages, respectively. sij is the area of the initial i land type converted to the final j land type. Notably, the transition matrix was calculated based on land cover classification maps in vector format and then presented as a Sankey diagram in this study.
(3) Hot spot analysis
Hot spot analysis (Mitchel, 2010) is a spatial clustering method that can measure the change dynamics of the land cover in a given area and period. High values and low values with statistical significance are called hot spots and cold spots, respectively. The hot spots and cold spots correspond to areas with dramatic land cover changes and areas with minor land cover changes, respectively. In addition, the change frequency map of the land cover was used as the base map for hot spot analysis in this study.

2.3.3 Analysis of mainstream water level change and its relationship with the area of various land cover types

The trend line (see Formula 1 for details) was also used to analyse the trend of the WLd, WLw and WLa. As 80%-90% of the discharge of the LMR occurs from May to October (Delgado et al., 2010), we considered the wet period to be from May to October, and the other months were considered the dry season.
The Spearman correlation coefficient (Spearman, 2010) was used to analyse the relationship between the areas of different land cover types and the mainstream water level. The Spearman correlation coefficient does not require the original variables to have a certain distribution, and its value is between -1 and 1. The greater the absolute value is, the stronger the correlation. Suppose the datasets X(xi) (i=1, 2, 3, …, n) and Y(yi) (i=1, 2, 3, …, n) represent the land cover type area and mainstream water level, respectively. The datasets Q(qi) and S(si) are the level datasets of datasets X and Y, respectively. The formula to calculate the Spearman coefficient (ρ) is as follows:
$\rho=\frac{\sum_{i=1}^n\left(q_i-\bar{q}\right)\left(s_i-\bar{s}\right)}{\sqrt{\left(\sum_{i=1}^n\left(q_i-\bar{q}\right)^2\right)\left(\sum_{i=1}^n\left(s_i-\bar{s}\right)^2\right)}}$
where ρ indicates the degree of correlation. qi and si denote the levels of samples xi and yi in the datasets, respectively. $\bar{q}$ and $\bar{s}$ denote the means of the level datasets Q and S, respectively. n denotes the total number of samples. The significance and the significance level are denoted by p and α, respectively, and α was set to 0.05.
Notably, in the correlation analysis, the areas of various land cover types in the current year corresponded to the average water levels of the mainstream (WLd, WLw and WLa) of the previous year, indicating a time lag response of the lake water level and land cover types to the mainstream water level, as well as the acquisition time of the Landsat images at the beginning of each year.

3 Results

3.1 The accuracy evaluation of the annual scale land cover classification dataset

According to the evaluation results (Figure 2), the producer’s accuracy and user’s accuracy of a single land cover type were greater than 80% for all land cover classification maps during the period from 1988 to 2020. Among these land cover types, the two precision indicators in farmland and water bodies were generally greater than those in other land use types, followed by woodland and grassland, and the indicators in wasteland and degraded land were relatively low.
Figure 2 The accuracy evaluation results of the annual scale land cover classification maps during 1988-2020

3.2 Spatiotemporal changes in land cover

3.2.1 Spatial dynamics of land cover

As shown in Figure 3, the distribution of land cover in the TSLRZ was characterized by an obvious ringed belt structure. Woodland, wasteland and farmland were distributed in the first, second and third belts, respectively, from the inside to the outside around the TSL, and the other land cover types were scattered among the three belts. The spatial pattern of land cover in the TSLRZ during the period from 1988 to 2020 showed the following changes.
Figure 3 Spatial patterns of land cover in the Tonle Sap Lake riparian zone during 1988-2020. The last picture is a schematic diagram of the belt distribution.
First, with the expansion of farmland towards the lakeside, the continuous woodland and grassland areas, especially in the southeastern part, and the wasteland in the second belt were obviously fragmented and gradually disappeared. Second, interactive conversion often occurred between wasteland and farmland. After 2006, the conversion gradually decreased. Third, the conversion of woodland and degraded land was often interactive. Clearly, more woodland was converted to degraded land in 2016 and 2020 than in other years. In particular, most of the degraded land located on both sides of the lake along the northwest‒southeast axis, which was converted from woodland in 2017, was not restored to woodland until 2020. Fourth, the distribution of the water body areas fluctuated greatly, and its area in 1996, 2001, 2009 and 2014 was larger than that in other years. Notably, some water body areas were reclaimed and converted to farmland after 2015, especially around the Tonle Sap River.
Figure 4 shows that in the period from 1988 to 2020, the area of land cover change hot spots in the TSLRZ was 4337.47 km2, accounting for 28.63% of the study area, and most of these hot spots were located in the second belt, the lakeside and the southeastern part. In the prechange period, the area of the hot spots was 3773.93 km2, accounting for 24.91% of the study area, and these hot spots were mostly located in the second belt. In the postchange period, the area of the hot spots was 4367.48 km2, accounting for 28.83% of the study area, and these hot spots were mainly located along the lakeside and in the southeastern part.
Figure 4 Hot spots of land cover changes in the Tonle Sap Lake riparian zone in the whole study period (1988-2020), the prechange period (1988-2007) and postchange period (2009-2020)

3.2.2 Temporal dynamics of land cover

Figure 5 shows that the proportion of farmland area in the study area has always been higher than that of other land use types, with values between 39.30% and 58.17%, followed by that of woodland. The area proportions of these two land cover types were greater than 70% in all years except 1996. In 1988, the proportions of farmland, woodland, wasteland, degraded land, water body and grassland areas in the study area were 41.72%, 32.61%, 13.14%, 5.42%, 4.33% and 2.78%, respectively. By 2020, the proportions of farmland and degraded land areas had increased by 16.46% and 4.40%, respectively, while the proportions of wasteland, woodland, grassland and water body areas decreased by 10.23%, 7.98%, 2% and 0.66%, respectively.
Figure 5 Structure of land cover in the Tonle Sap Lake riparian zone during 1988-2020
Figure 6 shows that during the period from 1988 to 2020, the area of farmland significantly increased, the areas of wasteland, grassland and woodland significantly decreased, and the areas of water bodies and degraded land did not significantly change. Compared with their areas in the prechange period, the increase in farmland and the decrease in grassland in the postchange period became significant, the change in the area of water body transitioned from a significant increase to a significant decrease, and the change in the area of degraded land transitioned from a nonsignificant decrease to a nonsignificant increase.
Figure 6 Change trends of various land cover types in the Tonle Sap Lake riparian zone. k denotes the linear slope. Subscripts 1, 2 and 3 correspond to the prechange period (1988-2007), postchange period (2009-2020) and the whole study period (1988-2020), respectively. p denotes significance, the significance level is set to 0.05, and * indicates p ≤ 0.05.

3.2.3 Conversion of land cover types

As shown in Figure 7a, from 1988 to 2020, the land cover type with the largest added area was farmland, the land cover type with the largest reduction was woodland, and the conversion type with the largest area was the conversion from wasteland to farmland. Figures 7b and 7c show that, first, the land cover type with the largest added area was farmland for both periods. However, the added farmland in the postchange period was approximately 2.02 times that in the prechange period. Second, the land cover type with the largest reduction was woodland, and 47.64% and 41.58% of the reduced part was converted to water bodies and degraded land in the prechange period. The land cover type with the largest reduction was water body, and 55.75% and 26.89% of the reduced part was converted to farmland and woodland in the postchange period. Third, the conversion type with the largest area was from wasteland to farmland in the prechange period, while it was from water body to farmland in the postchange period. Fourth, the area of woodland converted to degraded land was much larger than the area of woodland converted to farmland in each period. In addition, the area of woodland converted to degraded land in the postchange period was 1.78 times that in the previous period.
Figure 7 Sankey diagrams for conversions of the areas of various land cover types in the Tonle Sap Lake riparian zone from 1988 to 2020 (a), from 1988 to 2007 (b), and from 2009 to 2020 (c)

3.3 Changes in mainstream water level and its correlation with the area of land cover type

3.3.1 Changes in mainstream water level

Figure 8a shows that the highest daily water level was 21.94 m in 1996, and the lowest daily water level was 4.24 m in 1993. In addition, the daily water levels in 1998, 2015, 2016, 2019 and 2020 were lower than those in other years. Figures 8b-8d show that WLd and WLa significantly increased in the period from 1987 to 2007, while they did not significantly change in the period from 2008 to 2020. WLw showed no significant change trend in these two periods. Additionally, compared with that in the period from 1987 to 2007, the multiyear average of WLd and WLa in the period from 2008 to 2020 increased by 6.31% and 0.46%, respectively, and the multiyear average of WLw decreased by 2.72%. The changes in the multiyear averages of WLd and WLw indicated that the water level of the mainstream increased in the dry season and decreased in the wet season, and this phenomenon was related to the regulating effect of the upstream mega dams. These results can be supported by previous studies (Lu and Chua, 2021; Chua et al., 2022), although there have been some differences in the hydrological indicators analysed and time series.
Figure 8 Changes in the water level of the Lancang-Mekong River mainstream. k denotes the linear slope. Subscripts 1 and 2 correspond to the two periods of 1987-2007 and 2008-2020, respectively, which indicate the period before and the period after impoundment of the Xiaowan hydropower station, respectively. p denotes significance, the significance level is set to 0.05, and * indicates p ≤ 0.05.

3.3.2 Correlation between land cover type area and mainstream water level

The results of the Spearman correlation analysis (Figure 9 and Table 3) showed that the water body area was significantly related to WLd, WLw and WLa; the area of degraded land was significantly related to WLd and WLa; and the areas of two other land cover types (wasteland and grassland) were significantly related only to WLd. Compared with the prechange period, in the postchange period, the correlations between the areas of the four land cover types (water body, degraded land, wasteland and grassland) and WLd, WLw and WLa were reduced to varying degrees and did not reach significant levels. The results further indicated a weak linkage between the interannual change in local land cover and the mainstream water level and the operation of upstream mega dams.
Figure 9 Scatter plots of the areas of various land cover types in the Tonle Sap Lake riparian zone and the mainstream water levels. WLd, WLw and WLa represent the same meanings as those in Figure 8.
Table 3 Spearman correlation coefficient of the areas of various land cover types in the Tonle Sap Lake riparian zone and the mainstream water levels
Water
levels		 Statistics Study
periods		 Land cover types
Water body Wasteland Grassland Woodland Degraded land Farmland
WLd r 1988-2007 0.75* -0.64* -0.54* -0.15 -0.52* -0.02
2009-2020 0.29 -0.11 -0.18 0.21 -0.38 -0.29
p 1988-2007 0.00* 0.01* 0.02* 0.57 0.03* 0.93
2009-2020 0.44 0.78 0.65 0.59 0.31 0.44
WLw r 1988-2007 0.68* -0.37 -0.29 -0.16 -0.42 -0.09
2009-2020 0.28 -0.18 -0.13 0.30 -0.62 -0.17
p 1988-2007 0.00* 0.14 0.26 0.53 0.09 0.72
2009-2020 0.46 0.64 0.73 0.43 0.08 0.67
WLa r 1988-2007 0.75* -0.44 -0.36 -0.13 -0.52* -0.13
2009-2020 0.28 -0.18 -0.22 0.22 -0.53 -0.17
p 1988-2007 0.00* 0.08 0.16 0.63 0.03* 0.63
2009-2020 0.46 0.64 0.58 0.58 0.14 0.67

Note: WLd, WLw and WLa denote the average water level during the dry season and the wet season and the annual average water level; r denotes the Spearman correlation coefficient; p denotes significance, the significance level is set to 0.05, and * indicates p ≤ 0.05.

4 Discussion

4.1 Notable land cover changes after 2008

Three notable land cover changes were observed in the TSLRZ after 2008 based on the annual scale land cover classification dataset, namely, accelerated expansion of farmland, intensified fragmentation of woodland and significant shrinkage of water bodies (Figures 6 and 7). First, the increasing trend of the farmland area became significant in the postchange period (Figure 6), and the added farmland in the postchange period was approximately 2.02 times that in the prechange period (Figure 7). This result is supported by the study of Mahood et al. (2020), who observed that the added area of farmland was 441 km2 in the A-TSLRZ from 1993 to 2008, while it reached 2356 km2 from 2008 to 2018. Second, the area of woodland converted to degraded land in the postchange period was 1.78 times that in the previous period (Figure 7). This noticeable change was in sound agreement with the results of the study by Chen et al. (2022), who reported that forests have been rapidly fragmented in the A-TSLRZ since 1992 and that hardly any intact forests remained after 2010 due to logging and agricultural farming. Third, the change in the area of the water body significantly decreased in the postchange period (Figure 6). This is consistent with the results of several previous studies (Ji et al., 2018; Wang et al., 2020; Ng and Park, 2021). For example, Ji et al. (2018) found that the multiyear average area of the LWB in the TSL decreased by 1.5% in the period from 2008 to 2014 compared with the period from 2000 to 2007. Wang et al. (2020) also reported that the monthly average area of the LWB in the TSL changed only slightly in the period from 1988 to 2000 but showed a significant decreasing trend in the period from 2000 to 2018. Notably, compared with the results of these studies, the water body change detected in this study was more significant because our study did not include the smallest lake body, which is regarded as fixed.
Furthermore, some special phenomena have been documented based on the annual scale dataset. First, the interactive conversion of farmland and wasteland occurred often before 2006, and the areas of occurrence were unstable (Figures 3 and 7). This result may have been related to the local industrial structure. When local tourism and fishing are booming, labour tends to shift from agriculture to these industries, and vice versa. After 2006, this phenomenon was rare, which further reflected the acceleration of farmland expansion. Second, the conversion of woodland and degraded land was interactive, and the conversion was more likely to occur in the drier period. For example, 2015 and 2019 were typical drought years in the TSL area (Frappart et al., 2018; Sok et al., 2021). Correspondingly, more woodland was converted to degraded land in 2016 and 2020 than in other years (Figure 3). In addition, in contrast to the above interactive conversion caused by drought, most of the degraded land located on both sides of the lake, which was converted from woodland in 2017, was not restored to woodland until 2020. This difference can be ascribed to fires. According to the report by Lovgren (2020), massive fires burned as much as one-third of the 750,000 acres of flooded forest in the TSL area in 2016.

4.2 Response of land cover change to mainstream water level

The correlations between the water body area in the TSLRZ and the mainstream water levels (WLd, WLw and WLa) were statistically significant during the prechange period (1988-2007) but insignificant during the postchange period (2009-2020) (Figure 9 and Table 3). This result indicated that the water body shrinkage in the last 10 years (Figure 6) was not dominated by the mainstream water level and was still affected by other factors. This result is supported by the study of Ji et al. (2018), who reported that the decrease in the water area of the TSL during the dry season after 2008 could not be attributed solely to the change in mainstream runoff. The changes in the correlations between the water body area and the mainstream water levels during the postchange period can be attributed to the following three aspects. First, according to previous studies (Morovati et al., 2021; 2023; Chua et al., 2022), the flood pulse of the LMR has declined in recent years due to the development and operation of upstream mega dams in the upstream, climate variations, local irrigation and channel incision (caused by declining sediments combined with sand-mining operations). This may be an important reason for the shrinkage of water body areas. Second, the components of the water balance model for this lake-floodplain system mainly include the LMR mainstream, TSL’s own tributaries, rainfall, evaporation, groundwater infiltration and overland flow via floodplain (Kummu et al., 2014; Morovati et al., 2023), which suggested that the change in other components (excluding the mainstream) can also lead to the change in the water body areas in the TSLRZ. Daly et al. (2020) noted that the average temperature in the TSL increased by 0.03°C/year in the period from 1988 to 2018, which meant that evaporation increased in the area. The study of Ng and Park (2021) showed that tributary discharge and rainfall, which provide approximately 50% of the supply of lake water (the proportion of the supply may vary in different periods), did not display any significant trend over the investigated period (1980 to 2018) and were not the primary factors contributing to the shrinking of the lake. Third, based on the results of our study, the water body areas have been subjected to greater human interference during the postchange period. Under natural conditions, water body areas will be transformed into natural land cover types, such as woodland and grassland. However, approximately 55.75% of the reduced water body areas in 2009 were converted into farmland in 2020 (Figure 7), especially around the Tonle Sap River (Figure 3). This reflects that local agricultural farming interfered with the original response mode between the water body area of the TSLRA and the mainstream water level. Moreover, in the postchange period, the area of added farmland was approximately 2.02 times that in the previous period (Figure 7), which meant that the irrigation water consumption increased accordingly and would also interfere with the original response mode. Based on the above results, it can be concluded that the changes in water body areas in the TSLRA were dominated by mainstream water levels during the prechange period but were influenced by multiple factors together during the postchange period.
The other land cover types (farmland, woodland, wasteland, degraded land, and grassland) in the TSLRZ indirectly responded to the mainstream water level through the conduction of the lake water level. In the prechange period, the area of degraded land was significantly related to WLd and WLa, and the areas of two other land cover types (wasteland and grassland) were significantly related only to WLd (Figure 9 and Table 3), in which the correlation between the area of wasteland and WLd was strong. However, as shown in Figure 7b, most changed wasteland was converted into farmland; therefore, the causality of this statistically significant correlation was weak. In addition, the correlations between the areas of two land cover types (degraded land and grassland) and mainstream water levels were weak during the prechange period but reduced to varying degrees and did not reach significant levels during the postchange period. These fluctuations in both the areas of these land cover types and the mainstream water levels may relate to the more intense human activities, as well as more frequent and higher extreme changes in the climate in the postchange period.
According to the above analysis, the land cover changes in the last 30 years were mainly the conversion of nonagricultural land to agricultural land by local reclamation and logging, which stressed the importance of agriculture and the imperative of food security in Cambodia. The main driving force of land cover changes was the ever-growing population in Cambodia. The population of Cambodia changed from 7,975,597 in 1987 to 16,767,842 in 2022 (data sourced from the World Bank public database), and the demand for food continues to increase accordingly. Another driving force may be international market demand. The policy document on promoting rice production and rice export issued by Cambodia (Royal Government of Cambodia, 2010) highlighted how to turn Cambodia into a “rice basket” and a major global rice exporter and proposed an export target of 1 million tons in 2015. Moreover, Cambodia’s rice exports have accelerated since the launch of the China-ASEAN Free Trade Area in 2010 (Prom, 2016). In addition, although the reduction in the flood pulse of the LMR was not a driving force, it indirectly provided convenient conditions for agriculture activities to a certain extent. Usually, human activities such as reclamation and logging are constrained by higher flood levels and wider inundation ranges. After the flood pulse weakened, the range of water level changes became relatively narrow. This change was conducive to the development and utilization of land in the TSLRZ, thereby increasing the area of farmland.

4.3 Limitations and way forward

The relationship between the water level change of the mainstream and the land cover change in the TSLRZ, as identified in this study, represents a linkage relationship rather than a quantitative direct relationship. This is due to the complexity of the indirect effect of the mainstream water level on the land cover and the lack of relevant hydrological factor data in the TSLRZ. A quantitative relationship between them needs further investigation through sequential analysis of the two quantitative relationships. The first is between the land cover change in the TSLRZ and the hydrological factor change in the TSLRZ, while the second is between the hydrological factor change and the water level change of the mainstream. The hydrological factors, which include but are not limited to inundation frequency, inundation period, and humidity, are located in the same two-dimensional space as land cover and directly affect its dynamics.
The interaction mechanism between LMR and TSL is indeed complex due to the influence of the annual reverse flow phenomenon. The reverse flow occurs when the water level of the LMR is higher than that of the TSL, which will enlarge the TSL inundation range and impact the land cover changes in the TSLRZ. Currently, scholars have deeply studied the interaction mechanism (Kummu et al., 2014; Li et al., 2019; Morovati et al., 2021; 2023; Chua et al., 2022). In contrast to previous studies, the focus of this study is to explore the changes in the ecological environment of the TSL when the hydrological regime of the LMR mainstream underwent distinct changes and its relationship with the water level of the LMR. In this analysis, the land cover changes in the TSL were used to represent its ecological changes, while the water level changes at the Kratie gauging station were used to represent the water level changes in the LMR mainstream. The selection of the Kratie gauging station mainly hinges on its historically statistical correlation within the TSL’s inundation range prior to the development of upstream mega dams, a timeframe that establishes the baseline for this study. Notably, the water level fluctuations at the gauging station do not necessarily indicate water body area changes in the TSLRZ, and the flood pulse at this station does not necessarily trigger the onset of reverse flow. Therefore, the linkage relationship obtained in this study can only reflect the localized aspects of the interaction mechanism between LMR and TSL. Future research to comprehensively decipher this mechanism necessitates a multifaceted approach and integrating additional hydrological indicators.
While we do not advocate ignoring the future implications of the mainstream hydrological regime, the land cover changes that we have documented and their correlations with the mainstream water level showed that the influences of local reclamation and logging on the land cover changes were far greater than those of the mainstream water level, greatly threatening the ecological environment in the TSLRZ. The continuous expansion of farmland may be unsustainable in the long run, endangering regional food security. Therefore, it is of great importance for government agencies, relevant investors and development agencies to take urgent, coordinated action to address the issue of the conversion of nonagricultural land to agricultural land. By doing so, it is conducive to mitigating possible risk in the future of the TSL and the livelihoods of those residing in the region.
We propose three measures to address the issue identified here. First, areas that are already zoned for protection, such as the three Biosphere Reserve core areas and one Ramsar site, must be prohibited for reclamation and logging and strictly protected so that they can provide ecosystem services to the surrounding communities. Second, secondary protected areas should be built in some areas that have already experienced extensive reclamation and logging, such as the areas located on both sides of the lake along the northwest‒southeast axis and areas around the Tonle Sap River, and implement ecological restoration actions such as reverting reclaimed land to its natural state, water source protection and waterway dredging. Third, local government agencies should establish a data sharing platform. Specifically, this platform should include different types of data, such as hydrological data, meteorological data, precious species data and grain production and export data. Through data sharing, more scholars, research institutions and environmental protectors can participate in the ecological protection of the TSL.

5 Conclusions

In this study, we created an annual scale land cover classification dataset with a 30 m spatial resolution in the TSLRZ during the period from 1988 to 2020, and the producer’s accuracy and user’s accuracy of a single land cover type were greater than 80%. On this foundation, this study clarified the interannual spatiotemporal changes in land cover in the TSLRZ and assessed their relationship with mainstream water level changes in the LMR. The main conclusions of the study were as follows:
(1) Farmland was the largest land cover type, and its expansion towards the lakeside was the most notable change in the TSLRZ in the last 30 years. During this period, the farmland area showed a significant increase, the wasteland, grassland and woodland areas all showed significant decreases, and the water body and degraded land areas showed stage-specific characteristics.
(2) There were three notable land cover changes in the TSLRZ after 2008, namely, accelerated farmland expansion, intensified woodland fragmentation and significant water body shrinkage. During the postchange period (2009-2020), the water body area significantly decreased, and the areas of added farmland and fragmented woodland were 2.02 and 1.78 times those in the prechange period (1988-2007), respectively.
(3) The responses of the water body, degraded land, wasteland and grassland areas to the mainstream water levels weakened in the TSLRZ after 2008. Their correlations were reduced and did not reach significant levels, indicating a weak linkage between the interannual change in local land cover and the mainstream water level and the operation of upstream mega dams.
(4) The land cover changes in the TSLRZ mainly resulted from local reclamation and logging and were driven by the ever-growing population in Cambodia and the international market demand.
The results above can provide a new scientific basis for the transboundary influence analysis of hydrological change, and they can be used to support resource utilization and ecological protection in the riparian zone of river-connected lakes under climate change and cascade dam building.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Arias M E, 2013. Impacts of hydrological alterations in the Mekong Basin to the Tonle Sap ecosystem[D]. Christchurch, New Zealand: University of Canterbury.

[2]
Asian Development Bank (ADB), 2019. Kingdom of Cambodia: Preparing the irrigated agriculture improvement project. https://www.adb.org/sites/default/files/project-documents/51159/51159-001-tacr-en.pdf.

[3]
Binh D V, Kantoush S A, Saber M et al., 2020. Long-term alterations of flow regimes of the Mekong River and adaptation strategies for the Vietnamese Mekong Delta. Journal of Hydrology: Regional Studies, 32: 100742.

DOI

[4]
Breiman L, 2001. Random Forest. Machine Learning, 45: 5-32.

DOI

[5]
Chen A F, Chen A P, Varis O et al., 2022. Large net forest loss in Cambodia’s Tonle Sap Lake protected areas during 1992-2019. Ambio, 51(8): 1889-1903.

DOI

[6]
Chua S D X, Lu X X, Oeurng C et al., 2022. Drastic decline of flood pulse in the Cambodian floodplains (Mekong River and Tonle Sap system). Hydrology and Earth System Sciences, 26(3): 609-625.

DOI

[7]
Daly K, Ahmad S K, Bonnema M et al., 2020. Recent warming of Tonle Sap Lake, Cambodia: Implications for one of the world’s most productive inland fisheries. Lakes and Reservoirs: Research and Management, 25(2): 133-142.

DOI

[8]
Dang H, Pokhrel Y, Shin S et al., 2022. Hydrologic balance and inundation dynamics of Southeast Asia’s largest inland lake altered by hydropower dams in the Mekong River basin. Science of the Total Environment, 831: 154833.

DOI

[9]
Delgado J M, Apel H, Merz B, 2010. Flood trends and variability in the Mekong River. Hydrology and Earth System Sciences, 14(3): 407-418.

DOI

[10]
Eng C, Ouch V, 2006. TSBR Land cover map derived from the orthophoto map (unpublished data from the Spatial Analysis and Information Laboratory, Ministry of Agriculture, Forestry, and Fisheries,

[11]
Foody G M, 2002. Status of land cover classification accuracy assessment. Remote Sensing of Environment, 80(1): 185-201.

DOI

[12]
Frappart F, Biancamaria S, Normandin C et al., 2018. Influence of recent climatic events on the surface water storage of the Tonle Sap Lake. Science of The Total Environment, 636: 1520-1533.

DOI

[13]
Gu Z K, Zhang Y, Fan H, 2021. Mapping inter- and intra-annual dynamics in water surface area of the Tonle Sap Lake with Landsat time-series and water level data. Journal of Hydrology, 601: 126644.

DOI

[14]
Guan Y, Zheng F, 2021. Alterations in the water-level regime of Tonle Sap Lake. Journal of Hydrologic Engineering, 26(1): 05020045.

DOI

[15]
He C, Zhang J, Liu Z, et al., 2022. Characteristics and progress of land use/cover change research during 1990-2018. Journal of Geographical Sciences, 32(3): 537-559.

DOI

[16]
He D M, Liu H, Feng Y et al., 2016. Perspective on theories and methods study of transboundary water resources under the global change. Advances in Water Science, 27(6): 928-934. (in Chinese)

[17]
Huang Z Y, Wang J D, Zhang Y B, 2020. A brief analysis of the relationship between landscape change and human activities in Tonle Sap Lake Basin, Cambodia. China Rural Water and Hydropower, (4): 30-38. (in Chinese)

[18]
Ji X, Li Y, Luo X et al., 2018. Changes in the lake area of Tonle Sap: Possible linkage to runoff alterations in the Lancang River? Remote Sensing, 10(6): 866.

DOI

[19]
JICA, 1999. Cambodia reconnaissance survey digital data (unpublished data from the Japan International Cooperation Agency).

[20]
Kim S, Sohn H G, Kim M K et al., 2019. Analysis of the relationship among flood severity, precipitation, and deforestation in the Tonle Sap Lake Area, Cambodia using multi-sensor approach. KSCE Journal of Civil Engineering, 23: 1330-1340.

DOI

[21]
Kummu M, Tes S, Yin S et al., 2014. Water balance analysis for the Tonle Sap Lake-floodplain system. Hydrological Processes, 28(4): 1722-1733.

DOI

[22]
Lamberts D, 2008. The unintended role of the local private sector in biodiversity conservation in the Tonle Sap Biosphere Reserve, Cambodia. Local Environment, 13(1): 43-54.

DOI

[23]
Lambin E F, Turner B L, Geist H J et al., 2001. The causes of land-use and land-cover change: Moving beyond the myths. Global Environmental Change, 11(4): 261-269.

DOI

[24]
Li C W, You Z Q, Li A Q et al., 2019. Variation characteristics, influencing factors and hydrological conditions of the reverse flow from Mekong River to Tonle Sap Lake. Applied Ecology and Environmental Research, 17(6): 13875-13895.

[25]
Li D, Long D, Zhao J et al., 2017. Observed changes in flow regimes in the Mekong River basin. Journal of Hydrology, 551: 217-232.

DOI

[26]
Li M M, 2003. The method of vegetation fraction estimation by remote sensing[D]. Beijing, China: Institute of Remote Sensing Applications, Chinese Academy of Sciences. (in Chinese)

[27]
Liaw A, Wiener M, 2002. Classification and regression by random forest. R News, 2(3): 18-22.

[28]
Lin Z, Qi J, 2017. Hydro-dam-A nature-based solution or an ecological problem: The fate of the Tonle Sap Lake. Environmental Research, 158: 24-32.

DOI

[29]
Lovgren S, 2020. Cambodia’s biggest lake is running dry, taking forests and fish with it. https://www.nationalgeographic.com/science/article/cambodia-tonle-sap-lake-running-dry-taking-flooded-forest-fish.

[30]
Lu X X, Chua S, 2021. River discharge and water level changes in the Mekong River: Droughts in an era of mega-dams. Hydrological Processes, 35(7): e14265.

DOI

[31]
Mahood S P, Poole C M, Watson J E M et al., 2020. Agricultural intensification is causing rapid habitat change in the Tonle Sap floodplain, Cambodia. Wetlands Ecology and Management, 28(5): 713-726.

DOI

[32]
Meng Q J, Zang S Y, Song K S et al., 2018. A comparison of conditions of Tonle Sap Lake wetlands in dry and wet seasons in 1990 with those in 2016. Wetland Science, 16(6): 801-807. (in Chinese)

[33]
Mitchel A, 2010. The ESRI Guide to GIS Analysis (Volume 2): Spatial Measurements and Statistics. RedLands: ESRI Press.

[34]
Morovati K, Nakhaei P, Tian F et al., 2021. A machine learning framework to predict reverse flow and water level: A case study of Tonle Sap Lake. Journal of Hydrology, 603: 127168.

DOI

[35]
Morovati K, Tian F, Kummu M et al., 2023. Contributions from climate variation and human activities to flow regime change of Tonle Sap Lake from 2001 to 2020. Journal of Hydrology, 616: 128800.

DOI

[36]
Ng W X, Park E, 2021. Shrinking Tonle Sap and the recent intensification of sand mining in the Cambodian Mekong River. Science of the Total Environment, 777: 146180.

DOI

[37]
Pal S K, Majumdar T J, Bhattacharya A K, 2007. ERS-2 SAR and IRS-1C LISS III data fusion: A PCA approach to improve remote sensing based geological interpretation. ISPRS Journal of Photogrammetry and Remote Sensing, 61(5): 281-297.

DOI

[38]
Pokhrel Y, Burbano M, Roush J et al., 2018. A review of the integrated effects of changing climate, land use, and dams on Mekong River hydrology. Water, 10(3): 266.

DOI

[39]
Prom S, 2016. Research on rice export trade of Cambodia[D]. Guilin, China: Guangxi Normal University. (in Chinese)

[40]
RAMSAR, 2010. The list of wetlands of international importance. http://www.ramsar.org/.

[41]
Royal Government of Cambodia, 2010. Policy paper on promotion of paddy rice production and export of milled rice. Supreme National Economic Council, Phnom Penh.

[42]
Siev S, Yang H, Sok T et al., 2018. Sediment dynamics in a large shallow lake characterized by seasonal flood pulse in Southeast Asia. The Science of the Total Environment, 631: 597-607.

[43]
Sok S, Chhinh N, Hor S et al., 2021. Climate change impacts on rice cultivation: A comparative study of the Tonle Sap and Mekong River. Sustainability, 13(16): 8979.

DOI

[44]
Spearman C, 2010. The proof and measurement of association between two things. International Journal of Epidemiology, 39(5): 1137-1150.

DOI PMID

[45]
Sterling S M, Ducharne A, Polcher J, 2013. The impact of global land-cover changes on the terrestrial water cycle. Nature Climate Change, 3(4): 385-390.

DOI

[46]
Takada T, Miyamoto A, Hasegawa S F, 2010. Derivation of a yearly transition probability matrix for land-use dynamics and its applications. Landscape Ecology, 25: 561-572.

DOI

[47]
Tangdamrongsub N, Ditmar P G, Steele-Dunne S C et al., 2016. Assessing total water storage and identifying flood events over Tonle Sap basin in Cambodia using GRACE and MODIS satellite observations combined with hydrological models. Remote Sensing of Environment, 181: 162-173.

DOI

[48]
Trung N V, Choi J H, Won J S, 2013. A land cover variation model of water level for the floodplain of Tonle Sap, Cambodia, derived from ALOS PALSAR and MODIS data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(5): 2238-2253.

DOI

[49]
Turner B L, Lambin E F, Reenberg A, 2007. The emergence of land change science for global environmental change and sustainability. Proceedings of the National Academy of Sciences of the United States of America, 104(52): 20666-20671.

DOI PMID

[50]
UNESCO, 2006. Biosphere Reserves: World Network. http://www.unesco.org/mab/.

[51]
Wang D, Liu Y, Zheng L, et al., 2023. Growing impacts of low-flow events on vegetation dynamics in hydrologically connected wetlands downstream Yangtze River Basin after the operation of the Three Gorges Dam. Journal of Geographical Sciences, 33(4): 885-904.

DOI

[52]
Wang H W, Qi Y, Lian X H et al., 2022. Effects of climate change and land use/cover change on the volume of the Qinghai Lake in China. Journal of Arid Land, 14(3): 245-261.

DOI

[53]
Wang T, Liu C L, Du D B, 2021. Spatio-temporal dynamics of international freshwater conflict events and relations from 1948 to 2018. Acta Geographica Sinica, 76(7): 1792-1809. (in Chinese)

DOI

[54]
Wang Y, Feng L, Liu J G et al., 2020. Changes of inundation area and water turbidity of Tonle Sap Lake: Responses to climate changes or upstream dam construction? Environmental Research Letters, 15(9): 0940a1.

DOI

[55]
Wen G, Wang Z, Xia S et al., 2006. Least-squares fitting of multiple M-dimensional point sets. Visual Computer, 22(6): 387-398.

DOI

[56]
Yigez B, Xiong D, Zhang B et al., 2023. Dynamics of soil loss and sediment export as affected by land use/cover change in Koshi River Basin, Nepal. Journal of Geographical Sciences, 33(6): 1287-1312.

DOI

[57]
Yun X, Tang Q, Wang J et al., 2020. Impacts of climate change and reservoir operation on streamflow and flood characteristics in the Lancang-Mekong River Basin. Journal of Hydrology, 590: 125472.

DOI

[58]
Zhan X, Sohlberg R A, Townshend J R G et al., 2002. Detection of land cover changes using MODIS 250 m data. Remote Sensing of Environment, 83(1/2): 336-350.

DOI

[59]
Zhao J C, Yang K, Zhu Y H et al., 2019. The spatial-temporal changes of wetland in the Tonle Sap Lake Basin from 1988 to 2009. Journal of Southwest Forestry University, 39(6): 130-136. (in Chinese)

[60]
Zhong R D, Zhao T T G, Chen X H, 2021. Evaluating the tradeoff between hydropower benefit and ecological interest under climate change: How will the water-energy-ecosystem nexus evolve in the upper Mekong basin? Energy, 237: 121518.

DOI

[61]
Ziv G, Baran E, Nam S et al., 2012. Trading-off fish biodiversity, food security, and hydropower in the Mekong River Basin. Proceedings of the National Academy of Sciences of the United States of America, 109(15): 5609-5614.

DOI PMID

Options
Outlines

/

〈 〉