Study On The Difference Of Spectral And Urban Green Land Biomass Estimation Based On Landsat OLI And ETM+ Images

Landsat series of satellites had been engaged in the continuous observation and research mission of the Earth for 44 years,with its moderate time and spatial resolution,covering the whole world,free accessed and other advantages,Landsat series of satellites were widely used in regional scale forest resource management,investigation and biomass real-time,continuous dynamic monitoring.Landsat 8 OLI sensor,launched in February 2013,was optimized for scanning mode,band spectral range,radiation resolution,etc.compared to TM and ETM + sensors.Obviously,there is a certain difference between the spectral information of OLI and ETM + sensors,and whether the difference of spectral information will be transmitted to the biomass estimation results.Whether this difference affect biomass estimates of a long time series using two sensor images? In this paper,OLI and ETM + sensor images with strip numbers of 119/39 were selected as remote sensing data sources to analysze the differences of spectral information such as apparent reflectance,surface reflectance and NDVI index,taking the variables such as the original band,vegetation index and texture information of OLI and ETM + sensor in Hangzhou area,then using multiple stepwise regression and random forest to estimate the aboveground biomass of the green space in the study area,and analysising the differences between the OLI and ETM + sensor images in the biomass estimation.The results showed that:1.The standard deviation of the difference is less than ± 0.06 between OLI and ETM + sensors.The atmospheric apparent reflectance and surface reflectance of the two sensors are greater than 0.7.The determination coefficient of visible light band,shortwave infrared band and NDVI index is greater than 0.8,and the significance is less than 0.001 between each band and NDVI.The correlation of visible light band,shortwave infrared band and NDVI index is higher than apparent reflectance.So,atmospheric correction can reduce the difference between OLI and ETM +2.The coefficient of stepwise regression were 0.611 and 0.568 based on OLI and ETM+ image.After the feature variables were selected,the R2 of random forest model were increased from 0.735,0.703 to 0.777,0.743.The accuracy of the random forest model was improved after the variable selection,the accuracy of the random forest model was generally higher than that of the stepwise regression.3.The OLI image estimation results had a wider range of values than the ETM+ estimation,the OLI range of 0~102.425 t/hm2 and the ETM + range of 0~99.753 t/hm2.The R2 of the OLI image estimation and the measured value was 0.696,and corresponding the ETM+ was 0.6783.The difference between two images was basically 0,and the correlation between two estimates was high,the R2 was 0.8148,and its significance was less than 0.001.The results of biomass estimation can be converted by YOLI = 0.760 + 0.9667 XETM + or YETM + = 4.631 + 0.8429 XOLI.4.The spectral information of OLI and ETM + have a small difference,which can be used for estimating the biomass of urban green space.The difference between the two sensor biomass estimation results is also small.Therefore,it is not necessary to consider the difference of the two kinds of sensors when using the Landsat series image for long time series biomass estimation.