Research On Remote Sensing Retrieval And Validation Of Typical Vegetation Canopy Parameters

Leaf area index(LAI)is an important canopy structure parameter to characterize the vegetation growth status,and has been a hot spot for terrestrial vegetation canopy studies for many years.As an important tool for regional large-scale monitoring,remote sensing retrieval is of great significance for global regional LAI assessment.However,the validation of remote sensing LAI products is lagging behind,which greatly restricts the application and development of remote sensing data.To address this problem,thesis takes three typical vegetation types of crops,woodland and grasslands as examples,and uses ground truth data obtained by LAI grouping monitoring system to verify the remote sensing validation products of canopy parameters represented by LAI,and carries out the study of scale conversion method and time series effect.The main research contents are as follows:(1)The retrieval of Sentinel-2 leaf area index based on PROSIAL algorithm was performed,and the ground data was used to verify the accuracy of the retrieval results.The study found that the official SNAP algorithm can be better applied to crops and grasslands,and there is a very significant correlation between the retrieval results and ground data(p<<0.01),and the correlation coefficients reach 0.82 and 0.68,indicating that Sentinel-2 retrieval results have high accuracy and can reflect the real situation on the surface.However,for the woodland area,most of the pixel values are exceed the input range of the algorithm,and the algorithm has low applicability,and there is still much room for improvement.(2)Combined with the imaging principle of MODIS and the response characteristics of the sensor to the light source,the mean method,Gaussian point spread function(GPSF)and triangular point spread function imaging method(TPSF)are used for the scale conversion study,and are compared and evaluated with various methods such as overlay resampling method(CR),pixel aggregation and bilinear resampling.The comparison results show that GPSF and TPSF have the best performance in upscaling conversion results,with Moran index reaching 0.85 and 0.83,and similarity with standard data structure SSIM of 0.958 and 0.951 respectively,both of which have high conversion accuracy.(3)In order to verify the accuracy of MODIS LAI,Sentinel-2 was used as the medium,and the CR and TPSF methods in(2)were used to upscale the Sentinel-2 LAI data and to validate the MODIS LAI.The results of data analysis showed that the conversion results of the two methods and MODIS LAI showed good correlation,with correlation coefficients of 0.7 and 0.74,respectively,and MODIS LAI was somewhat overestimated in a few areas such as towns and cities,and underestimated in most areas.(4)In order to verify the temporal variation law of MODIS Lai products,the MODIS LAI products and ground data of Yucheng,Guyuan and Qianyanzhou experimental stations from March 2020 to March 2021 were analyzed for growth trends,and the results showed that: in crops and grasslands,the MODIS LAI products and ground data variation trends have high consistency,and the correlation coefficients of the two reached 0.71 and0.67,indicating that MODIS LAI can better reflect the real variation pattern of ground LAI;however,there are more abnormal values of MODIS LAI products in woodland,and the data fluctuate greatly,indicating that there is greater uncertainty of remote sensing retrieval products in woodland.