The future of weather forecasting

Predicting weather accurately doesn’t just help our daily lives but has deeper impact for food security and disaster management. Good news for monsoondependent India is that we are getting better at predicting. New technologies, such as Internet of Things (IOT) and Artificial Intelligence (AI) are helping meteorological experts to give better information to predict agricultural output and natural disasters.

In early 2020, Google researchers released a paper that claimed that they have been able to predict weather just six hours in advance. The researchers trained their AI model to use radar data of cloud movement to predict rains with higher speed and accuracy.

“This precipitation nowcasting, which focuses on 0-6 hour forecasts, can generate forecasts that have a 1km resolution with a total latency of just 5-10 minutes, including data collection delays, outperforming traditional models, even at these early stages of development,” the Google paper said.

Increasingly, unpredictable global and micro weather patterns are emerging and catching meteorologists by surprise.

The Indian Meteorological Department (IMD) and private players like Skymet have been working on various technology models to support farm sector while giving improved warning for potentially disastrous weather phenomenon.

Skymet has now moved to provide hyper local weather predictions in Mumbai. For some years it has been using Automatic Weather Stations (AWS) for gathering and disseminating crucial data. An AWS automatically captures and reports data to the central system at 30-60 minute intervals. It records wind direction, speed, humidity and atmospheric pressure.

Skymet deploys a combination of AI and IOT solutions for generating critical weather information. As the first private weather company in the country, Skymet “owns and operates 6500+ AWS, 600 agricultural sensors, 400+ air pollution sensors and 70+ lightning sensors across 24 states in India.”

The data generated by all these is processed by its Skysense machine learning engine. The information is then customised for banks, insurance companies, farmers and government departments. Cloud detection and movement is also analysed with artificial intelligence-based models to help farmers daily.

Apart from weather, an important service is to use AI for predicting crop production and output using images from drones or satellites.

Algorithms scan high resolution satellite images to automatically derive plant count which can help in yield and production forecasts. This is done through Computer Vision or Machine Learning applications for precision detection and limited human effort.

For example, a high-resolution photo of wheat stalk is assessed for the number of grains it may have. Hidden and partial grains are assessed too. Many such random samples are used to build predictions on how the overall crop yield in a field may be.

Weather data and remote sensing parameters are also used to predict the crop health, crop yield, pest/disease based on the geographical location. Over time, such detailed data mapped with weather and other conditions can help identify patterns with such speed that human researchers would find tough to match. Government and private agencies must intensify the usage of such technology in meteorological predictions. Sending information over IOT devices needs improved connectivity and information for real time predictions. Such information can be critical in times of natural phenomenon. “Met data should be made available freely and the government should allow us to set up weather radars. Access to government super computing facilities will help improve predictions. To encourage private players, defence and space organisations should be allowed to work with us,” says Skymet founder Jatin Singh.

While IMD is an important institution, India should also boost private players for enhancing a strong techled weather services industry. This sector is well suited for self-reliance.

Apart from predicting weather,algorithms can be used to scan satellite images to automatically derive plant count and production estimates