Mangrove Ecosystem Health Assessment In Beibu Gulf,Guangxi Based On Multi-Source Remote Sensing

Mangroves are one of the most productive and valuable wetland ecosystems in the world.The health assessment of wetland ecosystems is a necessary prerequisite and foundation for the conservation and restoration of wetland resources,and the health assessment of mangrove ecosystems is of great significance for the restoration of degraded ecosystems.The magnitude of the physicochemical parameter of canopy chlorophyll content（CCC）and the structural parameter of canopy height（h）can not only reflect the productivity of mangroves but also judge their health condition.Current studies on the health assessment of mangroves tend to ignore information on physicochemical and structural parameters.There is often a lack of suitable data sources and inversion models for estimating mangroves’physicochemical and structural parameters.Therefore,in this paper,taking the Beibu Gulf region as the main study area and the mangrove area near Suwu and Shajiao villages as the typical study area,we selected multi-source optical and polarization data from Zhuhai-1Hyperspectral Satellite（OHS）,Sentinel-2A,Gaofen-3（GF-3）satellite and Sentinel-1A satellite.We compared the optical data with the joint radiative transfer model,respectively The accuracy of CCC estimation by stacked machine learning regression algorithm was reached,and the applicability of polarization data fused with optical images to invert the canopy height was investigated.A pressure-state-response（P-S-R）model-based mangrove ecosystem health evaluation index system was constructed to assess the health status of mangroves in Beibu Gulf,Guangxi.The results of the study are as follows:（1）Characteristic variables were preferred by correlation analysis and variable importance assessment.After correlation analysis and variable importance assessment,RSI_（12,17）,DSI_（12,18）,and NDSI_（6,12）combined vegetation indices had the highest sensitivity to CCC in mangroves;b17＿VV,b16＿VV,and b6＿VV were the critical parameters for canopy height estimation with the highest contribution of inversion in the study of inversion of canopy height from fused data of GF-3 and OHS.（2）Optimal inverse models are constructed by machine learning combined with multi-source data.By combining OHS data with the Gradient Boosting Regression Tree（GBRT）integrated learning regression model,the mangrove CCC can be inverted with high accuracy,and its precision（training accuracy:RMSE=0.963μg/cm²）is better than that of Random Forest（RF）model and linear regression model.In addition,the fusion of GF-3 and OHS data combined with the Extreme Gradient Boosting（XGBoost）machine learning regression algorithm can obtain more accurate inverse canopy height fitting results,resulting in better estimation with an accuracy of R²of 0.710 and RMSE of 0.243m.（3）Assessment of mangrove ecosystem health by P-S-R model.After quantitative inversion and remote sensing classification techniques,this paper determined 12 categories of indicator parameters and used hierarchical analysis（APH）to determine the weight of each indicator,and finally constructed the judgment matrix of each indicator in the project layer,factor layer,and indicator layer,where the CR of each layer was less than 0.1,meeting the requirements of consistency test.The CHI,PHI,SHI,and RHI health indices were calculated and analyzed to find that the mangrove ecosystem in Guangxi Beibu Gulf was in a healthy state from 2019 to 2022,and a healthier state in 2023,down one level compared to other years.