Estimates of tree species diversity via traditional optical remote sensing are based only on the spectral variation hypothesis (SVH); however, this approach does not account for the vertical structure of a forest. The relative height (RH) indices derived from GEDI spaceborne LiDAR provide vertical vegetation structure information through waveform decomposition. Although RH indices have been widely studied, the optimal RH index for tree species diversity estimation remains unclear. This study integrated GF-1 optical imagery and GEDI LiDAR data to estimate tree species diversity in a warm temperate forest. First, random forest plus residual kriging (RFRK) was employed to achieve wall-to-wall mapping of the GEDI-derived indices. Second, recursive feature elimination (RFE) was applied to select relevant spectral and LiDAR features. The random forest (RF), support vector machine (SVM), and k-nearest neighbor (kNN) methods were subsequently applied to estimate tree species diversity through remote sensing data. The results indicated that multisource data achieved greater accuracy in tree species diversity estimation (average R² = 0.675, average RMSE = 0.750) than single-source data (average R² = 0.636, average RMSE = 0.754). Among the three machine learning methods, the RF model (R² = 0.760, RMSE = 2.090, MAE = 1.624) was significantly more accurate than the SVM (R² = 0.571, RMSE = 2.556, MAE = 1.995) and kNN (R² = 0.715, RMSE = 2.084, MAE = 1.555) models. Moreover, mean_mNDVI, mean_RDVI, and mean_Blue were identified as the most important spectral features, whereas RH30 and RH98 were crucial features derived from LiDAR for establishing models of tree species diversity. Spatially, tree species diversity was high in the west and low in the east in the study area. This study highlights the potential of integrating optical imagery and spaceborne LiDAR for tree species diversity modeling and emphasizes that low RH indices are most indicative of middle- to lower-canopy tree species diversity.

Ecosystem service value (ESV) is a critical indicator of regional ecological well-being. Assessing and forecasting ESV are essential for achieving the coordinated development of environmental and economic systems. This study employs the SD-PLUS model, integrating Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs) to assess the spatiotemporal dynamics of land use and land cover change (LUCC), as well as ESV in Zhengzhou from 2030 to 2040. It analyses the impact of various driving factors on ESV and examines the spatial correlations among ecosystem services across different regions. The results indicate that the total ESV is expected to decrease by 73.53 × 10⁷ yuan, primarily due to significant reductions in cropland and water areas. By 2040, ESV is projected to increase by 14.51 × 10⁷ yuan under the SSP126 scenario, decrease by 73.18 × 10⁷ yuan under the SSP585 scenario, and show a moderate decline under the SSP245 scenario. Climate factors, transportation location, and topographical features have a significantly positive impact on ESV, while environmental and socioeconomic factors exert a negative influence. The analysis of interrelationships among ecosystem services shows that synergies dominate, especially between supporting and cultural services, with only localised trade-offs observed. These findings contribute valuable insights for the development of scientifically sound, well-reasoned, and efficient strategies for ecological conservation and sustainable development.