All-sky Near-surface Air Temperature Estimation And Validation Based On Multi-source Data

Near-surface air temperature,as an important climate variable,has been used in a wide range of fields such as ecology,hydrology,climatology,epidemiology,and environmental science.Understanding the spatial and temporal distribution pattern of global air temperature is conducive to a more comprehensive understanding of the earth’s climate.However,ground measurements are limited by poor spatial representation and inconsistency,while reanalysis and meteorological forcing datasets suffer from coarse spatial resolution and inaccuracy.The existing air temperature products usually do not have high spatial and temporal resolutions,spatial and temporal continuity and large scale coverage at the same time,so it is difficult to meet the needs of some scientific research and other applications for such data.In view of the current problems of air temperature estimation and products,a complete framework for medium-high spatial resolution all-sky air temperature estimation based on multi-source data was constructed.This paper estimated daily mean air temperature over Chinese mainland using moderate-Resolution Imaging Spectroradiometer（MODIS）data and Global Land Data Assimilation System（GLDAS）data,and estimated daily scale air temperature using Global LAnd Surface Satellite（GLASS）data.The main research conclusions are as follows:（1）In this paper,a framework for estimating the all-sky daily mean temperature based on MODIS and GLDAS data was established.MODIS land surface temperature（LST）was firstly gap-filled,and then combined with GLDAS air temperature and other data to establish random forests models under three weather conditions for estimating daily mean temperature.The daily mean air temperature product over Chinese mainland was developed（2003-2019）.The validation results showed that the accuracy of the method met expectations and the model performance was acceptable under all conditions.The R~2 range of the three models was 0.984-0.986,and the root mean square error（RMSE）range was 1.342-1.440 K.The total R~2,RMSE and bias of the estimated all-sky air temperature were 0.985 K,1.409 K and 0.030 K,respectively.The model performed well at most stations with an average RMSE of 1.383 K,97%of which was less than 2 K.All the data required by this method are easy to obtain,and the method is simple and feasible,which can be transplanted to the research of air temperature estimation in other regions.The spatial and temporal patterns of the generated product was similar to the reanalysis data,with higher accuracy and spatial resolution.（2）This paper mainly used a series of data products of GLASS to establish machine learning models to estimate the all-sky daily air temperature,and the 1 km daily maximum,mean and minimum air temperature products of the global land surface（2000-2020）were developed.The validation results showed that the total R~2,RMSE and bias of the estimated all-sky daily maximum air temperature were 0.971,2.347 K and-0.135 K,respectively.The total R~2,RMSE and bias of the daily mean air temperature were 0.984,1.670 K and-0.056 K,respectively.The total R~2,RMSE and bias of daily minimum air temperature were 0.967,2.339 K and 0.036 K,respectively.The RMSE of the maximum,mean and minimum daily air temperature products in2006 were 2.310 K,1.681 K and 2.360 K,respectively.The developed air temperature products have the characteristics of high accuracy,moderate resolution and global land surface covering,and has advantage on the spatial resolution and accuracy compared with the same type product.And the products are suitable for long time series analysis and application research,and can be used as a data support product for research in many scientific fields such as climate change,water cycle and energy balance.