Estimation And Validation Of Global Physically True Daily Mean Land Surface Temperature

Physically true daily mean land surface temperature(termed as Tdm)was widely used in various fields,such as surface thermal property detection,evapotranspiration mapping and global climate change analysis.Satellite thermal remote sensing is the major way to obtain LSTs regularly over an extensive region.However,limited by the trade-off between spatial and temporal resolution and the cloud contamination,current satellite thermal remote sensing products cannot provide temporally continuous LST products to calculate Tdm.Facing this challenge,traditional studies usually used temporally aggregated cloud-free LSTs for compromise,which on the one hand,causes systematic sampling bias compared with physically true Tdm,on the other hand,causes deviations in the trending analysis.In order to estimate the physically true Tdm,we designed a framework(termed as ADTC-based framework)by combining annual temperature cycle(ATC)model which is responsible for reconstructing under-cloud instantaneous LSTs,and diurnal temperature cycle(DTC)model which is responsible for providing temporally continuous LSTs.We describe the major scientific contributions in two aspects.First,the estimation of global Tdm and the trending analysis of LST.Based on the ADTC-based framework,we designed the acceleration algorithm to solve the problem that the running speed is too slow when estimating global Tdm with the ADTC-based framework.Consequently,global 1-km Tdm from 2010 to 2017 were fast estimated.Results show that traditional method to calculate Tdm(averaging all discrete cloudfree LSTs)has positive systematic sampling bias(around 2.0 K)over global scale,but negative sampling bias(around-1.0 K)in high latitude regions.The trending analysis of LST shows that both the results from traditional method to calculate Tdm and results from our estimated Tdm display the warming trend of global LST(about +0.07 K/year),but when comparing the trend in local scale,the MAE(mean absolute error)of the trend differences is around 0.056 K/year.Moreover,the trend differences between these two results are unevenly distributed in global filed.Second,the comprehensive validation of the ADTC-based framework.We validated the ADTC-based framework under cloud-free and overcast conditions with geostationary satellite data and in-situ measurements,respectively.In addition,we analyzed the separate contributions of ATC and DTC models as well as their uncertainties.Under cloud-free condition,the MAE of ADTC-based framework is around 0.5 K.Under overcast condition,the MAE of ADTC-based framework is around 1.0 K at daily scale and 0.5 K at monthly scale,respectively.Under both conditions,the ADTC-based framework is able to correct the systematic sampling bias(larger than 2.0 K).Contribution analysis shows that taking the MAEs of traditional method to calculate Tdm at daily scale as the baseline,the incorporation of ATC model decreases the MAEs from 4.2 to 2.0 K;The incorporation of DTC model further decreases the MAEs from 2.0 to 1.0 K.Uncertainty analysis reveals that although ATC model has systematic warm-biased at around 2.6 K during the daytime and cold-biased at around 1.1 K during the nighttime,DTC model is robust to the reconstruction error of ATC model while averaging hourly LSTs to calculate Tdm.The proposed ADTC-based framework provides the technique support for obtaining large-scale Tdm.The generated global Tdm products based on the ADTCbased framework provides useful data for climate change studies.In the future,through improving ATC and DTC models,the current ADTC-based framework is expected to extend to the estimation of hourly LSTs so that we can provide more various and valuable data for geoscience research.