Jupyter Notebook for threshold detection and ecological zoning using the Priority Index framework

This repository contains the comprehensive Python source code and analytical framework developed for the study "A nonlinear threshold-based study on ecological zoning for water conservation in Guangdong Province, China". It provides a reproducible pipeline that integrates spatial analysis with advanced machine learning to delineate environmental management zones and interpret the non-linear drivers of water conversation.Methodological Workflow: The codebase is structured around two independent but complementary analytical phases:Spatial Zoning and Priority Index (PI): Processes multi-source environmental and socio-economic variables (e.g., Temperature, Precipitation, DEM, GDP, Population) to compute a spatially-weighted Priority Index. It couples these metrics with Gi* spatial statistics to classify regions into specific ecological management types.Machine Learning and SHAP Interpretation: Utilizes an Optuna-optimized XGBoost model to evaluate the complex, non-linear impacts of the driving factors. It includes extensive SHapley Additive exPlanations (SHAP) analyses to map spatial SHAP distributions, extract zero-SHAP thresholds, and visualize feature interactions.Repository Structure:config.py: Global configurations, dynamic file mappings, and spatial weights.spatial_utils.py: Core geospatial algorithms for raster alignment and PI computation.shap_utils.py: Utilities for SHAP calculation, polynomial fitting for thresholds, and spatial visualization.main_zoning.py: Execution script for the spatial zoning workflow.main_xgboost.py: Execution script for model training and SHAP interpretation.Usage Requirements: Researchers looking to reproduce or adapt this framework should place their aligned raster datasets (.tif), spatial grids (.shp), and tabular data (.csv) into the designated local ./data/inputs/ directory. For detailed setup instructions and dependencies, please refer to the included README.md file.本仓库包含为学术研究《中国广东省水源涵养生态分区的非线性阈值研究》开发的完整Python源代码与分析框架。该框架提供了可复现的工作流，整合空间分析与先进机器学习技术，用于划定环境管理分区并解析水源涵养的非线性驱动因素。 方法学工作流：代码库围绕两个独立且互补的分析阶段构建： 1. 空间分区与优先级指数（Priority Index, PI）：处理多源环境与社会经济变量（如气温、降水、数字高程模型（Digital Elevation Model, DEM）、国内生产总值（Gross Domestic Product, GDP）、人口），以计算空间加权优先级指数。该模块将上述指标与Gi*空间统计方法相结合，将区域划分为特定的生态管理类型。 2. 机器学习与SHAP解释：采用经Optuna优化的极限梯度提升（Extreme Gradient Boosting, XGBoost）模型，评估驱动因子的复杂非线性影响。其中包含大量Shapley可加解释（SHapley Additive exPlanations, SHAP）分析，用于绘制空间SHAP分布、提取零SHAP阈值，并可视化特征交互作用。 仓库文件结构： config.py：全局配置、动态文件映射与空间权重设置。 spatial_utils.py：用于栅格对齐与优先级指数计算的核心地理空间算法工具。 shap_utils.py：用于SHAP计算、阈值多项式拟合与空间可视化的工具函数。 main_zoning.py：空间分区工作流的执行脚本。 main_xgboost.py：模型训练与SHAP解释的执行脚本。 使用要求：欲复现或适配本框架的研究人员，需将已对齐的栅格数据集（.tif格式）、空间矢量网格数据（.shp格式）与表格数据（.csv格式）放入指定的本地./data/inputs/目录中。有关详细的安装说明与依赖项，请参阅附带的README.md文件。