There are various satellite data products that can help predict rainfall. Which one is best for the purpose?
M. Venkatarami Reddy, and Imranali M. Momin from the National Centre for Medium Range Weather Forecasting, and Ashis K. Mitra and D. S. Pai from the India Meteorological Department got together to compare different satellite data products.
The Global Precipitation Measurement (GPM), launched in February 2014, by NASA and the Japan Aerospace Exploration Agency, to provide rain and snow observations in near real-time, came up with two data products.
Integrated Multi-satellite Retrievals for GPM, or IMERG, has an advanced 13-channel passive microwave radiometer and a microwave Imager with dual-frequency precipitation radar at the Ka and Ku band, to estimate light as well as extreme precipitation more precisely.
The other product was Global Satellite Mapping of Precipitation, a multi-satellite rainfall product that blended passive microwave and infrared data. A moving vector with Kalman filter, an algorithm, enhanced the product’s performance in estimating rainfall.
Then there is our own satellite, INSAT 3D, launched slightly earlier, in July 2013. The INSAT 3D imager generates images of the Earth’s disc from a geostationary altitude of 36,000 kilometres every 26 minutes and provides information on various parameters, using six different bandwidths of the electromagnetic spectrum. The combination gave better land–cloud discrimination, could detect surface features such as snow, extracted sea-surface temperature with accuracy…
And then there is the Merged rainfall product from the India Meteorological Department and the National Centre for Medium Range Weather Forecasting, where the near real time data from Global Satellite Mapping of Precipitation is taken as a guess and corrected on the basis of the data from the gauges. The error estimation – between the guess and the correction needed is iterated and the ‘weight’ so gained is applied to the guess from the satellite data.
The team compared these data products for their accuracy in estimating Indian summer monsoon rainfall in 2016. They evaluated these data products against the daily gridded gauge-based rainfall data from the India Meteorological Department, for comparison.
They first re-sampled all rainfall products at 0.25°× 0.25° spatial grids using the linear averaging method. Then they statistically analyzed the data from all the products to understand the spatial distribution of rainfall, and the variation from ground-based rainfall measurements from the gauges.
All the products agree on the gross regional estimates of the mean seasonal rainfall: there are two maximas, one along the west coast and the other in the North-East where there is high summer monsoon rainfall. The north-west India, J&K and southeast peninsular India receives lower rainfall.
But the team found that all products either underestimate or overestimate the rainfall in different geographical locations. They also found that their own Merged and IMERG data products estimate spatial distributions of mean monsoon rainfall realistically, except in areas such as the Western Ghats and the North-East. When they compared the results, the researchers observed that the Merged rainfall product can estimate rainfall better than do the other satellite-based rainfall products.
However, on the sub-regional scale, no satellite-based estimate was very reliable and all need improvement especially in regions where there are mountainous terrains.
Accurately estimating precipitation remains challenging. We need more rain gauges to correct biases in satellite products, especially from inaccessible areas in difficult terrain. The report’s assessment of currently available satellite data products can help tweak existing algorithms to correctly estimate precipitation, once more data from such regions are available.
Int. J. Remote Sens., 40: 4577-4603;
STEAMindiaReports: steamin' hot science news tips for media channels