Assessing Climate-Sensitive Nutrient Pollution Hazards in Urban-Rural Transitional Watersheds: An Artificial Intelligence and Remote Sensing Data Assimilation Approach

This dataset contains the supplementary materials for the paper "Assessing Climate-Sensitive Nutrient Pollution Hazards in Urban-Rural Transitional Watersheds: An Artificial Intelligence and Remote Sensing Data Assimilation Approach" by Jingyuan Xue, Can Yuan, Xiaoliang Ji, Michael L. Grieneisen, Minghua Zhang, Liuyue He. Contents: The supplementary file includes 8 figures (Figure S1-S8) and 9 tables (Table S1-S9) that support the main findings of the article. Figure S1. Field photograph of sampling site 17. Figure S2. Flowchart of the Random Forest Regression (RFR) workflow. Figure S3. Delineation of the 221 sub-watersheds in the study area. Figure S4. Effects of the number of trees (k) and the number of split variables (m) on out-of-bag (OOB) mean square error for (a) TN and (b) TP. Figure S5. Scatter plots of the optimal regression equations used in the RSI models for each sampling period. (Panels compare measured and remotely sensed values for TN and TP, with fitted regression curves overlaid.) Figure S6. Long-term simulations of TN concentrations in the WRTR watershed from January 2021 to December 2023 using the RFR, RSI, and ASS models. Figure S7. Long-term simulations of TP concentrations in the WRTR watershed from January 2021 to December 2023 using the RFR, RSI, and ASS models. Figure S8. Receiver Operating Characteristic (ROC) curves for construction land and water area thresholds in relation to TN (a-b) and TP (c-d). (The diagonal lines represent random classification, and curves farther above the diagonal indicate stronger discriminatory performance.) Table S1. Watershed characteristic variables and data source information Table S2. Performance of four machine learning algorithms Table S3. 254 band combination forms required for remote sensing inversion modelling Table S4. Optimal spectral band combinations used to construct RSI models for TN and TP Table S5. Comparison of regression functions used in RSI models for TN and TP across five sampling periods Table S6. RMSE of assimilated and measured TN and TP concentration under different ensembles Table S7. RMSE of assimilated and measured concentration for (a) TN and (b) TP under different error levels (Q = model error, R = observation error) Table S8. Changes in the model RMSE after each assimilation step Table S9. ROC curve analysis for (a) TN and Construction land, (b) TP and Construction land (c) TN and water area, (d) TP and water area