The previous comparative studies on watersheds were mostly based on the comparison of dispersive characteristics, which lacked systemicity and causality. We proposed a causal structure-based framework for basin comparison based on the Bayesian network (BN) and focus on the basin-scale water–energy–food–ecology (WEFE) nexus. We applied it to the Syr Darya River basin (SDB) and the Amu Darya River basin (ADB), of which poor water management caused the Aral Sea disaster. The causality of the nexus was effectively compared and universality of this framework was discussed. In terms of changes in the nexus, the sensitive factor for the water supplied to the Aral Sea changed from the agricultural development during the Soviet Union period to the disputes in the WEFE nexus after the disintegration. The water–energy contradiction of the SDB is more severe than that of the ADB, partly due to the higher upstream reservoir interception capacity. It further made management of the winter surplus water downstream of the SDB more controversial. Due to this, the water–food–ecology conflict between downstream countries may escalate and turn into a long-term chronic problem. Reducing water inflow to depressions and improving the planting structure prove beneficial to the Aral Sea ecology, and this effect of the SDB is more significant. The construction of reservoirs on the Panj River of the upstream ADB should be cautious to avoid an intense water–energy conflict such as the SDB's. It is also necessary to promote the water-saving drip irrigation and to strengthen the cooperation.

Despite the growing interest among researchers, satellite-based prediction of soil salinity remains highly uncertain. The improvements in prediction accuracy reported in previous studies are usually limited to a single area. We performed a meta-analysis of regional satellite-based soil salinity predictions combined with in situ soil sampling and machine learning. Based on $R^{2}$ and root-mean-square error (RMSE) collected, we evaluated the effects of various features on the model accuracy and established a Bayesian network to evaluate the joint causal effect of multifeatures. Most significant differences were found in soil sampling schemes and characteristics of the study area, including the mean and variability (averaged $R^{2}$ of 0.75 for soil sample sets with lower salinity variation and 0.62 for others) of the salinity, climate type ( $R^{2}$ of 0.64 in arid areas and 0.74 in others), soil texture ( $R^{2}$ of 0.66 in sandy areas and 0.57 in others), and the interval between sampling date and satellite data acquisition date ( $R^{2}$ of 0.53 under the condition of over 15 days and 0.65 in others). Generally, using different satellite data has limited effects on model performance among which Sentinel-2 performed better ( $R^{2} $ = 0.72) than Landsat ( $R^{2} $ = 0.66). The sampling of subsamples for each sample should focus on their subpixel-scale spatial heterogeneity across satellite data rather than the number of subsamples. It is also necessary to select appropriate vegetation and salinity indices for different satellite data under different vegetation conditions. Among algorithms, random forests ( $R^{2} $ = 0.70) and support vector machines ( $R^{2} $ = 0.71) performed best.

Simulating the carbon-water fluxes at more widely distributed meteorological stations based on the sparsely and unevenly distributed eddy covariance flux stations is needed to accurately understand the carbon-water cycle of terrestrial ecosystems. We established a new framework consisting of machine learning, determination coefficient (R²), Euclidean distance, and remote sensing (RS), to simulate the daily net ecosystem carbon dioxide exchange (NEE) and water flux (WF) of the Eurasian meteorological stations using a random forest model or/and RS. The daily NEE and WF datasets with RS-based information (NEE-RS and WF-RS) for 3774 and 4427 meteorological stations during 2002–2020 were produced, respectively. And the daily NEE and WF datasets without RS-based information (NEE-WRS and WF-WRS) for 4667 and 6763 meteorological stations during 1983–2018 were generated, respectively. For each meteorological station, the carbon-water fluxes meet accuracy requirements and have quasi-observational properties. These four carbon-water flux datasets have great potential to improve the assessments of the ecosystem carbon-water dynamics.

Net ecosystem exchange (NEE) is an important indicator of carbon cycling in terrestrial ecosystems. Many previous studies have combined flux observations and meteorological, biophysical, and ancillary predictors using machine learning to simulate the site-scale NEE. However, systematic evaluation of the performance of such models is limited. Therefore, we performed a meta-analysis of these NEE simulations. A total of 40 such studies and 178 model records were included. The impacts of various features throughout the modeling process on the accuracy of the model were evaluated. Random forests and support vector machines performed better than other algorithms. Models with larger timescales have lower average R² values, especially when the timescale exceeds the monthly scale. Half-hourly models (average R² = 0.73) were significantly more accurate than daily models (average R² = 0.5). There are significant differences in the predictors used and their impacts on model accuracy for different plant functional types (PFTs). Studies at continental and global scales (average R² = 0.37) with multiple PFTs, more sites, and a large span of years correspond to lower R² values than studies at local (average R² = 0.69) and regional (average R² = 0.7) scales. Also, the site-scale NEE predictions need more focus on the internal heterogeneity of the NEE dataset and the matching of the training set and validation set.

Shortfalls in regular evapotranspiration (ET) monitoring and evaluation pose a huge challenge to agricultural water resource distribution in arid Central Asia (CA). In this study, a first detailed regional assessment of GLEAM, ERA5, MERRA2, CLSM and NOAH ET products in CA was performed by systematically implementing the triple collocation (TC) method, in which about 36,936 grid cells for each ET data (within a six-triplet design) were collocated, at 0.25° and with monthly resolutions during 2003–2020. The reliability of the strategy adopted was confirmed in four arid biomes using standard evaluation metrics (R, RMSE and BIAS), and by spatiotemporal cross-validation of the six ET triplets across CA. Results show that the systematic TC method produced more robust ET product assessment metrics with reduced RMSEs compared to the initial ET product validation using in-situ, which showed weak-positive correlation and high negative bias-range (i.e., −21.02 ≤ BIAS < 16 mm) in the four arid biomes of CA. The spatial cross-validation by TC showed that the magnitude of ET random errors significantly varies, and confirms the systematic biases with site-scale measurements. The highest ET uncertainties by CLSM (27.43%), NOAH (29.16%), MERRA2 (38.28%), ERA5 (36.75), and GLEAM (41%) were more evident in the shrubland, cropland, grassland, cropland again, and desert biomes, respectively. Moreover, error magnitudes in high altitudes (Tianshan Mountain range) are generally lower than in plain-desert areas. All ET products spatially captured ET dynamics over CA, but none simultaneously outperformed the other. These findings are invaluable in the utilization of the assessed ET products in supporting regional water resource management, particularly in CA.