ERA5 Reanalysis Of Surface Temperature Downscaling Under Complex Terrain And Underlying Surface Conditions

Land Surface Temperature(LST)is one of the most important environmental parameters to describe the interaction between land surface and atmosphere,energy exchange and water cycle process in regional and global regions.It is used in urban heat island effect research,surface energy flux estimation,soil moisture estimation It plays an irreplaceable role in many fields.The surface temperature retrieved in the thermal infrared band often restricts each other in temporal and spatial resolution,and the validity of the data is closely related to the weather conditions at the time of imaging,which makes the temporal resolution of the actually available data lower,limiting its full application.Compared with remote sensing monitoring data,reanalysis data can provide continuous spatiotemporal surface temperature data.For example,the ERA5 reanalysis surface temperature product of the European Centre for Medium-Range Weather Forecasts provides the hourly surface temperature of the global land,but the spatial resolution is only 0.1°.In order to meet the demand for high temporal and spatial resolution surface temperature in related fields,spatial downscaling of high temporal resolution surface temperature data to improve the spatial resolution of surface temperature is an effective way to solve the above situation.(1)In order to provide an accurate reference surface temperature for downscaling work,consider topographic factors including elevation,slope,slope aspect and sky openness,establish a thermal infrared radiation transfer model suitable for complex terrain,and develop complex terrain in the study area.The thermal infrared remote sensing inversion of surface temperature under the condition of the underlying surface is used to directly verify the measured surface temperature of the site,and the indirect verification is performed using the near-surface air temperature data and the MOD11A1 surface temperature product.The results show that the inversion results of the surface temperature after considering the topographic effect are numerically reduced.The difference between the inversion results before and after considering the terrain effect is negatively correlated with the sky openness.In the two study areas,because the three-dimensional spatial distribution information of urban buildings is not considered,the suburban study area is relatively flat in terrain,the overall sky is more open,and the overall terrain fluctuation of the mountainous study area is more obvious.The difference values of temperature inversion methods are more reflected in the mountainous study area.(2)Using the Landsat 8 inversion surface temperature as the reference surface temperature,the ERA5 surface temperature correction linear model is established.The results show that the ERA5 reanalysis surface temperature is in good agreement with the spatial distribution of Landsat 8 surface temperature,and it can accurately express the relative level of surface temperature in the spatial distribution.product potential.For the low-resolution ERA5 surface temperature data,classical downscaling methods such as Dis Trad and multiple regression are used to downscale the surface temperature under the complex terrain and underlying surface conditions under study,because the explanatory factor is single or the fitting process is simple,the desired result cannot be obtained.In the process of downscaling using the random forest model,the importance of each surface parameter in the random forest model does not change much with the spatial resolution,which better reflects the "scale invariance" feature of the downscaling method.The importance of explanatory factors for different seasons reflects certain seasonal characteristics.In spring and summer,NDVI played a larger role in the downscaling process in these two seasons.In autumn and winter,two indices,NDBSI and NDBI,which can reflect surface dryness,play a greater role.For different regions,the importance of altitude is more obvious in mountainous regions with more undulating terrain.By adding a step-by-step downscaling link to the random forest downscaling method,the high and low values of the downscaling results are better restored,the error of the downscaling results is reduced,and the correlation with the reference surface temperature is higher.During the step-by-step downscaling process,the root mean square error(RMSE)of the ERA5 surface temperature downscaling results gradually increases as the downscaling spatial resolution gradually increases.There is an underestimation phenomenon in the high-value area of the surface temperature,and an overestimation phenomenon in the low-temperature area,and the downscaling error of the surface temperature in the mountainous area is larger.