秦岭南坡子午河中下游流域土地利用/土地覆被信息提取及其应用

Extraction and application of land use/land cover information in the middle and lower reaches of Ziwu River on the south slope of Qinling Mountains

精确的土地利用/土地覆被数据不仅可以反映区域的生态环境状况,为环境部门提供决策支持,也为实现区域生态环境更高质量发展发挥重要作用。以子午河中下游流域为研究区,利用多源多时相的landsat8卫星遥感数据,结合地面调查数据、文献调研等,探讨并研究支持向量机分类法(SVM)和随机森林模型(RFM)对该区的植被类型和土地利用现状类型进行识别,对两种方法的分类精度进行对比,并分析和评价光谱特征变量对模型的重要性和适用性。利用满足要求的土地利用现状数据,再结合修正的通用土壤流失方程RUSLE模型进一步计算出研究区的土壤侵蚀模数,绘制研究区土壤侵蚀分布图,结合土地利用/植被覆盖信息计算研究区的生态环境状况指数,从宏观上对子午河中下游流域进行生态环境评价。结果表明:①随机森林模型可以有效利用样本的特征因子,并与地形约束因子结合,从而对植被和土地利用类型进行分类,分类总体精度均达到80%以上,kappa系数分别为0.73和0.86,与传统的SVM方法相比,RFM方法均提高了森林类型和土地利用类型的分类精度。②研究区总体生态环境状况指数为87.12,生态环境状况为优,其中水源区附近由于土壤侵蚀流失量相对较大,所以生态环境状况为良,占研究区总面积的15.69%。

Accurate land use/land cover data can not only reflect the regional ecological environment and provide decision support for environmental departments,but also play an important role in achieving higher quality development of regional ecological environment.Taking the middle and lower reaches of Ziwu River Basin as the study area,using multi-source and multi temporal Landsat 8 satellite remote sensing data,combined with ground survey data and literature survey,this study discusses and studies the support vector machine classification(SVM)and random forest model(RFM)to identify the vegetation types and land use status types in this area,and compares the classification accuracy of the two methods.The importance and applicability of spectral characteristic variables to the model are analyzed and evaluated.Based on the data of land use status,combined with the modified general soil loss equation RUSLE model,the soil erosion modulus of the study area was further calculated,the distribution map of soil erosion in the study area was drawn,and the ecological environment index of the study area was calculated according to the land use vegetation cover information,and the ecological environment evaluation of the tourism basin under meridian river insects was carried out from a macro point of view.To the extent that it is not possible to make a difference between the two groups.The results show that:①the random forest model can effectively use the characteristic factors of samples and combine them with topographic constraints to classify vegetation and land use types.The overall classification accuracy is more than 80%,and the kappa coefficients are 0.73 and 0.86 respectively.Compared with the traditional SVM method,RFM method improves the classification accuracy of forest types and land use types.②The overall Eco-environmental Status Index of the study area is 87.12,and the eco-environmental status is excellent.Due to the relatively large amount of soil erosion near the water source area,the eco-environmental status is good,accounting for 15.69%of the total area of the study area.