Comparison of Landsat-8 and Sentinel-2 Imagery for Modeling Gross Primary Productivity of Tea Ecosystem

Accurately estimating gross primary productivity (GPP) is essential for understanding and managing carbon dynamics within an ecosystem. This study investigates the potential of imagery approaches, specifically utilizing Landsat‑8 and Sentinel‑2 data, to model GPP in tea ecosystem within the subtropical region of China. While extensive research has focused on cereal crop ecosystems, tea plantations, despite their global significance as a cash crop, have received limited attention regarding GPP modeling. To address this gap, a field campaign was carried out using the eddy covariance (EC) system to monitor net ecosystem exchange (NEE) within tea plantations at an ecosystem scale. Pruning is recognized as a crucial management practice in the growth of tea plants, leading to significant variations in NEE and its components (ecosystem respiration (RES)). Consequently, we selected the pruning period, from February to June for modeling GPP. Traditionally, vegetation photosynthesis models (VPMs) based on imagery data have required extensive parameterization, posing challenges for data-limited scenarios. In this study, we developed a parametric model based on vegetation indices such as normalized difference vegetation index (NDVI) and scaled photochemical reflectance index (sPRI), which describe both plant structure and physiology using EC and Landsat-8/Sentinel‑2 imagery data. Results indicate that while NDVI partially captures GPP variation using Landsat‑8 (R2 = 0.60) and Sentinel‑2 (R2 = 0.71) imagery, incorporating sPRI significantly enhances the agreement between modeled and observed GPP (Landsat-8 : R2 = 0.77, Sentinel-2 : R2 = 0.80). Furthermore, comparing GPP estimates derived from EC (GPPEC) with those from Sentinel (GPPSentinel) and Landsat (GPPLandsat) imagery reveals that GPPSentinel closely aligns with GPPEC (R2 = 0.80), outperforming GPPLandsat based on various evaluation indices (index of Agreement, Kling-Gupta efficiency, mean bias error, relative bias in percent).