Elevation Accuracy Evaluation And Correction Of SRTM?ALOS And ASTER GDEM In Shandong Province Based On ICESat/GLAS

SRTM1,ALOS and ASTER GDEM V2 data have high spatial resolution and wide coverage,it is widely used in disaster monitoring,ecological construction,digital city,etc.Since the publication of SRTM1,ALOS and ASTER GDEM V2 data,its accuracy evaluation and correction has been a research hotspot.However,the accuracy verification of large-scale areas lacks effective means.Traditional methods have poor reliability and high data acquisition costs.In 2003,the disclosure of ICESat-1 data provided the possibility to evaluate the accuracy of SRTM1,ALOS and ASTER GDEM V2.Considering the complex terrain of large areas,this paper focuses on SRTM1,ALOS and ASTER GDEM V2 in Shandong Province,with ICESat/GLAS as reference data,mainly doing the following work:(1)Data selection and processing.Download all the ICESat/GLAS observation data in the study area,including 14 cycles and 432 band data;the data is removed and cropped.The data of Shandong SRTM1,ALOS and ASTER GDEM V2 were downloaded and the coordinates were transformed.For the year of GDEM data,download the 100m land use data of Shandong Province(2000).(2)Error analysis.With the help of ICESat,the elevation accuracy of SRTM1,ALOS and ASTER GDEM V2 was evaluated;firstly,the gross error data was removed,and then the error evaluation indicators and error distribution were statistically obtained to obtain the error characteristics of SRTM1,ALOS and ASTER GDEM V2.Combined with the land use data and slope data of Shandong Province,the SRTM1,ALOS and ASTER GDEM V2 errors of Shandong Province were classified and statistically obtained,and the relationship between GDEM error and terrain factor was obtained.(3)Correction and verification of accuracy.Four interpolation methods of inverse distance weighting,ordinary kriging,terrain to grid and natural neighborhood are selected to construct the error surface of SRTM1,ALOS and ASTER GDEM V2,and the error surface is superimposed with the original GDEM.Finally,the SRTM error surface was superimposed with the original SRTM data to obtain the corrected SRTM.(4)Experimental analysis.The errors in the elevations of SRTM1,ALOS and ASTER GDEM V2 in Shandong Province were 3.34 m,3.29 m,and 7.20 m,which were all better than the nominal accuracy,and the overall error was normally distributed.As the slope increases,the elevation accuracy tends to decrease.The analysis of the relationship between land cover types and error distribution shows that:dry land and shrub forest land types have higher accuracy;rivers and lakes have lower accuracy.The results of the four interpolation methods show that the characteristics and accuracy of the error surface constructed by different interpolation methods are also different.Among them,the effect of inverse distance weighted correction is the best,followed by terrain to grid and natural neighborhood,and the ordinary Kriging correction is the worst.Compared with the original GDEM after being corrected by the multiple linear regression analysis method,the error in the error in the ASTER GDEM V2 elevation after the correction is reduced by 2.09 m,and SRTM1 and ALOS are reduced by 1.13 m and 1.12 m,respectively.