Long-term topographic effect on remotely sensed vegetation index-based gross primary productivity (GPP) estimation at the watershed scale

Abstract

• Topographic effect on GPP estimation was analyzed during a 21-year period. • MTG is more sensitive to the topographic variations than TG. • MTG could well capture the seasonal variations of radiation and temperature. • It is useful to consider the long-term topographic effect on GPP estimation. Accurate monitoring of gross primary productivity (GPP) is crucial to understanding the feedback to global climate change. Due to the availability of remotely sensed vegetation index (VI) since the 1970s, VI-based models have been an effective tool to obtain spatial-continuous GPP estimates at a large scale. However, most VI-based models neglect the topographic effect on the vegetation photosynthesis and are more suitable for GPP estimation in flat areas. Moreover, the limited existing studies only focused on validating the mountain VI-based model at the site and watershed scales, and the long-term topographic effect on VI-based GPP estimation at the watershed scale is not yet known. In this work, a flat VI-based model (Temperature and Greenness, TG) and a mountain VI-based model (Mountain TG, MTG) were adopted to obtain time-series GPP estimates over a mountainous watershed during 2001–2021. Then, the long-term topographic effect on VI-based GPP estimation was summarized by investigating the spatial and temporal differences between TG and MTG -simulated GPP, as well as their sensitivities to climatic factors. Results illustrated that the MTG-simulated GPP presented a higher spatial variation than those from TG, with the coefficient of variation values of 64% and 45%, respectively. The high similarity between MTG-simulated GPP and DSR and LST were observed in seasonal variation during the growing season. During the 21-year period, MTG-simulated GPP presented a higher sensitivity to the time-series radiation than the TG-simulated GPP (a higher correlation coefficient value of 0.10), whereas it showed a lower sensitivity to the time-series temperature than the TG-simulated GPP (a lower correlation coefficient value of 0.04). The sensitivities of MTG-simulated GPP to the time-series DSR and LST significantly changed with the variations of topographic attributes (i.e., elevation, slope, aspect). These results revealed that the long-term topographic effect could regulate radiation and temperature and then influence the seasonal variation of vegetation photosynthesis over a mountainous watershed. Our study suggests the necessity of coupling the info of surface topography in estimating time-series GPP and highlights that more attention should be given to the long-term topographic effect on remotely sensed VI-based GPP estimation over mountainous watershed.