地球信息科学学报 ›› 2020, Vol. 22 ›› Issue (10): 2023-2037.doi: 10.12082/dqxxkx.2020.200078

• 专栏:城乡生态环境综合监测 • 上一篇    下一篇

融合土地覆盖和土壤水分产品的近地表空气温度空间化方法

高亮1,2(), 杜鑫1, 李强子1,*(), 王红岩1, 张源1, 王思远1,2   

  1. 1.中国科学院空天信息创新研究院,北京 100101
    2.中国科学院大学,北京 100049
  • 收稿日期:2020-02-17 修回日期:2020-05-12 出版日期:2020-10-25 发布日期:2020-12-25
  • 通讯作者: 李强子 E-mail:gaoliang@aircas.ac.cn;liqz@aircas.ac.cn
  • 作者简介:高亮(1994— ),男,甘肃定西人,硕士生,主要从事农业遥感和农业气象灾害方面的研究。E-mail:gaoliang@aircas.ac.cn
  • 基金资助:
    国家重点研发计划资助项目(2017YFD0300404-1);国家重点研发计划资助项目(2017YFD0300402)

A Near-surface Air Temperature Spatialization Method Integrating Landuse and Soil Moisture Products

GAO Liang1,2(), DU Xin1, LI Qiangzi1,*(), WANG Hongyan1, ZHANG Yuan1, WANG Siyuan1,2   

  1. 1. Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2020-02-17 Revised:2020-05-12 Online:2020-10-25 Published:2020-12-25
  • Contact: LI Qiangzi E-mail:gaoliang@aircas.ac.cn;liqz@aircas.ac.cn
  • Supported by:
    National Key Research and Development Program of China(2017YFD0300404-1);National Key Research and Development Program of China(2017YFD0300402)

摘要:

空气温度是评价人居环境的重要指标,与人类的生产生活息息相关;其观测对于水文、环境、生态和气候变化等方面的研究具有重要意义。传统的大范围空气温度观测数据一般通过气象站点获取,但由于气象观测站点空间分布离散稀疏的特点,所获取的数据不能精确描述空间连续的空气温度变化情况。因此,实现基于遥感数据的近地表空气温度精准估算具有重要的现实意义。本研究基于精细的地表覆盖类型、空间连续的土壤水分、地表温度(LST)数据,并结合其他辅助数据,构建了近地表空气温度空间化模型,并对近地表空气温度影响因子进行评估,发现地表覆盖类型对近地表空气温度的影响最大,土壤水分为最活跃的影响因素,经验证,模型精度较高,R2接近0.85,RMSE为0.5℃。本研究获取的精确空间连续的近地表空气温度信息,能够充分表达其空间异质性,为农业气象灾害灾变过程监测、农作物生长过程模拟、区域气候变化分析等研究提供良好的近地表空气温度数据支撑。

关键词: 近地表空气温度, 地表温度, 空气温度影响要素, 机器学习, 土地覆盖, 土壤水分, 空间化, 变量重要性分析

Abstract:

Air temperature is an important attribute for evaluating the living environment, and its studies and observations are closely related to human production and life. Air temperature observation data is of great significance for the study of hydrology, environment, ecology, and climate change. Traditional description of large-scale air temperature is generally obtained through meteorological stations. As affected by land surface condition and atmospheric state, the air temperature is spatially heterogeneous. However, due to the sparse spatial distribution of meteorological station sites, the data obtained from these meteorological stations cannot accurately describe the continuous spatial variation of air temperature across large areas. Hence, accurate inversion of near-surface air temperature based on remote sensing data is regarded as an effective and reasonably practicable solution. There are already some studies about obtaining spatially continuous near-surface air temperature using land surface temperature and other remote sensing data. In this study, we have used the remote sensing data, specifically the precise surface coverage type and spatially continuous soil moisture data, as the new input to improve the accuracy of temperature inversion. On this basis, we built a near-surface air temperature spatialization model using Land Surface Temperature (LST), land cover, soil moisture, land surface temperature, NDVI, DEM, aspect, and slope as the influencing factors. In order to fit the complex relationship between air temperature and its influencing factors, we chose four widely used machine learning algorithms and compared their accuracy to select the most reasonable model. At the same time, we also validated the results and evaluated the contribution of the influencing factors. Based on the results of the designed experiments, we found that precise surface cover type and spatially continuous soil moisture data played the most important role in near-surface air temperature spatialization model. The surface cover type has the greatest influence on the near-surface air temperature, and soil moisture is the most active influencing factor. The model validation results showed that the spatialization model has a relatively high accuracy, with an R2value close to 0.85, and a RMSE of 0.5℃. Comparing with traditional methods, the results of near-surface air temperature spatialization model in our study could express more refined spatial distribution pattern. The high precision near-surface air temperature inversion model proposed by our research is expected to provide effective data support to the study on the dynamic monitoring of agricultural meteorological disasters, simulation of crop growth processes, and analysis of regional climate change.

Key words: Near-surface air temperature, Land surface temperature, Factors affecting air temperature, Machine learning, Land cover, soil moisture, spatialization, importance analysis