Accurate estimation of leaf chlorophyll content (Chl) is essential for monitoring vegetation photosynthetic capacity and simulating carbon budgets at regional to global scales. Under sparse canopy conditions, however, Chl retrieval is often hindered by strong soil background effects. Vegetation indices (VIs) remain widely used tools for Chl estimation, yet their performance in sparse canopies has not been comprehensively evaluated. In this study, coupled leaf and canopy radiative transfer models were used to simulate canopy reflectance under sparse low vegetation and sparse forest scenarios. We further analyzed the effects of canopy structure, soil background, and sun-observer geometry, and evaluated the performance of multiple VIs in retrieving Chl. To validate these results, the large-scale remote sensing data and image simulation framework (LESS) model and field-measured dataset were employed. Our results show that, for sparse low vegetation, Chl, leaf area index (LAI), and soil background are the dominant drivers of canopy reflectance variability, whereas for sparse forests, Chl, crown diameter (CD), viewing zenith angle (VZA), and stem density (SD) exert the greatest influence. Notably, the Sentinel-2 red-edge 1 (RE1) band simulated from model outputs exhibited the highest sensitivity to Chl variations under sparse canopy conditions. Among the 101 VIs evaluated, Datt99, REP, REP3, and Macc01 demonstrated the most robust performance. These findings provide a scientific basis for selecting VIs suitable for Chl estimation in sparse canopies and support vegetation monitoring in arid and semi-arid ecosystems as well as during early crop growth stages.