Spatiotemporal dynamics of vegetation cover: integrative machine learning analysis of multispectral imagery and environmental predictors

This study investigates the spatiotemporal dynamics of Fractional vegetation cover (FVC) influenced by environmental changes across diverse landscapes. High-resolution multispectral imagery from the Landsat series was used to analyze the interactions between FVC and climatic variables, supporting targeted conservation efforts. The study utilized precipitation data from CHIRPS, temperature, vapor pressure deficit (VPD), solar radiation (SR), and soil moisture (SM) from Terra Climate from 2000 to 2023. By integrating these environmental datasets with machine learning algorithms, such as Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), alongside traditional statistical analyses, the research models and predicts FVC dynamics across the regions. The RF model achieved an R² of 0.8971, a Mean Squared Error (MSE) of 0.003228, and a Root Mean Squared Error (RMSE) of 0.056815 on testing datasets, indicating high predictive accuracy. Similarly, XGBoost showed strong performance with an R² of 0.8695, an MSE of 0.004008, and an RMSE of 0.063310. The study reveals a statistically significant increase in FVC in the Khyber Pakhtunkhwa (KPK) region, with an annual growth rate of 0.002749/year and an R² value of 81.93% (p < 0.01), suggesting substantial and consistent vegetation enhancement. In contrast, Azad Jammu and Kashmir (AJK) exhibits a more variable vegetation response, with an even higher growth rate of 0.004408 annually but a lower R² of 73.44% (p < 0.01), reflecting uneven growth across different areas. Gilgit-Baltistan (GB) shows marginal vegetation growth with a rate of 0.000290/year, mainly due to its high-altitude terrain, extreme climatic conditions, and limited water availability. This study highlights the need to understand spatiotemporal vegetation dynamics under environmental changes and applies an innovative machine learning approach to predict FVC across diverse landscapes. The findings provide critical insights for region-specific conservation strategies, supporting policymakers in designing adaptive land-use policies and enhancing the precision of environmental management practices.

Text 122311.pdf - Published Version Restricted to Repository staff only Download (8MB) Official URL or Download Paper: https://link.springer.com/article/10.1007/s12145-0...

Actions (login required)

View Item