Retrospective Retrieval of Long-Term Global Leaf Area Index (1982-2010) by Fusion of AVHRR and Modis Data

As a key parameter in estimating carbon, water and energy exchanges between the land surface and the atmosphere, leaf area index (LAI) is used as an indispensable input in most climate, hydrology, biogeochemistry and ecosystem models. A global multi-decade long-term LAI record from remote sensing measurements is required for global change modeling and analysis. The Advanced Very High Resolution Radiometers (AVHRR) onboard NOAA 7-14 satellite series observed the earth since 1981 and is a valuable source of data for long-term LAI retrieval. However, retrievals of LAI from AVHRR are often highly uncertain owing to the atmospheric effect, bands information and data quality. In recent decades, more advanced sensors, such as MODIS and VEGETATION, are launched and provide more reliable data for LAI estimation. Combination of observations from these sensors makes it possible to derive long-term LAI records. However, the differences in sensor characteristics and data quality make it difficult to produce a highly consistent long-term LAI series. In this presentation, we show an approach for generating a long time series (1982-2010) of global LAI products through the fusion of MODIS and historical AVHRR data. The main idea is to establish an AVHRR SR-LAI relationship pixel by pixel based on AVHRR reflectance and LAI derived form MODIS data rather than retrieving LAI from AVHRR reflectance based on physical model inversing. In this way, we can partially overcome uncertainties due to the low quality of AVHRR sensor calibration, atmospheric contamination and the assumption of the background (soil, litter, moss, understory) optical property. Since the SR-LAI relationship derived this way is based on the same LAI dataset, it ensures that the LAI from these two sensors are consistent. The following steps are followed to derive the long-term LAI dataset: (1) A LAI series from 2000 to 2010 is derived from MODIS land surface reflectance data based on the global LAI algorithm proposed by Deng et al. (2006); (2) The pixel-based LAI-SR relationship was constructed from the derived MODIS LAI and AVHRR SR for the overlapping period from 2000 to 2006; (3) This relationship was applied to historical AVHRR data with modification of band differences among AVHRR series sensors. The consistency between the AVHRR and MODIS dataset is evaluated by comparing the results in the overlapping years from 2000-2006. The inter-comparison of the long-term dataset and other global LAI products, including MODIS/CYCLOPES, are also performed. The temporal consistency is analyzed. Finally, field measurements in several locations in China with various vegetation types are used to perform validation.