基于GF-1 WFV数据和随机森林算法的太湖水生植被时空变化分析OA

To explore a classification method for aquatic vegetation applicable to domestic Gaofen data,this study systematically evaluates the spectral,index,and texture features of GF-1 WFV data for discriminating aquatic vegetation types.A random forest-based method for aquatic vegetation classification is constructed,achieving high-accuracy classification of submerged vegetation,floating-leaved/emergent vegetation,and algal blooms.The spatiotemporal evolution characteristics of aquatic vegetation in Taihu Lake over the past decade(2016-2025)are analyzed.The results show that the feature-selected random forest method using visible and near-infrared data achieves classification accuracy comparable to that of the VBI(vegetation and bloom indices)algorithm based on shortwave infrared.The overall accuracy is 92.33%and the Kappa coefficient is 0.897 8.Independent validation indicates an overall classification accuracy of 89.23%and a Kappa coefficient of 0.855 0,demonstrating stable and reliable performance.Over the past decade,the spatiotemporal variation patterns of algal blooms,submerged aquatic vegetation,and floating/emergent aquatic vegetation in Taihu Lake remain relatively stable,each exhibiting distinct characteristics:Algal blooms exhibit significant seasonal spatial migration patterns.Over the past decade,the maximum area of algal blooms fluctuates,while the past five years show a downward trend.Submerged aquatic vegetation and floating-leaved/emergent vegetation are characterized by"expanding in place",while submerged aquatic vegetation exhibits significant ecological vulnerability.