NNwobi-354/semiarid-land-degradation-trajectory

Machine learning and multi-source geospatial analysis of semi-arid land degradation trajectories across Northeast Nigeria (Borno and Yobe) and Kalmykia, Russia. # Multi-Scale Ecohydrologic Dynamics & Vegetation Tipping Points Across Semi-Arid Biomes This repository contains the complete spatial data processing pipeline, machine learning architectures, panel econometrics scripts, and replication data for the multi-biome comparative study: **"Multi-scale ecohydrologic dynamics and vegetation resilience across semi-arid biomes: A comparative analysis of the Nigerian Sahel and the Caspian Steppe."** --- ## Quick Replication via Google Colab To reproduce the full analytical workflow without local installation: 1. Open the primary Jupyter Notebook located in the root directory: **`Nigeria_Kalmykia_Land_Degradation_ML_Analysis.ipynb`**. 2. Click the **"Open in Colab"** badge at the top of the notebook (or at the top of this README). 3. In Google Colab, select **Runtime > Run all** to execute all scripts sequentially. --- ## Repository Overview This project investigates non-linear vegetation tipping points, atmospheric drying constraints, and soil moisture boundaries across two contrasting semi-arid biomes (2003–2025): 1. **Supply-Limited Regime:** Borno and Yobe States, Sub-Saharan Sahel (Nigeria). 2. **Demand-Driven Regime:** Republic of Kalmykia, Caspian Steppe (Russian Federation). The analytical framework integrates Google Earth Engine (GEE) spatial harmonization, spatial block cross-validated XGBoost modeling, TreeSHAP feature attributions, and Two-Way Fixed Effects (TWFE) panel econometrics. --- ## Repository Structure