Estimation Of Soil Organic Carbon In Semi-arid Area By Remote Sensing

Soil carbon pool is the largest carbon stock in terrestrial ecosystems,and Soil Organic Carbon(SOC)has a significant impact on the global carbon cycle.The arid and semi-arid region occupies an important position in the global ecosystem and it's a ‘potential desertification area'.And the ecological environment is fragile and vulnerable to natural and human factors.Therefore,accurately estimating regional SOC and analysis spatial distribution rules have practical significance for regional carbon cycle process and regulation mechanism,and at the same time it can provide scientific reference for the study of carbon stocks in semi-arid area.With the development of quantitative remote sensing technology,the use of remote sensing data to measure and monitor SOC has become a new tool,and how to use the remote sensing data to accurately estimate the regional SOC is important extremely.In our study,it took the Otindag sandy lands and its surrounding areas as the research area,and used MODIS and GF-1 WFV from 2014 to 2016 as data sources.Different Soil Organic Carbon Density(SOCD)estimation models were constructed and verificated.Then the the optimal model was used to estimate the SOCD in the study area.Firstly,based on the MODIS data,the improved CASA model was used to estimate the annual average Net Primary Productivity(NPP),and the annual Soil Basic Respiration(SBR)was inverted combined with the soil respiration model.Based on the correlation analysis between the average annual NPP,the annual SBR and the measured SOCD,the optimal SOCD estimation model was constructed and the model was used to estimate the research area SOCD.Secondly,using GF-1 WFV as the data source,the Spatial and Temporal Adaptive Reflectance Fusion Model(STARFM)was used to construct the high spatial resolution NDVI and then estimated average NPP and average annual SBR.Similar to the previous text,The estimation model was constructed and the optimal SOCD model was selected to estimate soil carbon in Otindag sandy land.By comparing different SOCD estimation models,the optimal model of SOCD estimation basedon measured data and multi-scale remote sensing data was discussed,and the natural and human factors that affected the SOCD were analyzed.The main findings were as follows:(1)Based on the measured soil profile data in the semi-arid area,combined with the remote sensing data with high spatial and temporal resolution,SOCD could be more accurately estimated.By comparing different estimation models,compared to the average annual NPP,the optimal SOCD-SBR model that integrated the annual growth rate of vegetation growth and soil respiration had better accuracy.The coefficient of determination(R2)was 0.70 and the accuracy of the model could be achieved 77.87%.Then the R2 between the measured SOCD and estimated SOCD reached 0.86,and the RMSE was 0.266 kg C/m~2.(2)The verification results of the SOCD-NPP and SOCD-SBR based on the MODIS and GF-1 data indicated that the SOCD-SBR model based on GF-1 data was more accurate and the error was lower with the measured SOCD.And it could describe the SOCD and its spatial distribution pattern of the Otindag sandy land in a more detailed and hierarchical manner.There was great significance for fencing,afforestation and sand prevention and control in the study area.(3)The estimation results of SOCD based on the optimal model showed that the spatial distribution of SOCD in the Otindag sandy lands and its surrounding areas was significantly uneven,with large differences between east and west,and the SOCD gradually increased from west to east.As a result,a region with a low SOCD centered on the west and a high soil carbon density centered on the east are formed.The highest SOCD in the study area was about approximately 6.77 kg C/m~2,the lowest SOCD was approximately 0.66 kg C/m~2,and the average SOCD was approximately about 1.22 kg C/m~2.(4)By analyzing the correlation between the measured SOCD and the annual average temperature and annual precipitation,we could see that at the P<0.01 level,the SOCD of the Otindag sandy lands was significantly negatively correlated with the annual average temperature,and at a certain level,SOCD was positively correlated with annual precipitation.On the other hand,it was found that the SOCD was closely related to human activities and its influence was chiefly there was great differences in SOCD in different landuse/landcoverpatterns.The SOCD in the Otindag sandy lands was higher and lower in order of tree>Cultivated land > Shrubwood > grassland.