Leaf Area Index Retrieval And Product Validation Over Heterogeneous Land Surfaces

Leaf area index(LAI),which is defined as one half of the total green leaf area per unit ground surface area,is a critical structural variable for quantifying the exchange processes of energy and matter between the land surface and atmosphere,it is thus identified as a key parameter in most terrestrial ecosystem models.To acquire long-term LAI records at the global scale,several remote sensing LAI products have been generated from various satellite sensors.However,global LAI products are only available at 500 m or even coarser spatial resolution.The coarse resolution introduces more errors for LAI retrievals and product validation because of the heterogeneous nature of Earth's surface.First,since the LAI retrieval algorithm only assumes one land cover type as input for each pixel,land cover mixtures within a pixel introduce the main error in the moderate-or coarse-resolution LAI estimations.Second,long-term ground LAI measurements from the global networks of sites have emerged as a promising data source to validate global LAI product time-series.However,the spatial scale-mismatch issue between site and satellite observations introduces undesired errors in the validation of products because the spatially heterogeneous land surface results in incomparability between observations from sites and satellites.To achieve the goal that how to improve the accuracy of LAI retrievals and evaluate products using the ground LAI measurements from global networks of sites over the heterogeneous land surface,the main content and conclusion of this dessertation are as follows.(1)A new method was developed to improve the accuracy of LAI retrievals by considering the mixture of different land cover types over the heterogeneous surface.A case study for the pixel mixed with water was conducted using the 30-m land cover maps.This method employs the spetral linear mixture(SLM)model and the latest MODIS LAI inversion algorithm to remove the water effect in the generation of high accuracy of LAI retrievals.Results show that the water reflectance in mixed pixels can be calculated accurately,with the absolute errors for Red and NIR band were only 0.0067 and 0.0082,respectively.Moreover,the simulation of water reflectance errors also show that these uncertainties can only cause less than 10% relative errors for LAI retrievals.Additionally,the absolute and relative errors of LAI retrievals caused by water can be achieved about 1.3 and 80%,respectively.The uncertainty of improved LAI retrievals can be reduced by 0.8 for pixels contained large water area.More importantly,the uncertainty of LAI retrievals was not related to the variation of the water area fraction after removing the water effects.(2)An approach was proposed to evaluate the spatial representativeness of site-based LAI measurements,which can improve the reliability of product validation.The spatial representativeness of LAI measurements was graded from Level 0-4 based on the ranges of three indicators,i.e.,DVTP,RAE and CS,and only the Level 0 measurements are the ideal dataset for the product validation.The site-based LAI measurement from Chinese Ecosystem Research Network(CERN)was analyzed in this study.The result shows the RMSE of spatial representativeness error at Level 0 was only 0.094,and the RMSE ranged from 0.256 to 2.564 for Level 1 to Level 4.This evaluation approach can effectively distinguish the quality levels of spatial representativeness in the product pixel grid,and the product validation result can be more reliable using the best representativeness of ground measurements in the validation practice.(3)Based on the ground LAI measurements from global networks of sites,an approach(Grading and Upscaling of Ground Measurements,GUGM)that integrates a spatial representativeness grading criterion and a spatial upscaling strategy was proposed to resolve this scale-mismatch issue and maximize the utility of time-series of site-based LAI measurements.Considering all the evaluation results together,the result suggests that the proposed GUGM approach can significantly reduce the uncertainty from spatial scale mismatch and increase the size of the available validation dataset.Furthermore,GUGM was successfully implemented to validate global LAI products in various ways with advantaging frequent time-series validation dataset.The GUGM can be used to to better understand the structure of LAI product uncertainties and their evolution across seasonal or annual contexts.In turn,this method can provide fundamental information for further LAI algorithm improvements and the broad application of LAI product time-series.(4)The consistency of different LAI products was analyzed to imbue confidence that generating the long-term LAI records based on the use of multiple sensors.A multi-year global LAI product generated from VIIRS and MODIS was used to evaluate for spatiotemporal consistency.Also,the uncertainty of VIIRS product was quantified by utilizing available ground measurements.Results indicate that the LAI retrievals from VIIRS and MODIS are consistent at different spatial and temporal scales.The estimate of mean discrepancy meets the stability requirement for long-term LAI datasets from multi-sensors as suggested by the GCOS.Therefore,the method to generate VIIRS product based on MODIS LAI retrieval alogrithm can be referred for other long-term products using multiple sensors.Additionally,since the clumping effect was taken into account in the LAI retrieval algorithm,VIIRS displays better agreement with true LAI as expected.However,the uncertainty of VIIRS LAI product cannot meet the accuracy requirement of GCOS(<max(0.5,20%)),indicating that the LAI retrieval algorithm needs to be developed in the future.The LAI retrieval algorithm for the pixels mixed with water,spatial representativeness grading method and GUGM method proposed in this dessertation are of great potential to improve the accuracy of LAI retrievals using high-resolution land cover maps and obtain the reliable product validation results using site-based LAI measurements over heterogeneous land surface.