基于GF-1遥感数据的大兴安岭林区雪水当量反演  

Comparison between GF-1 data and Landsat-8 data in inversing SWE of the forest in the northern of Daxing'an Mountain

作　　者：俞正祥[1] 韩广忠[2] 蔡体久[1] 琚存勇[1] 朱宾宾 YU Zhengxiang HAN Guangzhong CAI Tijiu JU Cunyong ZHU Binbin(School of Forestry, Northeast Forestry University, Harbin 150040, China Songling Forestry Bureau of Daxing＇an Mountain, Songling 165012, China)

机构地区：[1]东北林业大学林学院,黑龙江哈尔滨150040 [2]大兴安岭松岭林业局,黑龙江松岭165012

出　　处：《南京林业大学学报（自然科学版）》2017年第3期105-111,共7页Journal of Nanjing Forestry University：Natural Sciences Edition

基　　金：国家自然科学基金项目(31370460);东北林业大学学术名师支持计划(PFT-1213-21)

摘　　要：【目的】以Landsat-8卫星数据为对照,探索国产GF-1卫星数据对林区积雪特征的估测能力,实现基于国产卫星对东北林区雪水水文过程的监测。【方法】以大兴安岭北部林区为研究区,结合同期的地面雪水当量野外观测数据,利用偏最小二乘回归与BP神经网络两种方法建立线性与非线性雪水当量反演模型。通过平均均方根误差(E_(RMSE))、平均相对均方根误差(r_(RMSE))和平均估测精度这3个评价指标对所建模型进行对比评价。同时,利用两种遥感数据建立的最优模型对研究区内雪水当量分布特征进行反演,并对反演结果进行对比分析。【结果】基于GF-1数据所建立的线性与非线性模型性能均略低于以Landsat-8数据构建的模型,其中GF-1数据最优反演模型精度为80.3%,较Landsat-8反演模型低1.6%;基于GF-1数据反演的雪水当量值与Landsat-8的基本相同;两类遥感数据反演得到的雪水当量在空间分布特征上基本一致,均反映了雪水当量与地形、植被及土地利用类型的高度相关性;由于山地林区植被和地形复杂,并且春季升温过程中地面积雪日消融速率快,GF-1数据以其高空间与时间分辨率上的优势能够更好地对研究区雪水水文过程进行监测。【结论】国产GF-1卫星能够替代Landsat-8卫星成为对大兴安岭北部林区雪水水文过程监测的遥感数据源。[Objective] With a large number of independently developed satellites launched, China has made tremendous advancements in remote sensing. The successful launch of the GF-1 satellite indicates that the observation of earth has entered the ＂high-resolution epoch＂ in China. This study was conducted to monitor the snow hydrological process of forests based on data from this Chinese satellite. [ Methods ] The accuracy of forest snow characteristics based on the Chinese GF-1 satellite data was estimated by comparison with Landsat-8 satellite data. The research site was the forest of the northern Daxing＇an Mountains. Based on the two methods of PLS （partial least squares） and BP （back propagation） neural networks, the linear and nonlinear SWE （ snow water equivalent） inversion models were built using synchronous ground SWE field observation data. The models were evaluated through the indicators of L＇ RMSE, rRMSE and average estimation accuracy. The best models based on the two types of remote sensing data were used to determine the inverse of the distribution characteristics of the SWE in the study area. [ Results ] The results showed that the accuracy of the models established using GF-1 data was slightly lower than that of those established using Landsat-8 data. The highest accuracy of a GF-1 model was 80.3%, which was 1.6% lower than that of the equivalent Landsat-8 model. The inverse SWE values based on the data of GM-1 were little higher than that of Landsat-8. The inversed SWE from the two types of remote sensing data showed the same spatial distributions, which showed that the SWE is highly correlated with topography, vegetation and land use type. Both terrain and vegetation are complicated in mountain forests. Snow ablation is relatively fast in spring. Owing to a higher spatial and time resolution than that of Landsat-8, the GF-1 satellite is more efficient for monitoring snow hydrological processes. [ Conclusion] Our results indicated that data from the GF-1 satellite can replace Landsat-8

关 键 词：雪水当量 遥感数据 GF-1 Landsat-8 大兴安岭北部 

分 类 号：S715-3[农业科学—林学]