Search

• The flux footprint and physical mechanism are integrated into the convolutional neural network model. • Coupled flux footprint is effective for enhancing the ET estimates in highly heterogeneous and variable wind direction regions. • Coupled physical mechanisms significantly improve the simulation capability of extreme ET events. • Urban ET presents a greater spatial variability with the lowest ET in impervious surface. • The decline in urban ET is due to the combined effect of the decrease in radiation and the increase in impervious surfaces. Estimating urban evapotranspiration (ET) is of great significance for urban water resource allocation and assessing the urban heat island effect. However, most current urban ET models are based on the energy balance theory to estimate urban ET. These models lead to significant errors in urban ET simulation due to the surface heterogeneity and the existence of anthropogenic heat fluxes in urban areas. To solve this issue, this study proposes a modified machine learning-based urban ET method that can estimate urban ET at the site and regional scales. To better characterize the heterogeneity of urban surfaces, the flux footprint of in-situ ET and physical mechanism of ET process are integrated into the convolutional neural network (CNN) model. The modified CNN model is tested in a fast-developing city: Shenzhen, China based on two Eddy Correlation (EC) observations. The verification results indicated that coupling flux footprint and physical mechanism into the CNN model could effectively improve the accuracy of urban ET simulation at the site scale. The modified CNN model significantly reduced the root-mean-square-error (RMSE) of 25.8 W/m2 and increased the determination coefficient (R2) of 0.17 compared to the CNN-O model (The CNN-O model is defined as the CNN model do not integrate flux footprint and physical mechanism of ET). Further analyses suggested that fusing flux footprint data into a machine learning model helps enhance ET estimation in regions with high heterogeneity and highly variable wind directions. Moreover, the integration of physical mechanisms significantly enhanced the model capability to simulate extreme ET events. The modified CNN model is further applied to map the spatial distribution of urban ET and reconstruct long-term urban ET changes. The spatial pattern of urban ET exhibited large spatial variability, where the urban ET in water bodies (mean λET larger than 480 W/m2) and vegetation-covered areas (mean λET larger than 260 W/m2) are substantially higher than the impervious surfaces (mean λET less than 30 W/m2). Long-term trend analyses demonstrated that urbanization resulted in decline in urban ET. The average decreasing rate of urban ET is 1.61 mm/yr (P < 0.05), with a 18 % decrease relative to the long-term ET average. The leading causes for the decline of urban ET are the increased impervious surfaces and the decreased radiation. This study improved the simulation accuracy of urban ET and revealed the response of urban ET to urbanization. [ABSTRACT FROM AUTHOR]

The accurate estimation of coastal water quality parameters (WQPs) is crucial for decision-makers to manage water resources. Although various machine learning (ML) models have been developed for coastal water quality estimation using remote sensing data, the performance of these models has significant uncertainties when applied to regional scales. To address this issue, an ensemble ML-based model was developed in this study. The ensemble ML model was applied to estimate chlorophyll-a (Chla), turbidity, and dissolved oxygen (DO) based on Sentinel-2 satellite images in Shenzhen Bay, China. The optimal input features for each WQP were selected from eight spectral bands and seven spectral indices. A local explanation strategy termed Shapley Additive Explanations (SHAP) was employed to quantify contributions of each feature to model outputs. In addition, the impacts of three climate factors on the variation of each WQP were analyzed. The results suggested that the ensemble ML models have satisfied performance for Chla (errors = 1.7%), turbidity (errors = 1.5%) and DO estimation (errors = 0.02%). Band 3 (B3) has the highest positive contribution to Chla estimation, while Band Ration Index2 (BR2) has the highest negative contribution to turbidity estimation, and Band 7 (B7) has the highest positive contribution to DO estimation. The spatial patterns of the three WQPs revealed that the water quality deterioration in Shenzhen Bay was mainly influenced by input of terrestrial pollutants from the estuary. Correlation analysis demonstrated that air temperature (Temp) and average air pressure (AAP) exhibited the closest relationship with Chla. DO showed the strongest negative correlation with Temp, while turbidity was not sensitive to Temp, average wind speed (AWS), and AAP. Overall, the ensemble ML model proposed in this study provides an accurate and practical method for long-term Chla, turbidity, and DO estimation in coastal waters. • A machine learning (ML) based ensemble model was proposed for the estimation of water quality parameters. • The feature selection was conducted for the ML-based ensemble model. • The ML-based ensemble model improved the estimation accuracy of non-optical active constituent. • SHAP method was employed to quantify the contribution of the input features. • Climate factors driving mechanism was considered on interannunal dynamics of water quality. [ABSTRACT FROM AUTHOR]