What is "relative error" in TRMM 3B42 and 3B43?

[GES DISC - ingridgs]

Answer:
The "relative error" is the random error discussed below.

The accuracy of the precipitation products can be broken into systematic departures from the true answer (bias) and random
fluctuations about the true answer (sampling), as discussed in Huffman (1997). The former are the biggest problem for climatological averages, since they will not average out. However, for short averaging periods the low number of samples and/or algorithmic inaccuracies tend to present a more serious problem for individual microwave data sets. That is, the sampling is spotty enough that the collection of values over (for example) one day may not be representative of the true distribution of precipitation over the day. For VAR, the sampling is good, but the algorithm likely has substantial RMS error due to the weak physical connection between IR Tb's and precipitation.

Accordingly, the "random error" is assumed to be dominant, and estimates could be computed as discussed in Huffman (1997). Random error cannot be corrected.

The "bias error" is likely small, or at least contained. This is less true over land, where the lower-frequency microwave channels are not useful for precipitation estimation with our current state of knowledge. The state-of-the-art at the monthly scale is reflected in the study by Smith et al. (2006). Studies of the sub-monthly bias have not yet been performed.

References: Huffman, G.J., 1997: Estimates of root-mean-square random error contained in finite sets of estimated
precipitation. Journal of Applied Meteorology, 36, 1191-1201. (Link to full text of paper, PDF version)

Smith, T.M., P.A. Arkin, J.J. Bates, and G.J. Huffman, 2006: Estimating bias of satellite-based precipitation estimates. Journal of Hydrometeorology, 7(5), 841-856. (Link to abstract; full text requires subscription or payment)