Boston based micro weather forecasting company ClimaCell has just extended its short-range flood prediction service with global availability and alerts up to 24 hours before, anticipated events such as rivers bursting banks and flooding a nearby town.
ClimaCell’s unique proposition is based on its global network of 500 million virtual sensors to feed data to its Hydro-Meteorological platform designed to predict when floods will strike, including flash floods. The company says it has already deployed this in megacities more prone to flooding, embracing 500 million people across Asia, South America, Africa, and Europe.
Micro weathercasting predated modern computer models when meteorology was in the hands of supposedly local experts, but has grown recently with increased availability of sensor data on the ground and development of finer grained forecasting models over short time scales. There is certainly pent up demand for forecasts down to scales of meters and minutes across numerous sectors from agriculture to sporting events, where it can help determine whether to delay starts and employ covers for example just before a heavy shower arrives.
The idea is to complement rather than compete with large scale computer forecasts from the major agencies, by filling in local detail beyond their scope. The other point is that the relatively high precision data from established stations takes a while to process and input to global forecasting models, such that ultra-short-range local forecasts have still been confined to locals or small firms whose predictions are often unreliable.
This has created room for start-ups like ClimaCell to pick up the baton from the major agencies and enhance their large-scale models by filling in local detail and then applying their specialised models to make bespoke predictions either for specific customers in say agriculture, or for potentially devastating events in a small area, such as floods or severe thunderstorms.
As ClimaCell’s Chief Scientist, Daniel Rothenberg, put it, “we fill in the gaps left by traditional sources of weather observations with hundreds of millions of virtual weather sensors from repurposed wireless and IoT networks. By leveraging a ‘big data’-centric approach to the issue, we can make a significant contribution to improving weather awareness across wide swaths of society. That includes providing weather data to parts of the world which have not typically had access to it — including the developing world.”
Indeed, the company is targeting India, citing World Bank figures suggesting that events such as floods associated with long-term weather variability could cost the country up to 2.6% of its GDP by 2050. Although such figures are somewhat meaningless, there is no doubt that India suffers greatly from events such as floods given its high population density and dependence on agriculture, coupled with the fact it has relatively few functioning weather radars to serve its 1.3 billion people and provide early warning of intense rainfall developing in a local area that could lead to flash flooding.
Even in the US, which has a high density of rainfall detecting radars, there is scope for local flood prediction given the country is prone to intense rainfall that can only be predicted down to neighbourhood level very close to the event.
Given that flash flooding accounts for 85% of overall serious flooding events, usually when rivers burst their banks, the model first predicts rainfall totals in a given catchment area and then assesses whether this is enough to cause an overflow. It then decides by how much it is going to overshoot that threshold to work out how serious the flood is likely to be.
Currently ClimaCell has four flood forecasting products. The first is the Global Urban Flooding Forecast (CGUFF), which sends alerts 24 hours before a possible flood event, using by-the-minute precipitation forecasting. Second is the Global Flooding Forecast (CGFF), providing river discharge and flash flood alerts. This uses a time/space basin scale for improved accuracy.
Third is a much longer-range product that is almost certainly less well proven, the ClimaCell Seasonal Extreme Weather Forecast (CSEWF), aiming to predict the future likelihood of changes in typical seasonal weather, with potential benefits including better water, energy, and agriculture resource management. Finally is the ClimaCell Sub-Seasonal Extreme Weather Forecast (CSSEWF), aiming to estimate changes in the likelihood of a weather event happening within a given week to a month.
Firms such as ClimaCell have a window of opportunity at present, but it will not be long before the big players muscle in on micro weather forecasting. Just two months ago in January 2019 IBM and its subsidiary The Weather Company unveiled a new global weather forecasting system with the claim it would provide the most accurate local weather forecasts ever seen worldwide when released later in the year. This would combine unprecedented computer power with a smaller grid for the forecasting model, with the challenge again being to get sufficiently localized data in on time, with talk of crowdsourcing to obtain that. In a time of climate change, this type of system needs to be in place and attached to many other ioT systems everywhere.