Estimating ecosystem evapotranspiration (ET) is important to understanding the global water cycle and to study land‐atmosphere interactions. We developed a physics constrained machine learning (ML) model (hybrid model) to estimate latent heat flux (LE), which conserves the surface energy budget. By comparing model predictions with observations at 82 eddy covariance tower sites, our hybrid model shows similar performance to the pure ML model in terms of mean metrics (e.g., mean absolute percent errors) but, importantly, the hybrid model conserves the surface energy balance, while the pure ML model does not. A second key result is that the hybrid model extrapolates much better than the pure ML model, emphasizing the benefits of combining physics with ML for increased generalizations. The hybrid model allows inferring the structural dependence of ET and surface resistance (r_s), and we find that vegetation height and soil moisture are the main regulators of ET and r_s.

Water resources, which are considerably affected by land use/land cover (LULC) and climate changes, are a key limiting factor in highly vulnerable ecosystems in arid and semi-arid regions. The impacts of LULC and climate changes on water resources must be assessed in these areas. However, conflicting results regarding the effects of LULC and climate changes on runoff have been reported in relatively large basins, such as the Jinghe River basin (JRB), which is a typical catchment (> 45 000 km²) located in a semi-humid and arid transition zone on the central Loess Plateau, northwest China. In this study, we focused on quantifying both the combined and isolated impacts of LULC and climate changes on surface runoff. We hypothesized that under climatic warming and drying conditions, LULC changes, which are primarily caused by intensive human activities such as the Grain for Green Program, will considerably alter runoff in the JRB. The Soil and Water Assessment Tool (SWAT) was adopted to perform simulations. The simulated results indicated that although runoff increased very little between the 1970s and the 2000s due to the combined effects of LULC and climate changes, LULC and climate changes affected surface runoff differently in each decade, e.g., runoff increased with increased precipitation between the 1970s and the 1980s (precipitation contributed to 88 % of the runoff increase). Thereafter, runoff decreased and was increasingly influenced by LULC changes, which contributed to 44 % of the runoff changes between the 1980s and 1990s and 71 % of the runoff changes between the 1990s and 2000s. Our findings revealed that large-scale LULC under the Grain for Green Program has had an important effect on the hydrological cycle since the late 1990s. Additionally, the conflicting findings regarding the effects of LULC and climate changes on runoff in relatively large basins are likely caused by uncertainties in hydrological simulations.

Overlying water, sediment, rhizosphere sediment and mangrove seedlings in the Futian mangrove forest were analyzed for heavy metals. The results showed that mangrove plant acidified sediment and increased organic matter contents. Except for chromium (Cr), nickel (Ni) and copper (Cu) in Aegiceras corniculatum sediment, heavy metals in all sediments were higher than in overlying water, rhizosphere sediment and mangrove root. Heavy metals in Avicennia marina sediments were higher than other sediments. The lower heavy metal biological concentration factors (BCFs) and translocation factors (TFs) indicated that mangrove plant adopted exclusion strategy. The geo-accumulation index, potential ecological risk index and risk assessment code (RAC) demonstrated that heavy metals have posed a considerable ecological risk, especially for cadmium (Cd). Heavy metals (Cr, Ni, Cu and Cd) mainly existed in the reducible fractions. These findings provide actual heavy metal accumulations in sediment-plant ecosystems in mangrove forest, being important in designing the long-term management and conservation policies for managers of mangrove forest.