Remote Sensing Estimation Of GPP Over Mountainous Areas Using The Information Of Neighbor Space And Investigation Of Its Spatial Scaling

Gross Primary Productivity(GPP)is an essential ecological index in assessing terrestrial ecosystems'capacity to absorb atmospheric carbon dioxide.To realize China's 2060 carbon-neutral goal,it is crucial to improve the observation,simulation,and prediction over mountainous areas due to the importance of mountain ecosystems in the global carbon balance.Compared to flat areas,the mechanism of vegetation photosynthesis over mountainous areas is more complicated under the influence of neighbor space.However,the existing remote sensing GPP products almost all neglect the influence of neighbor space on vegetation photosynthesis,and their low spatial resolutions also cannot adapt to the high spatial heterogeneity of mountainous vegetation productivity.Therefore,it is particularly necessary to develop modeling strategies for the remote sensing estimation of GPP over mountainous areas,not always treating the surface as flat terrain.In this context,the main aims of this study are to analyze the latest research progress of terrestrial GPP,develop strategies for remote sensing estimation of mountain GPP,characterize the accumulation rule of spatial scale error in the process of mountain GPP estimation,and explore the mechanism for mountain GPP spatial scaling.Finally,the proposed strategies of GPP estimation over mountainous areas were applied in a typical mountainous area of southwest China(i.e.,Wanglang integrated observation and experiment station for mountain ecological remote sensing).The main research contents and conclusions of this study can be summarized as follows:(1)Uncertainty analysis of GPP estimation from typical vegetation remote sensing models over mountainous areas.This study evaluated three typical vegetation GPP remote sensing models(i.e.,Light Use Efficiency model,vegetation index model,and process-based model)based on the global Eddy Covariance(EC)data,and then comprehensively discussed the uncertainties of GPP estimation from these three typical models over mountainous areas.It was found that the mean absolute error between the simulated GPP and EC GPP over mountainous areas was higher than that over flat areas(i.e.,?115 gC m^-2yr^-1),suggesting that the GPP estimation over mountainous areas is more challenging.Most current vegetation GPP remote sensing models neglect the redistribution effects of neighbor space on radiation,temperature,and water.Besides,remote sensing models intended for estimating regional or global GPP are often executed at coarse resolutions,and the surface heterogeneity within each modelling grid is usually ignored.(2)Remote sensing estimation of GPP over mountainous areas based on neighbor space.To address the above issues existing in current remote sensing GPP estimation over mountainous areas,this study made a step to incorporate the information of neighbor space into current vegetation GPP remote sensing models.By considering the spatial heterogeneity of radiation and temperature in the neighbor space,two representative remote sensing models for GPP estimation over mountainous areas,namely Mountain Two-Leaf LUE(MTL-LUE)model and Mountain Temperature and Greenness(MTG)model,were proposed in this study.Besides,the effectiveness of an eco-hydrological model named Boreal Ecosystem Productivity Simulator(BEPS)-Terrain Lab(combining the effects of neighbor space on water,radiation,and temperature)in simulating GPP over mountainous areas was also discussed.MTL-LUE,MTG,and BEPS-Terrain Lab were validated at the global typical mountainous sites or watersheds.MTL-LUE effectively improved the GPP estimation over mountainous areas(i.e.,root mean square error between simulated GPP and EC GPP was reduced by 0.55 gC m^-2d^-1)through considering the alteration of direct radiation,the reduction of diffuse radiation,and the variation of sunlit canopy area.MTG also effectively improved the GPP estimation over mountainous areas(i.e.,root mean square error between simulated GPP and EC GPP was reduced by 5.43 gC m^-28d^-1).This study highlights that incorporating the redistribution effect of neighbor space on radiation,temperature,and water into current vegetation remote sensing models is a practical approach to improve GPP estimation over mountainous areas.(3)Investigating the mechanism of GPP spatial scaling over mountainous areas.This study analyzed the accumulation rule of the spatial scale error in the process of GPP remote sensing estimation over mountainous areas.Results showed that the spatial scale error of mountainous GPP estimation increased with a decrease in its modeling spatial resolution.To reduce the spatial scale error,this study proposed a spatial scaling index of GPP using the information of neighbor space and vegetation heterogeneity.The effectiveness of the proposed spatial scaling index of GPP was proved at 16 mountainous watersheds across the globe(i.e.,root mean square error was reduced by 169 gC m^-2yr^-1).This study suggests that incorporating the information of neighbor space and vegetation heterogeneity into the spatial scaling index is useful for improving coarse resolution GPP estimates over mountainous areas.(4)Application of mountain vegetation GPP remote sensing models over a typical mountainous area in southwest China.In this study,these three mountain vegetation GPP remote sensing models(i.e.,MTL-LUE,MTG,and BEPS-Terrain Lab)were applied in a typical mountainous area in southwest China-Wanglang integrated observation and experiment station for mountain ecological remote sensing.The limitation of BEPS-Terrain Lab over large areas is that its complicated model structure requires a large number of input data and ecological parameters.Besides,the MTL-LUE model also requires a large number of meteorological data.Although the understanding of the vegetation photosynthesis process in MTG is not as detailed as that in BEPS-Terrain Lab and MTL-LUE,the MTG model takes advantage of extensive remotely sensed data and topography information,and can effectively track mountainous GPP over large areas.Based on the results over Wanglang station,the following strategies were suggested in the process of GPP remote sensing estimation over mountainous areas:1)if relevant ecological parameters and meteorological data are sufficient and reliable,the BEPS-Terrain Lab model is the best choice,2)if only reliable meteorological data are available in the study area,the MTL-LUE model is more suitable than BEPS-Terrain Lab and MTG,3)if there is no reliable ground data in the study area,the MTG model can be used as the simulation tool,and 4)running models at high spatial resolutions would be better for GPP estimation over mountainous areas,otherwise the model output of GPP at coarse spatial resolutions should be corrected by the spatial scaling index.To obtain large-scale GPP estimates over mountainous areas,it is necessary to develop reliable remote sensing GPP models and spatial scaling indices based on neighbor space.The main innovation points and contributions of this study are that:(1)two vegetation GPP remote sensing models based on neighbor space(MTL-LUE and MTG)were established innovatively to improve the GPP estimation over mountainous areas,and(2)the accumulation rule of spatial scale error in the process of GPP estimation over mountainous areas was analyzed systematically,and a spatial scale index of GPP using the information of neighbor space and vegetation heterogeneity was first proposed to reduce the spatial scale error.Future work will apply the above research results to the generation of global high-resolution GPP remote sensing products over mountainous areas,characterize the response of the global mountainous ecosystem to the environment under the scenario of global climate change,in the hope of providing data support and theoretical basis for realizingChina's 2060 carbon-neutral goal.