Deep learning–based downscaling of summer monsoon rainfall data over Indian region

Abstract

Accurate short-range weather forecasting has significant implications for various sectors. Machine learning based approaches, e.g. deep learning, have gained popularity in this domain where the existing numerical weather prediction (NWP) models still have modest skill after a few days. Here we use a ConvLSTM network to develop a deep learning model for precipitation forecasting. The crux of the idea is to develop a forecasting model which involves convolution based feature selection and uses long term memory in the meteorological fields in conjunction with gradient based learning algorithm. Prior to using the input data, we explore various techniques to overcome dataset difficulties. We follow a strategic approach to deal with missing values and discuss the model's fidelity to capture realistic precipitation. The model resolution used is ( 25 km). A comparison between 5 years of predicted data and corresponding observational records for 2 days lead time forecast show correlation coefficients of 0.67 and 0.42 for lead day 1 and 2 respectively. The patterns indicate higher correlation over the Western Ghats and Monsoon trough region ( 0.8 and 0.6 for lead day 1 and 2 respectively). Further, the model performance is evaluated based on skill scores, Mean Square Error, correlation coefficient and ROC curves. This study demonstrates that the adopted deep learning approach based only on a single precipitation variable, has a reasonable skill in the short range. Incorporating multivariable based deep learning has the potential to match or even better the short range precipitation forecasts based on the state-of-the-art NWP models.

Machine learning‐based statistical downscaling of wind resource maps using multi‐resolution topographical data

“…ERA5 has a better performance than other reanalysis data, 29 contains global long‐term data from 1979 to the present, and is easily accessible. Therefore, many previous downscaling studies 17,30–32 have used ERA5 as input data. As ERA5 includes hundreds of hourly weather variables, it is necessary to select and process the wind‐related variables.…”

“…ERA5 has a better performance than other reanalysis data, 29 contains global long-term data from 1979 to the present, and is easily accessible. Therefore, many previous downscaling studies 17,[30][31][32] have used ERA5 as…”

“…14,15 Another approach is regression or machine learning (ML) using other highresolution variables. Several studies have been conducted for various weather factors 16,17 and wind maps 9,[18][19][20][21][22][23] using these methods, which show better results than spatial interpolation because they can consider other variables such as topography. Many ML techniques such as neural networks, 19 support vector regression (SVR), 20 random forest (RF), 9,21 and deep learning 18 have been used.…”

A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts

“…Many authors have approached the problem as a pure super-resolution task by coarsening their 'truth' data and inputting this to their model, then trying to retrieve the lost resolution. Papers that take this approach include Sha et al (2020), F. Wang et al (2021), and Kumar et al (2021). However, we argue that this is not sufficient to tackle the full downscaling problem, since it does not account for the inevitable errors in the input forecast data.…”

Identifying meaningful covariates that can improve the interpolation of monsoon rainfall in a low‐lying tropical region

“…The meticulous investigation in regional context in other parts of the world is also under studied. There are several studies carried out in monsoon rainfall context (Hussain et al ., 2010; de Amorim Borges et al ., 2016; Gonga‐Saholiariliva et al ., 2016; Kumar et al ., 2021), however, a little attention has been given to identify the meaningful covariates to be employed. Therefore, there is an opportunity to explore the covariates that are in fact making an impact to the interpolation accuracy.…”