Study On Estimation Of Vegetation Gross Primary Productivity Based On CNN Deep Learning Model In China

Terrestrial ecosystems are the basis for human survival and sustainable development,and in the context of increasing global climate change and human activities,carbon cycling processes are greatly affected.Gross Primary Productivity(GPP)is an important factor in estimating the Earth's supporting capacity and evaluating the sustainability of terrestrial ecosystems,and plays an important role in global change and the carbon cycle,making accurate assessment of vegetation GPP particularly important.In order to enrich the estimation tools of GPP,this paper takes China as the study area and uses deep learning algorithms of Convolutional Neural Networks(CNN)based on FLUX observation data and remote sensing data to estimate the vegetation GPP for 2001-2017.The spatial and temporal patterns of GPP were analyzed,followed by the relationship between GPP and temperature,precipitation,solar radiation and CO₂using the partial derivative method and correlation coefficient.The results of the study show that the deep learning algorithm is an effective method for estimating vegetation GPP;the vegetation GPP in China showed an increasing trend from 2001 to 2017.The main conclusions drawn from this paper are as follows.(1)The monthly GPP simulated values were significantly correlated with FLUX GPP,with R²of 0.84(P<0.01)and RMSE of 49.02 g C·m^-2·a^-1;the annual GPP was also significantly correlated with FLUX's annual GPP,with R²of 0.86(P<0.01),which was better than other GPP products with FLUX GPP.The annual GPP correlated well with other data products at the spatial scale,and the total GPP was consistent with the results of previous studies;the GPP simulated in this paper reflected the different vegetation cover types and the boundaries were also obvious,with strong spatial heterogeneity,so the accuracy of the Chinese vegetation GPP results simulated by the CNN deep learning algorithm was high.(2)There are obvious differences in monthly GPP changes across the country during the study period.The national average and spatial variation of vegetation GPP values varies greatly between seasons,with the highest GPP values in summer and the lowest GPP values in winter.The national multi-year average GPP value was 707.36g C·m^-2·a^-1,with a total of 6.41 Pg C·a^-1,showing a decreasing distribution characteristic from southeast to northwest.The mean value of the vegetation GPP trend was 5.96g C·m^-2·a^-1,with an overall increasing trend,indicating that the vegetation grew well nationwide during 2001-2017.(3)There are differences between the multi-year average GPP of different vegetation types across the country,with forestland>cropland>grassland>desert,so forests have a pivotal position in the national carbon sink.There are obvious spatial distribution characteristics in different physical geographic regions,with the highest GPP value in the eastern monsoon region and the lowest in the arid and semi-arid northwest region.(4)The annual average GPP and different seasonal GPP of the national vegetation show obvious vertical zonal distribution characteristics within different altitude zones.There are three decreasing and two increasing zones of GPP variation over the whole altitude.The annual average GPP of vegetation shows a clear"double peak and double valley"distribution with the change of slope direction,with the maximum value of GPP occurring on the southeast and west slopes,the maximum value on the southeast slope,the minimum value of GPP on the southwest and north slopes,and the minimum value on the north slope.(5)The contribution of temperature,precipitation and solar radiation to vegetation GPP change is also different.The area with the largest positive contribution of precipitation to GPP is 53.59%.The area with positive contribution of temperature,precipitation and solar radiation is higher than the area with negative contribution.Both precipitation and solar radiation promote the growth of GPP;from the perspective of different vegetation types,the contribution of different climatic factors to different types of GPP is consistent,but the contribution of precipitation to desert GPP is more significant.(6)Increases in temperature,precipitation,solar radiation and CO₂concentration all contributed to vegetation growth to varying degrees,with increased CO₂concentration increasing GPP by the largest area of 69.19%,followed by precipitation,temperature and solar radiation.Overall,CO₂was the main driver of vegetation GPP change.Among the climatic factors,increased precipitation increased GPP by the largest area,and precipitation was the main climatic driver of vegetation GPP change in China,followed by precipitation as the main driver of GPP change in the eastern monsoon and northwest arid and semi-arid regions,and solar radiation as the main driver of GPP change in the Qinghai-Tibet alpine region.