Please use this identifier to cite or link to this item:
Title: Rainfall retrieval from TRMM radiometric channels using artificial neural networks
Authors: Kumar, Rajesh
Das, I M L
Gairola, R M
Sarkar, A
Agarwal, Vijay K
Keywords: Artificial neural network;Brightness temperature;Multivariate regression;Multilayered feed forward
Issue Date: Apr-2007
Publisher: CSIR
PACS No.: 92.60.Jq; 84.35.+i
Abstract:  An algorithm for the retrieval of rainfall has been developed from the radiometric measurements of TRMM microwave Imager (TMI) of Tropical Rainfall Measuring Mission (TRMM) satellite using multilayer feed forward Artificial Neural Network (ANN) over different oceanic and land regions of India and its neighborhoods. The “back propagation with momentum” has been used as a learning algorithm for the ANN architecture. The inputs to the ANN are TMI-brightness temperatures (TB) and the output TMI-rain rates, to demonstrate its capability to retrieve rain rates within limited computer time and with reasonable accuracy. The training data has been split up randomly in three parts, viz. training, validation and test data sets. The performance of the network is evaluated for independent data sets (which were not included in the training) after training and cross validation. Instantaneous precipitation estimates demonstrate very high correlation coefficients with the observed rainfall. Although the rainfall estimation using ANN are influenced by many factors such as the representativeness and sufficiency of the training dataset, the generalization capability of the network to new data sets, seasonal and location changes, it is found that the model can still be used for the retrieval of precipitation at high spatial and temporal resolutions. The ANN is shown to quickly reproduce the results of a long time series of data. The ANN derived rain rates have been compared with the estimates obtained from non-linear multivariate regression (MR) techniques using the identical set of data. It has been found that the ANN method, in general, is far superior to the MR technique in its ability to reproduce rainfall intensity in very short time.
Page(s): 114-127
ISSN: 0975-105X (Online); 0367-8393 (Print)
Appears in Collections:IJRSP Vol.36(2) [April 2007]

Files in This Item:
File Description SizeFormat 
IJRSP 36(2) 114-127.pdf1.97 MBAdobe PDFView/Open

Items in NOPR are protected by copyright, with all rights reserved, unless otherwise indicated.