Yimin Ji

An overview of the TRMM/TSDIS fire algorithm and product

Since December 1997, the Tropical Rainfall Measuring Mission (TRMM) has provided tremendous opportunities for the understanding of global rainfall variability and mechanisms, as well as for retrieving and understanding of surface properties such as sea surface temperature and forest fires. The TRMM Science Data and Information System (TSDIS) fire algorithm routinely produces daily and monthly global fire products. These products have been achieved since January 1998. Any interested users may download these products from the TSDIS web page. The TSDIS night-time fire algorithm is a traditional threshold method using only the Visible and Infrared Scanner (VIRS) thermal band brightness temperatures. The daytime algorithm is a contextual approach that uses VIRS visible/near-infrared band reflectance and the University of Maryland (UMD) 1 km land type data to reject false fire pixels. This study compared the TSDIS fire products with the European Commission (EC) Joint Research Center (JRC) 1 km AVHRR fire products. The two products were in agreement on spatial and temporal variations over areas where both data were available. An analysis of the TSDIS monthly fire products during the two years of 1998 and 1999 manifested seasonal cycles of biomass fires over Southeast Asia, Africa, North America and South America. The data also showed interannual variations associated with the ENSO cycle in Central America and the Indonesian region during the two years. We compared the TSDIS monthly fire products with the aerosol index products of the Total Ozone Mapping Spectrometer (TOMS) of the NASA Earth Probe satellite. The results indicated that the major aerosol variations of the global atmosphere during the two years were associated with the smoke released by the biomass fires.

Ground validation of TRMM and AMSU microwave precipitation estimates

Global rain gauge data were used to validate the instantaneous and 3-hourly TRMM TMI and AMSU microwave rainfall products for a two-year period from January 2001 to December 2002. In the comparison of instantaneous rainfall products, all satellite data were aggregated to a 0.1/spl deg//spl times/0.1/spl deg/ grid with gauge location at the grid center and compared to the 60-minute gauge rain. The biases against gauge rain are +15% and -20% for TMI and AMSU land/coastal rain estimates, respectively. The gauge data from ocean area in this study were only from Kwajalein, The biases as compared to the Kwajalein gauge data are +12% and +65% for TMI and AMSU ocean rain estimates respectively. The correlation is about 0.56 between TMI and gauge data and about 0.50 between AMSU and gauge data.

TRMM fire algorithm, product and applications

The TRMM Science Data and Information System (TSDIS) fire algorithm is a contextual approach that uses the TRMM Visible Infrared Scanner (VIRS) thermal channel brightness temperature, visible/near-infrared channel reflectance, and global land type data to retrieve fire pixels. The products include global images of daily hot spots and monthly fire counts at 0.5/spl deg//spl times/0.5/spl deg/ resolution, as well as text files that details necessary information of all fire pixels. In order to understand the variability of global land fires and their effects on the distribution of atmospheric aerosols, statistical methods were applied to the TSDIS fire products as well as the Total Ozone Mapping Spectrometer (TOMS) aerosol index products for a period of four years from January 1998 to December 2001. The statistical results showed contrast between North and South hemispheres and also inter-continental transitions in Africa and America. These analyses also identified 25-60 day intra-seasonal oscillations that were superimposed on the annual cycles of both fire and aerosol data. The intra-seasonal variability of fires showed similarity of Madden-Julian oscillation mode. The VIRS fire data were also compared with coincident TRMM rainfall data to investigate the interaction between fire and cloud.

Diurnal and Seasonal Cycles of Land Fires from TRMM Observations

This Chapter summarizes methodologies of detecting land fires from the TRMM/VIRS measurements. The TRMM science data and information system (TSDIS) fire products include global images of daily hot spots and monthly fire counts at 0.5°×0.5° resolution, as well as text files that details necessary information of all fire pixels. These products have been archived since January 1, 1998. Diurnal and seasonal cycles of TRMM land fire products in the Eastern United States, Africa, as well as the South America and Asia, are discussed and compared to other satellite products. Statistical methods were applied to the TSDIS fire products as well as to the total ozone mapping spectrometer (TOMS) aerosol index products for a period of seven years from January 1998 to December 2004. The variability of global atmospheric aerosol is consistent with the fire variations during this period. The TRMM fire products were also compared to the coincident TRMM rainfall and other rainfall products to investigate the interaction between rainfall and fire.

Diurnal cycle and intra-seasonal variability of TRMM rainfall and fire products

The diurnal cycle from 7-years TRMM TMI rain data showed validity as compared to those from ground observations and TRMM merged products. We show that the variances of diurnal cycle are less than 5% of intra-seasonal variances over tropical oceans. Parallel spectrum analyses on TMI and Global Precipitation Climatology Project (GPCP) rain products indicate a near-perfect match of intra-seasonal spectral distribution. However, the variance of diurnal cycle can be as large as 25% of the intra-seasonal variance over land. Inter-comparisons of TMI and GPCP, as well as TRMM merged data indicate a false or exaggerated 25-day oscillation of single TMI rainfall product for a 10 o х 10 o grid box in Amazon and western Africa areas. The diurnal cycle of TRMM land fire products and its aliasing effect on intra-seasonal variability are also studied using TRMM VIRS data and independent rain data in the past seven years. We show that the seasonal and intra-seasonal fire variations are clearly related to the rainfall variability during fire seasons over tropical land. False fire diurnal cycle and intra-seasonal variations occur frequently during the non- fire seasons.

Comparisons of real-time microwave rainfall products from TRMM/TMI, DMSP/SSM/I, and NOAA/AMSU

The real time rain estimates from the TRMM/TMI and the SSM/I Goddard Profiling Alforithm (GPROF), along with the Climate Prediction Center real time IR rain estimates, have been merged to a preparatory version of the TRMM global 3-hourly rain product. This paper discusses the issue of merging NOAA/AMSU rain estimates as an addition into this preparatory product and provides comparisons of rainfall among TMI, SSM/I GPROF and AMSU products. The orbit data of the three microwave sensors were aggregated to 3-hourly gridded rain products with horizontal resolution of 0.25°. The histogram data based on the coincident rainfall among TMI, SSM/I and AMSU estimates from the 3-hourly data series showed about +50% bias of AMSU rain as compared to the SSM/I-TMI rain over tropical ocean. However, over land, the AMSU data showed about -20% bias as compared to the SSM/I-TMI products. The correlation between TMI and AMSU oceanic data is about 0.5. In order to justify the results from microwave rainfall estimates, the TMI rain estimate were compared with ground validation data for a 4-year period. The results indicate a lower (-15%) TMI rain amount over ocean and a higher (+25%) TMI estimate over land as compared to the ground observation.

Seasonal and interannual variations of global fire and aerosol from satellite observations

Statistical analyses of the Tropical Rainfall Measuring Mission Science Data and Information System (TSDIS) pentad fire composite data identified 25-60 day intra-seasonal oscillations that are superimposed in the annual cycles of global bio-mass burnings. The results also indicated interannual variations of fires associated with 1998/1999 ENSO cycle in Indonesian and Central America. Similar statistical methods were used to analyze the Total Ozone Mapping Spectrometer aerosol index product. The results were consistent with those derived from the TSDIS fire data.