Reconstruction And Driving Factors Analysis Of MODIS Time Series Land Surface Temperature

Land surface temperature(LST)is an important characteristic physical quantity that characterizes the surface water-heat condition and energy balance process,and it plays an important role in the application research of crop evapotranspiration and growth monitoring,water cycle,climate change and so on.Due to the sparse distribution of ground observation stations,thermal infrared remote sensing technology has become an important means to quickly obtain ground temperature in a large range.However,the surface temperature acquired by thermal infrared is susceptible to clouds,and the lack of values in the image caused by cloud cover is an important factor restricting the application and development of thermal infrared surface temperature data.Previous studies have made it difficult to reconstruct the true surface temperature of the ground covered by clouds.In this paper,based on the strong representation of the ground station observation data on the surface temperature under the cloud cover,an effective reconstruction model of the fusion station observation data and the cloudless part of the thermal infrared surface temperature is established.Therefore,the high-precision MODIS surface temperature data set from 2003 to 2018 was obtained.Based on the reconstructed surface temperature data,the spatial-temporal variation pattern of surface temperature in the past 16 years was studied.At the same time,considering the complex mechanism of surface temperature changes,based on the time series data of five remote sensing surface temperature driving factors and two ocean index data,the correlation and driving effects of various driving factors and surface temperature changes were discussed.The main work and conclusions are as follows:(1)Based on the high accuracy of the site observation data and not being affected by the cloud,it can reflect the advantage of the true surface temperature under the cloud cover.An effective fusion model between the site observation surface temperature and the available surface temperature in clear sky was established Year-time continuous spatial temperature data set.This data set effectively reconstructs the true surface temperature value that can represent the cloud cover,and overcomes the limitation that the surface temperature can only be reconstructed under the assumption of cloudless conditions in the past.Using independent site observation data for accuracy verification,the results show that the reconstruction results have higher accuracy,with an average RMSE of 1.42 ?,MAE of 1.32?,and R2 of 0.97.The data set has been released and can be downloaded by users worldwide.(2)Using the reconstructed data,we explored the spatio-temporal change pattern of surface temperature in the past 16 years,and captured many areas with significant linear changes.According to the pixel-by-pixel linear trend analysis results,62.5%of the country's surface temperature is on the rise,of which 18.1%are significantly higher(P<0.05),mainly distributed in the central and western regions of the Inner Mongolia Plateau,the southern region of Tibet,and the Huanghuai Sea Near the plain,and the temperature rise is large,the slope of the change is slope>0.075.The areas with a significant cooling trend are mainly concentrated in the vicinity of the Songnen Plain and parts of southern China,but the cooling range is not large(-0.075<change slope<0.05).(3)The systematic study of the correlation between the surface temperature and the five driving factors and the response of the surface temperature to the El Nino phenomenon.The study found that among the five driving factors,NDVI has the highest correlation with the surface temperature in the study area,with a generally negative correlation,followed by cloudiness and atmospheric water vapor content,while aerosols and soil moisture have little effect on the surface temperature.At the same time,the correlation study between the surface temperature and ENSO found that the surface temperature is more significantly affected by the warming event El Nino event.The probability of the surface temperature anomaly occurring in the corresponding month reaches 76.74%,mainly reflected in the autumn and winter seasons.The correlation between LST anomalies in January and ENSO is the best.