Leveraging GEE and machine learning algorithm in dynamic modeling of eco-environmental quality

Abstract

The impact of human activities and natural resources on ecological quality (EQ) is a global issue facing the world. To address these problems, this study used a Remote Sensing Ecological Index (RSEI) to assess the temporal and geographical dynamics of Iran's eco-environmental conditions. To prepare this index, Principal Component Analysis (PCA) was utilized on four main indices, namely Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI), Wetness Index (WET), and Normalized Difference Bare Soil Index (NDBSI), using the Google Earth Engine (GEE) platform. This study suggests an innovative approach to modeling RSEI using a Random Forest (RF) model and an independent variable to assess EQ, considering the drawbacks of traditional methods. The results illustrate significant variations in the EQ from 2018 to 2022. The percentage area with a constant EQ decreased from 63 % in 2018–2019 to 53 % in 2021–2022. Areas with improved EQ experienced dramatic fluctuations, peaking at 75.37 % in 2020–2021 but plummeting to just 1 % in 2021–2022. Conversely, areas with worsened EQ started relatively low at 22 % in 2018–2019, increased to 3 % in 2019–2020, then rose sharply to 45 % by 2021–2022. These trends highlight a substantial shift, particularly in 2020–2021, with a remarkable improvement, followed by a steep decline in improvement and a significant rise in deterioration in the following year. In addition, RSEI values showed that 2018, among the years studied, had a moderate level of ecological vulnerability in the western and northwestern parts, which trended towards a weak level of vulnerability. Moreover, the results of the environmental factors revealed that evapotranspiration and PM2.5 were critical contributors to EQ. These assessments emphasize the need for targeted strategies to reduce environmental degradation, including vegetation restoration, pollution control, and sustainable land use. This study provides a roadmap for land managers and planners to identify vulnerable areas and implement optimal conservation strategies.