综合典型地表组成的城市地表温度估算  

Estimating urban surface temperature based on the typical land surface compositions

作　　者：李斌侠 臧淑英[1] 吴长山[1] 李苗[1] 占红[1] 解瑞峰[1] 

机构地区：[1]哈尔滨师范大学黑龙江省普通高等学校地理环境遥感监测重点实验室,哈尔滨150025

出　　处：《测绘科学》2016年第2期87-91,共5页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41171322);黑龙江省自然基金重点项目(ZD201308);黑龙江省自然科学基金面上项目(D201407)

摘　　要：针对单一的地表物质组成并不能充分反映城市地表热环境特点这一问题,该文基于热混合影像,利用线性光谱分解方法获取地表组成信息,然后利用光谱分解热混合、线性回归、决策树方法估算地表温度。结果表明:只研究单一地表组成对地表温度的影响,有可能扩大其环境效应;决策树模型在不同规则下能更好地模拟地表温度的空间异质性;光谱分解热混合模型只需要两组数据即可估算出不同地表覆盖下的地表温度,且估算精度较其他模型高;光谱分解热混合模型和多元回归模型结合4种地表组成监测其对地表温度的影响,决策树方法通过不透水面、水体、植被预测地表温度,前两者估算精度比后者高,因此综合考虑城市典型地表组成能更好反映其对地表温度的作用。This paper used a spectral unmixing method to estimate the fractional land covers from TM imagery,then the spectral unmixing thermal mixing,linear regression and decision tree models were developed to estimate LST.The results indicated that with a single land surface composition,its impact on environment might be over-estimated;the impact of land surface compositions on LST were nonlinear and LST＇s distributions were of spatial heterogeneity;with just a pair of data,SUTM approach accurately reflected the interplay between LST and various land biophysical compositions,and comparing to the other models,SUTM performed well with low RMSE and MAE;SUTM and multiple regression models could examine the impacts of four urban biophysical compositions on LST,but the decision tree model was with only three land compositions;and as a result,the former models＇ estimation accuracy was better,which proved the advantage of accommodating all typical land surface compositions.

关 键 词：地表组成 地表温度 光谱分解 热混合 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]