Overview This repository contains the Jupyter Notebook (Main_figure.ipynb) used to analyze global water balance anomalies and generate the figures for the associated research paper. The notebook is self-contained, with output figures displayed immediately following their respective code blocks to facilitate reproducibility and visualization. ContentsThe notebook covers the following analyses: Figure 1: Global time series analysis (1991-2024) of streamflow, surface water storage, precipitation, and temperature, including uncertainty quantification. Figure 2: Continental-scale spatial analysis of hydrological anomalies, featuring basin-level maps and contribution bar charts. Figure 3: SHAP (SHapley Additive exPlanations) attribution analysis identifying dominant climatic drivers (Precipitation, Temperature, LAI) using ternary color mapping. Figure 4: Regional precipitation anomaly analysis across CORDEX domains using a multi-product ensemble (JRA-3Q, ERA5, MSWEP). Workflow & Data Integration This notebook represents the final visualization stage of our study's analytical pipeline. It integrates the outputs from the previous machine learning and data processing steps: Data Ingestion: The notebook reads the consolidated datasets generated from our prior steps, specifically: Simulated monthly streamflow data (from the ML framework). Satellite-derived surface water storage and hydroclimatic variables (from this dataset). Basin-level SHAP attribution values (from this dataset). Anomaly Calculation & Aggregation: The code processes the raw time-series data to calculate baseline climatology (e.g., 1991–2022 averages) and computes the abrupt anomalies observed during the 2023–2024 record warming period. It aggregates basin-level data (HydroSHEDS Level 4) to continental and global scales for macroscopic analysis. Spatial & Statistical Rendering: Utilizing geospatial libraries (geopandas, cartopy), the notebook maps the basin-level anomalies and SHAP dominant drivers onto global projections. It also calculates and plots the uncertainty bounds (standard deviations) across the multi-product ensembles. Technical Details Language: Python 3.12 Key Dependencies: pandas, numpy, matplotlib, geopandas, cartopy, xarray, rioxarray, py-cordex. Installing the required dependencies via `pip` or `conda` typically takes 10–15 minutes on a standard desktop computer. Expected Run Time: Executing the entire notebook cell-by-cell takes less than 15 minutes on a standard desktop computer.