Remote sensing and meteorological data-driven prediction of seasonal changes and optimization of urban roadside plant landscape configuration

This paper proposed a remote sensing and meteorological data-driven framework for optimizing urban roadside plant landscape configurations by addressing seasonal climate variations and plant response delays. The methodology combined remote sensing NDVI data and meteorological reanalysis to predict seasonal changes, utilizing the LightGradient Boosting Machine (LightGBM) model to forecast Normalized Difference Vegetation Index (NDVI) time series with a Coefficient of Determination (R²) of 0.87. A dynamic matching mechanism was developed to assess Plant Adaptability (PA), incorporating species heat and drought tolerance. A Multi-Objective Optimization (MOO) process, involving Thermal Comfort (TC), Ecological Suitability (ES), and Color Coverage (CC), was applied using a Genetic Algorithm (GA) to optimize plant species and planting density. Experimental results showed the method achieved PA scores of 0.82, 0.79, and 0.74 for main roads, secondary roads, and sidewalks, respectively, and a CC of 90% in July. This approach demonstrated the feasibility of dynamic, data-driven optimization for urban greening, enhancing TC, ecological compatibility, and aesthetic value across various road types.