UTH Pathfinder Retrieval

This page contains latest results for applying the UTH algorithm (Jackson and Bates, 2001, JGR, in press) to the HIRS clear-sky Pathfinder data.

I first began by applying the algorithm to one month of data, to two separate satellites, for the month January 1983. I wanted to look at the bias in UTH when comparing monthly grids that were constructed by applying the alogorithm to the swath data, pentad grid data and the monthly mean data.Figure 1 shows the monthly mean grids when computing the UTH data from swath data and from pentad grid data for NOAA-6. Both monthly grids show a large dry region across the Pacific around 20N. This dry region is associated with the El Nino event of 1982/83. Convective region just south of the equator has missing data indicating that the HIRS clear-sky data was not able to establish any clear-sky observations in this region for the entire month. Missing data regions in the poleward of 30 degrees are related to the algorithm's constraint on the temperature profile. The difference map indicates that the swath-averaged monthly grid is wetter than the UTH data constructed from the pentad grid data. This bias is caused by the swath data seeing the transient behavior of UTH and the exponential relationship between brightness temperature and UTH.Figure 2 gives is the same diagram except for NOAA-7 data. The results are very similar to NOAA-6.

The same analysis was carried out for the monthly mean data derived from the swath data to the UTH data derived from monthly mean data. Figure 3 indicates a significant bias between UTH data derived from the swath data and month mean data for NOAA-6. The bias is greater than the pentad data comparison and in the same direction for virtually all grid cells. Figure 4 for NOAA-7 gives a similar result.

The spatial distribution of the bias can be explained, in part, by a large standard deviation of the UTH derived from the swath data. Figure 5 shows how differences, particularly for the UTH Swath - Monthly data, relates well with the UTH standard deviation map. Large regions of agreement occur in the eastern Pacific and eastern north Africa. The Swath - Pentad does not show significant bias in these large areas but does show smaller regions at higher latitudes that coincide with the standard deviation map. Again, N07 results in Figure 6 gives a similar result to the NOAA-6 data.

A scatter diagram comparing the monthly mean UTH data derived here quanitifies the bias seen in these diagrams. Figure 7 indicates the bias between the swath and pentad derived UTH data is 0.8% and raises to 1.8% for the month-derived UTH. NOAA-7 results in Figure 8 indicate nearly the same bias but the mean values are wetter by about 2%. The intersatellite bias can be attributed to both sampling and instrument bias. RMS error range from about 1% to 1.5% for all results.

The mean bias seems rather small in the scatter diagrams compared to the larger regional biases seen in Figures 1-4. Based on these larger regional biases seen in mainly the UTH derived monthly grid data and to a lesser degree in the pentad data, I recommend not using the monthly mean data for deriving UTH. The problem with using either the pentad grid data or swath data is intercalibration. Currently, we have intercalibrated only the monthly mean data. To intercalibrate the pentad data, I would need to construct all the grid files for all satellites. That would require construction of about 20 Gbytes of pentad grid data. Interpolation of the pentad data may also be an issue; however, this one month study indicates the UTH fields constructed from the pentad data are reasonable filled. Intercalibration of the swath data has not been worked out yet; therefore, UTH data constructed with the swath data would be satellite dependent.

Darren Jackson
Last modified: Tue Aug 7 15:19:22 MDT 2001