Estimation Of Gross Primary Productivity Of The Salt Marsh Wetlands In The Yangtze Estuary By Integrating Eddy Covariance And Satellite Remote Sensing

Gross primary productivity（GPP）is one of the important parameters to measure ecosystem function.Accurate estimating of the GPP in coastal salt marsh wetlands is crucial for understanding carbon budget and carbon sink capacity of coastal wetlands.However,due to the characteristics of small area,fragmentation and tidal inundation effects,the GPP estimation of coastal wetlands at the regional scale may contain large uncertainties.There is currently a lack of applicable high-resolution GPP estimation methods for coastal salt marsh wetlands.In this study,the typical salt marsh wetlands the Jiuduansha wetland and the Chongming Xisha wetland in the Yangtze estuary were used as the study areas.In this paper,we examined two statistical methods,namely,simple linear regression and random forest regression,to carry out the construction and evaluation of the high spatial resolution GPP estimation model,explore the influencing factors and analyze the spatial pattern,in the salt marshes in the Yangtze estuary by integrating multi-source data,including eddy covariance,Landsat,and Moderate-resolution Imaging Spectroradiometer（MODIS）satellite data.Based on satellite remote sensing data sources of different resolutions,the satellite-derived photosynthetically active radiation（PAR）,enhanced vegetation index（EVI）,normalized difference vegetation index（NDVI）,and normalized difference water index（NDWI）were used individually and in different combinations to drive the two statistical methods and investigate their performances in estimating the GPP.The main findings are as follows:（1）This study used the VI×PAR model structure completely driven by remote sensing data to estimate the GPP of a typical salt marsh wetland in the Yangtze estuary.The results showed that together with the PAR,the EVI generally had the greatest potential for GPP estimation(R²>0.75,RMSE<6.80μmol m^-2 s^-1)based on simple linear regression.However,the NDVI outperformed the EVI in wetlands with little tidal flooding and plant litter（R²>0.78）.Compared to linear regression methods,the random forest method had the best performance in showing almost instantaneous GPP changes using a combination of EVI,NDVI,NDWI,and PAR as inputs.Moreover,using Landsat data with a high spatial resolution（30 m）yielded a much more accurate GPP estimation than using MODIS data with a 500 m resolution and Sentinel-2 data with a 10 m resolution.（2）The impact of environmental factors,especially tidal inundation,on the estimation of GPP using VI×PAR estimation forms was discussed,and the results showed that taking the impact of tidal inundation into consideration and removing data with NDWI>0.2,the simple linear regression between VI×PAR and measured GPP generally increased.The R² value of EVI×PAR reached 0.81.Moreover,the random forest method improved the performance in terms of upscaling the GPP measurements to large regions,and the combination of the EVI,NDVI,NDWI,and PAR performed the best,indicating that the use of the machine learning algorithm and the incorporation of a remote sensing index that reflects the tidal influence were beneficial to GPP estimation in coastal salt marshes.（3）In addition,vegetation classification and regional-scale GPP estimation were carried out in this study,taking the typical salt marsh wetland at the Yangtze estuary.The estimation results showed that the vegetation area of the salt marsh wetland showed an overall upward trend from 2005 to 2020,and the overall vegetation area of the salt marsh increased from 121.28 km² in 2005 to 101.07 km² in 2020,with a net loss of 20.21 km².The amount of change in the area of different types of Yangtze estuary salt marsh vegetation varied greatly,and the GPP values of C4 vegetation Spartina alterniflora were significantly higher than others.For example,the GPP estimation based on the simple linear regression model showed that the spatially averaged GPP of Spartina alterniflora in the Yangtze estuary was 20.96μmol m^-2 s^-1on August 16,2020,which was about 49%higher than that of Phragmites australis.Our results demonstrate that the use of high spatial resolution data,the proper use of remote sensing index,and the incorporation of a good combination of these indices through an advanced algorithm such as a machine learning algorithm are vital for capturing the nearly instantaneous GPP variations in coastal wetlands at large spatial scales and obtaining more accurate and comprehensive estimation results.