Spatial extrapolation in soil salinity and future land cover using hybrid machine learning and land change modeler: case study in the Mekong Delta and the Red River Delta

Spatial modeling can be used to predict future land cover changes based on past and present conditions. However, it is not yet known to what extent this model can be used to predict the future with reliable accuracy. Therefore, by using multi-temporal land cover data, this study aims to build an optimal model based on the calibration interval scenario. The optimal model is then used to predict and analyze changes in land cover in East Kalimantan in 2016–2036. 11 classified multi-temporal land cover maps from the Landsat Time Series using Random Forest in Google Earth Engine are used to model 14 calibration interval scenarios. A land Change Modeler is used to model and predict land cover change with 14 driving variables. The results of the classification of multi-temporal land cover maps show a good level of accuracy, with an Overall Accuracy value of 71.43–85.14% and a Kappa value of 0.667–0.827. Then 2016–2021 is one of the best scenarios with 5-year intervals where the accuracy of future predictions can still be relied upon for up to three prediction iterations. The calibration interval scenario approach in spatial modeling in East Kalimantan can be relied upon to show a decrease in forest cover from 2016 to 2021, with a deforestation rate of 651 km2/year. The prediction of land cover in 2036 estimates that the remaining forest cover area in East Kalimantan is 69.203 km2. It is believed that topography is the most influential variable driving land cover change in East Kalimantan.

Land change modeling (LCM) is a complex GIS procedure aiming at predicting land cover changes in the future, contributing thus to the design of interventions that help maintain ecosystem services and mitigate climate change impacts. In the present work, the land change model for Greece, a typical Mediterranean country, has been developed, based on historical information from remotely sensed land cover data. Land cover types based on the International Geosphere-Biosphere Program (IGBP) classification were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) land cover product, i.e. MCD12Q1, provided annually from 2001 to 2018 at a spatial resolution of 500 m. Initially, the dominant land cover changes and their driving variables for the decade of 2001 to 2011 were determined and the transition potential of land was mapped using a multi-layer perceptron (MLP) neural network. Four dominant land-cover transformations were found in Greece from 2001 to 2011, i.e. land transformation from Savannas to Woody Savannas, from Savannas to Grasslands, from Grasslands to Savannas, and from Croplands to Grasslands. Driving variables were found to be the Evidence Likelihood of Land Cover, i.e. the relative frequency with which different land cover categories occurred within the areas that transitioned, the Altitude as realized in the Digital Elevation Model of Greece from ASTER GDEM, the Distance from previously changed land and two climate variables i.e. Mean Annual Precipitation and Mean Annual Minimum Temperature. After the model was calibrated, its predictive ability was tested for land cover prediction for 2018 and was found to be 96.7%. Future land cover projections up to 2030 were developed incorporating CMIP6 climate data under two Shared Socioeconomic Pathways (SSPs), i.e SSP126 corresponding to a sustainable future and SSP585, which describes the future world based on fossil-fueled development. The results indicate that major historical land transformations in Greece, do not correspond to land degradation or desertification, as it has been reported in previous works. On the contrary, the land cover transitions indicate that the Woody Savannas gain areas constantly, whereas Grasslands and Croplands lose areas, and forested areas of all types demonstrate moderate gains. Concerning future land cover, the present work indicates that the direction of historical changes will also prevail in the next decade, with the most severe scenario, i.e. SSP585 slowing down the rate of changes and the most sustainable one, i.e. SSP126, accelerating the rate of expansion of woody vegetation land cover type.

Land cover has been changing rapidly throughout the world, and this issue is important to researchers, urban planners, and ecologists for sustainable land cover planning for the future. Many modeling tools have been developed to explore and evaluate possible land cover scenarios in future and time scales vary greatly from one study to another. The main objective of this study is to test land cover change prediction at different time scales in a Mediterranean catchment in SE France. Land cover maps were created from aerial photographs (1950, 1982, 2003, 2008, and 2011) of the Giscle catchment (235 Km2) and surfaces were classified into four land cover categories: forest, vineyard, grassland, and built area. Explanatory variables were selected through Cramer’s coefficient. Different time scales were tested in the study: short (2003-2008), intermediate (1982-2003), and long (1950-1982). To test the model’s accuracy, Land Change Modeler (LCM) of IDRISI was used to predict land cover in 2011 and predicted images were compared to a real 2011 map. Kappa index and confusion matrix were used to evaluate the model’s accuracy. Altitude, slope, and distance from roads had the greatest impact on land cover changes among all variables tested. Good to perfect level of spatial and perfect level of quantitative agreement were observed in long to short time scale simulations. Kappa indices (Kquantity = 0.99and Klocation = 0.90) and confusion matrices were good for intermediate and best for short time scale. The results indicate that shorter time scales produce better predictions. Time scale effects have strong interactions with specific land cover dynamics, in which stable land covers are easier to predict than cases of rapid change and quantity is easier to predict than location for longer time periods.

Soil salinization will affect 50% of global cropland areas by 2050 and represents a major threat to agricultural production and food sovereignty. As soil salinity monitoring is costly and time consuming, many regions of the world undertake very limited soil salinity observation (in space and time), preventing the accurate assessment of soil salinity hazards. In this context, this study assesses the relative performance of Sentinel-1 radar and Sentinel-2 optical images, and the combination of the two, for monitoring changes in soil salinity at high spatial and temporal resolution, which is essential to evaluate the mitigation measures required for the sustainable adaptation of agriculture practices. For this purpose, an improved learning database made of 863 soil electrical conductivity (i.e., soil salinity) observations is considered for the training/validation step of a Random Forest (RF) model. The RF model is successively trained with (1) only Sentinel-1, (2) only Sentinel-2 and (3) both Sentinel-1 and -2 features using the Genetic Algorithm (GA) to reduce multi-collinearity in the independent variables. Using k-fold cross validation (3-fold), overall accuracy (OA) values of 0.83, 0.88 and 0.95 are obtained when considering only Sentinel-2, only Sentinel-1 and both Sentinel-1 and -2 features as independent variables. Therefore, these results highlight the clear complementarity of radar (i.e., Sentinel-1) and optical (i.e., Sentinel-2) images to improve soil salinity mapping, with OA increases of approximately 10% and 7% when compared to Sentinel-2 and Sentinel-1 alone. Finally, pre-sowing soil salinity maps over a five-year period (2019–2023) are presented to highlight the benefit of the proposed procedure to support the sustainable management of agricultural lands in the context of soil salinization on a regional scale.

The incredible drying of the Aral Sea has resulted in a large area of exposed seafloor with saline soils, which has led to catastrophic consequences. This study investigated ground-truth soil salinity data and used Landsat data to map the soil salinity distribution of the exposed seabed of the Aral Sea from 1960 onwards. The soil salinity distribution, with the depth from 0 cm to 100 cm, was analyzed. The correlation analysis was applied to find the best performance index in describing soil salinity changes. The results showed that ground-truth data of topsoil salinity (depth of 0−5 cm) exhibited a significantly strong correlation with soil salinity index 4 (SI4) among seven indices, where the Pearson correlation coefficient (r) was up to 0.92. Based on the relationship between soil salinity sampling data and SI4, a linear regression model was employed to determine the capability of evaluating the soil salinity distribution of the Aral Sea with the coefficient of determination (R2), root mean squared error (RMSE), and ratio of performance to deviation (RPD) values of 0.84 and 0.86 dS m−1 and 2.36, respectively. The SI4 performed well and was used to predict the soil salinity distribution on the exposed seabed. The distribution showed that soil salinity increased from the former to current shoreline. In the North Aral Sea, compared to 1986, the water area remained stable, accounting for 50.3% in 2020, and the soil salinization level was relatively low. However, the moderately and slightly saline areas dominated 73.8% and 7.5% of the South Aral Sea in 2020, with an increase of 53% and 6% transformed from the water area. The area of salinized soils dramatically increased. The strongly and extremely saline areas were mainly located in the northeastern part of the eastern basin and western part of Vozrozhdeniya Island, respectively, and were the main source of salt-dust storms. These results support the dynamic monitoring and distribution patterns of soil salinization in the Aral Sea.

Groundwater contamination by arsenic is a serious environmental problem in the world. Yet there have been few studies conducted in Southeast Asian countries. This article surveys arsenic contamination in groundwater and residents from Vietnam, and is based on our previous studies. Samples of groundwater (n = 118), human hair (n = 59), and urine (n = 100) were collected in the Red River and Mekong River Deltas during 2001–2004. Arsenic was detected in most of the groundwater samples, and its level ranged from <0.1 to 486 µg/l. Elevated concentrations of arsenic were observed in groundwater at some locations in Ha Nam (up to 486 µg/l) in the Red River Delta and Dong Thap (up to 411 µg/l) in the Mekong River Delta. Remarkably, about 33% of these groundwater samples exceeded the WHO drinking water guideline of 10 µg/l. These results suggest that arsenic contamination in groundwater may be widely present in both the Red River and Mekong River Deltas, Vietnam. A significant positive correlation was observed between arsenic concentrations in groundwater and human hair. Arsenic speciation of human urine revealed the presence of inorganic arsenic, and these concentrations positively correlated with arsenic levels in groundwater. Thus, it is likely that residents in our study areas are chronically exposed to arsenic through drinking groundwater, suggesting that there is a health risk from arsenic in Vietnam.

Maps of soil salinity and toxic alkalinity in 2017–2018 were developed for the Chervlenoe irrigation massive of Svetloyarsk irrigation system. These maps were compared with the maps of soils in the virgin state and under irrigation in 1992 and 2006. The massive is located close to the Volga–Don Canal on the northern slopes of the Ergeni Upland (Volgograd oblast). Before irrigation, the soil cover was composed of associations (complexes) of loamy solonetzic light chestnut soils (Sodic Cambisols (Loamic, Ochric, Magnezic, Bathyprotosalilic)) and steppe solonetzes (Endosalic Solonetz (Albic, Loamic, Columnic, Cutanic, Differentic, Ochric, Magnezic)) developed from loesslike loam. The soil cover was strongly transformed by surface leveling during the irrigation system construction, and underwent secondary salinization in the 1970s–1990s because of excessive irrigation water supply. Changes in the soil salinity in the past 20–25 years are discussed. Both favorable (partial salt leaching) and negative (alkalization of topsoil horizons) processes are revealed.

In arid and semi-arid regions, soil moisture and salinity are important elements to control regional ecology and climate, vegetation growth and land function. Soil moisture and salt content are more important in arid wetlands. The Ebinur Lake wetland is an important part of the ecological barrier of Junggar Basin in Xinjiang, China. The Ebinur Lake Basin is a representative area of the arid climate and ecological degradation in central Asia. It is of great significance to study the spatial distribution of soil moisture and salinity and its causes for land and wetland ecological restoration in the Ebinur Lake Basin. Based on the field measurement and Landsat 8 satellite data, a variety of remote sensing indexes related to soil moisture and salinity were tested and compared, and the prediction models of soil moisture and salinity were established, and the accuracy of the models was assessed. Among them, the salinity indexes D1 and D2 were the latest ones that we proposed according to the research area and data. The distribution maps of soil moisture and salinity in the Ebinur Lake Basin were retrieved from remote sensing data, and the correlation analysis between soil moisture and salinity was performed. Among several soil moisture and salinity prediction indexes, the normalized moisture index NDWI had the highest correlation with soil moisture, and the salinity index D2 had the highest correlation with soil salinity, reaching 0.600 and 0.637, respectively. The accuracy of the BP neural network model for estimating soil salinity was higher than the one of other models; R2 = 0.624, RMSE = 0.083 S/m. The effect of the cubic function prediction model for estimating soil moisture was also higher than that of the BP neural network, support vector machine and other models; R2 = 0.538, RMSE = 0.230. The regularity of soil moisture and salinity changes seemed to be consistent, the correlation degree was 0.817, and the synchronous change degree was higher. The soil salinity in the Ebinur Lake Basin was generally low in the surrounding area, high in the middle area, high in the lake area and low in the vegetation coverage area. The soil moisture in the Ebinur Lake Basin slightly decreased outward with the Ebinur Lake as the center and was higher in the west and lower in the east. However, the spatial distribution of soil moisture had a higher mutation rate and stronger heterogeneity than that of soil salinity.

Soil salinity is one of the most brutal environmental factors reducing the surface area and productivity of salt-sensitive cultivated plants. This major problem hurts land cover and agricultural productivity, principal to a decline in soil fertility and quality. The objective of this paper is to study the possibility of using an image of the spectral remote sensing data and field survey to sample soil and map salinity by correlating electrical conductivity (EC) field measurements with soil salinity indices derived from the Landsat7 satellite image. In Habra plain (North-western Algeria), the field’s electrical conductivity was measured during the period from 15 to 19 November 1999. These data were used as ground truth for the correlation analysis with different indices of image band values. Landsat 7 images were used to calculate soil salinity indices such as the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Salinity Index (NDSI), the Soil Adjusted Vegetation Index (SAVI), and the Soil Salinity Index. A statistical analysis of the electrical conductivity (EC1:5 dS/m) and the environmental indices acquired from the Landsat 7 image was performed. Exponential regression was used to find the best indices, which were NDVI, NDSI, SAVI, and SI5, with a (NDVI R2=0.9, NDSI R2=0.9, SAVI R2=0.68, and SI5 R2=0.77) correlation with field truth data. A salinity map of the Habra plain was generated using this index with an acceptable level of accuracy. This study found that Landsat 7 images can effectively monitor soil salinity levels. The results of this study are relevant for agricultural operations, planners, and farmers by mapping and monitoring the soil salinity contamination. Finally, it is concluded that using remote sensing in salinity detection and mapping is highly significant.

Expansion of salt-induced soil in agriculture lands is major cause of desertification and crop yield loss. Detection of salt affected soils and land cover changes assessment is crucial to better understand the correlations between diverse land cover units and salinity.Salinity variations at local scales can be studied using a combination of satellite image analysis and field measurements.The purpose of this study is to assess and detect the variations of salt content in the soil and land cover in the study area located in Zaghouan Governorate (Tunisia). This includes the examination of soil salinization effect on land cover change between 2017 and 2021.Satellite imagery has shown that the study area has different soil cover units.Based on the combination between satellite imagery analysis and soil sampling, findings reveal a slight correlation between salinity levels and land cover classes.The study of soil salinity changes shows that salt content rates increased from 5.44 % in 2017 to 6.7 % in 2021, while the land cover has reversely changed to 9.95 % for the same time period.Furthermore, results related to the relationship between SI and NDVI indicates that soil salinity changes have a significant and direct impact on land cover over time.In addition to in situ data, comparing Sentinel-2 and Landsat 8 satellite data shows a great potential for detecting and monitoring salt-affected soils

Capability of Sentinel-2 MSI data for monitoring and mapping of soil salinity in dry and wet seasons in the Ebinur Lake region, Xinjiang, China

Wind erosion and dust emissions affect regions of the world with sparse vegetation cover or affected by agricultural practices that expose the soil surface to wind action. Although several studies have investigated the impact of soil moisture, land use and land cover on soil susceptibility to wind erosion and dust emissions, the effect of surface soil salinity and sodicity on dust emissions, remains poorly understood. Salt accumulation in agricultural soils is a major concern in agroecosystems with high evaporative demand, shallow water tables or irrigated with water rich in dissolved solids. Recent studies have focused on the effect of soil salinity on soil erodibility in dry atmospheric conditions, while the effect of soil salinity and sodicity in more humid conditions still needs to be investigated. Here we use wind tunnel tests to study the effect of high atmospheric humidity on wind erodibility and particulate matter emissions under saline and sodic conditions. We find that the threshold velocity for wind erosion significantly increases with increasing soil salinity and sodicity, provided that the soil crust formed by soil salts is not disturbed. Indeed, with increasing soil salinity, the formation of a soil crust of increasing strength is observed, leading to an increase in the threshold wind velocity and a consequent decrease in particulate emissions. Interestingly, after the threshold velocity was exceeded, soil crusts were readily ruptured by saltating sand grains resulting in comparable or sometimes even higher particulate matter emissions in saline and sodic soils compared to their untreated (‘control’) counterparts which can be explained by salinity‐induced aggregation and sodicity‐driven clay dispersion effects. Lastly, understanding the role of atmospheric humidity under changing climate scenarios will help to modulate the wind erosion processes in saline‐sodic soils and will help mitigate better dust emissions and soil management policies in arid and semi‐arid climate zones.

Soil salinization is considered one of the disasters that have significant effects on agricultural activities in many parts of the world, particularly in the context of climate change and sea level rise. This problem has become increasingly essential and severe in the Mekong River Delta of Vietnam. Therefore, soil salinity monitoring and assessment are critical to building appropriate strategies to develop agricultural activities. This study aims to develop a low-cost method based on machine learning and remote sensing to map soil salinity in Ben Tre province, which is located in Vietnam's Mekong River Delta. This objective was achieved by using six machine learning algorithms, including Xgboost (XGR), sparrow search algorithm (SSA), bird swarm algorithm (BSA), moth search algorithm (MSA), Harris hawk optimization (HHO), grasshopper optimization algorithm (GOA), particle swarm optimization algorithm (PSO), and 43 factors extracted from remote sensing images. Various indices were used, namely, root mean square error (RMSE), mean absolute error (MAE), and the coefficient of determination (R2) to estimate the efficiency of the prediction models. The results show that six optimization algorithms successfully improved XGR model performance with an R2 value of more than 0.98. Among the proposed models, the XGR-HHO model was better than the other models with a value of R2 of 0.99 and a value of RMSE of 0.051, by XGR-GOA (R2 = 0.931, RMSE = 0.055), XGR-MSA (R2 = 0.928, RMSE = 0.06), XGR-BSA (R2 = 0.926, RMSE = 0.062), XGR-SSA (R2 = 0.917, 0.07), XGR-PSO (R2 = 0.916, RMSE = 0.08), XGR (R2 = 0.867, RMSE = 0.1), CatBoost (R2 = 0.78, RMSE = 0.12), and RF (R2 = 0.75, RMSE = 0.19), respectively. These proposed models have surpassed the reference models (CatBoost and random forest). The results indicated that the soils in the eastern areas of Ben Tre province are more saline than in the western areas. The results of this study highlighted the effectiveness of using hybrid machine learning and remote sensing in soil salinity monitoring. The finding of this study provides essential tools to support farmers and policymakers in selecting appropriate crop types in the context of climate change to ensure food security.

Ciamis Regency experienced regional proliferation and land cover change dynamics. Information on existing and predicted land cover is needed to ensure alignment with spatial planning (RTRW). This study aims to: (1) analyze land cover change in Ciamis and its proliferated regions (2000-2018); (2) examine land cover prediction in 2031; and (3) analyze the alignment between land cover (2018) and predicted land cover (2031) with RTRW. Spatial analysis and Land Change Modeler were employed using ArcGIS and Idrisi Selva software. The alignment between land cover and RTRW was identified by a logic matrix based on land rent in 2031 with RTRW of Ciamis Regency, Banjar City, and Pangandaran Regency. The results showed that dry land dominated the 173,949 ha land cover in 2000 and increased to 183,231 ha in 2018. During 2000-2018, there was a decreasing trend in rice fields. The predicted land cover (2031) based on BAU and RTRW scenario shows that rice fields tend to decrease, while the built-up area has a significant increase. The alignment of RTRW with land cover (2018) is 97%, whereas its alignment with predicted land cover (2031) is 96% (BAU) and 93% (RTRW). The results are beneficial for land management and controlling land conversion.

Assessing the performance of machine learning algorithms for soil salinity mapping in Google Earth Engine platform using Sentinel-2A and Landsat-8 OLI data