Characteristics of brightness temperature with respect to rain rate over ocean and land and its implication on rain rate retrieval

This paper studies the characteristics of brightness temperature (BT) with respect to rain rate over ocean and land and its subsequent implication on rainfall retrieval utilizing artificial neural network (ANN). The BTs data are obtained from Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI). Variations of BT with rain rate are compared for all the nine channels over ocean and land, and over ocean 10.67 GHz is found to be most sensitive to rain rate. For 37.0 GHz channels, emission effect is not noticed over land, but it is prominent over ocean for lower value of rain rate. Over land, lower frequency channels (Nos 1-4) are least sensitive to rain rate. From a quantitative comparison of the 85.5 GHz channels over ocean and land, BT difference of ~10 K is observed. It is also observed that polarization difference between the BTs for this frequency channel is more over ocean. A difference of ~10 K and ~3 K are noticed over ocean and land, respectively. Neural networks are trained over ocean and land to give rain rate as output. The nine channel BTs from TMI are considered as input for training the networks. The collocated near-surface rainfall rate data from TRMM precipitation radar are considered as target. Correlation values of 0.92 and 0.82 are observed for validation set over ocean and land, respectively, with root mean square error (rmse) values of 2.26 mm/h and 3.54 mm/h. It is observed that rain rate retrieval is better over land after discarding the lower frequency channels. The ANN retrieved rain rate is compared with TMI rain rate and correlations of 0.71 and 0.57 are observed over ocean and land, respectively.