Temporally smoothed and gap-filled MODIS fPAR data to improve satellite-derived estimates of gross vegetation productivity

Satellite-derived measurements of terrestrial productivity such as the MOD17 8-day gross primary production (GPP) product generated at the University of Montana (UMT) enable a current and spatially comprehensive understanding of the global carbon cycle. These data are limited by the quality of upstream datasets including the MOD15 fPAR (fraction of photosynthetically active radiation) product. Problems associated with this dataset pertain to poor quality or missing data due to sub-optimal atmospheric conditions or sensor malfunction. The most recent MOD17 product (collection 4.8) utilizes fPAR data (UMT fPAR) that has undergone a temporal filling process based on simple linear interpolation. This process generally acts to increase fPAR, as unreliable data is replaced by linear interpolation of the nearest reliable values. We present an alternate smoothed and spatially gap-filled fPAR product derived using a modified version of the TIMESAT software. This approach provides a weighting mechanism based on the MODIS quality assessment layers maximizing the use of high-quality retrievals to fit an annual curve. The curve is used to replace missing or poor-quality observations. If large gaps exist in the time-series, TIMESAT does not fit a curve and a separate spatial gap-filling procedure is used. Processing has currently only been conducted for North, Central and the northern part of South America. Initial comparisons between four years (2002 - 2005) of the UTM and TIMESAT fPAR reveals that both exhibit similar spatial and temporal patterns reducing the noise that is present in the original MOD15 dataset. In regions with extensive periods of missing data due to high cloud cover, UMT fPAR values show greater deviations from the original high-quality MOD15 fPAR values compared to the TIMESAT-derived fPAR. We utilize the TIMESAT fPAR in the MOD17 GPP algorithm to test the sensitivity of the product to the TIMESAT smoothing algorithm. In addition, we compare seasonal and annual estimates of GPP derived from flux measurements at 30 field sites spanning representative vegetation types across North America.