Please use this identifier to cite or link to this item: http://dr.iiserpune.ac.in:8080/xmlui/handle/123456789/6390
Title: Application of Convolutional Recurrent Neural Network in Precipitation Forecasting
Authors: Kumar, Bipin
Chattopadhyay, Rajib
ABHISHEK, NAMIT
Dept. of Data Science
20161023
Keywords: Machine learning
Convolutional Recurrent Neural Networks
Medium Range Precipitation Forecasting
Spatiotemporal Forecasting
Deep Learning
Issue Date: Aug-2021
Citation: 50
Abstract: In this work, we use convolutional recurrent neural network-based architectures involving ConvLSTM and ConvGRU for precipitation forecasting over the Indian region. We first compare direct and iterative approach for forecasting and find the iterative approach better for our model. We use multiple variables such as specific humidity, orography, soil moisture and surface pressure as input features for better capturing the underlying dynamics. We also analyse the forecasts over different homogeneous rainfall regions. Finally, we compare ConvLSTM and ConvGRU based models. We find similar performance in both the models, ConvGRU being faster.
URI: http://dr.iiserpune.ac.in:8080/xmlui/handle/123456789/6390
Appears in Collections:MS THESES

Files in This Item:
File Description SizeFormat 
Thesis_20161023.pdf5.37 MBAdobe PDFView/Open    Request a copy


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