Here Comes the Rain Again Full Version Here Comes the Rain Again Full Version

Home → News → Here Comes the Pelting Again → Full Text

Here Comes the Rain Again

By Sandrine Ceurstemont
Commissioned by CACM Staff

September 21, 2021

---

---

Floods are the most common natural disaster caused by weather, according to data from the National Severe Storms Laboratory of the U.S. National Oceanographic and Atmospheric Administration, which says floods are expected to become even more frequent, due to climate change. Often acquired past extreme rainfall, floods can threaten lives and cause extensive harm to homes and infrastructure.

Alluvion prediction systems, nonetheless, tin help limit their touch on. Now, a combination of different types of models are used to develop forecasts, which can incorporate physics-based simulations to predict water levels based on surface flow in rivers, for example. Nonetheless, flood prediction system frequently are complex and crave a lot of computer ability and fourth dimension to produce forecasts. Furthermore, measurements such as rainfall are frequently required from conditions monitoring stations on the basis, but all-encompassing networks of such stations typically have only been set upward in developed countries.

"Not every land has admission to those resources, so predictions can be somewhat ad hoc depending on where you are," says John Kimball, professor of systems ecology at the University of Montana in Missoula.

Auto learning, therefore, is of involvement to help make predictions more attainable. One of the advantages to using motorcar learning is that the models can be trained with remote sensing data, such every bit satellite observations, which are available globally and often at no cost. Machine learning systems as well take much less fourth dimension to generate flood predictions. "The ability to brand predictions very quickly is peculiarly what'due south needed with these almost-existent-fourth dimension forecasts to get meaningful data to folks on the footing," says Kimball.

In contempo work, Kimball and his colleagues developed a machine learning model to forecast floods using satellite data. They focused on an expanse of eastern Zimbabwe and a part of neighboring Mozambique where whirlwind Idai triggered a major flooding outcome in March 2019. Their aim was to see whether satellite information could have helped to predict the flood, and therefore could exist used for future forecasts. "In that region, a amend early on warning arrangement is needed because information technology'south inundation-prone," says Kimball.

The team used a standard machine learning model called Classification and Regression Trees that was available within the Google Earth Engine platform. They trained it using data on their region of interest from satellite observations for several years prior to the flooding result, which was bachelor in the platform. Data from the Soil Moisture Active Passive (SMAP) satellite provided information near surface soil wet, while data from another satellite chosen Landsat revealed surface water weather. Global rainfall information, which did not come from a satellite, was used also, to calibrate and verify the model. The system also was validated using data from other satellites.

Kimball and his colleagues constitute their model accurately predicted the flooding acquired by cyclone Idai 24 hours in advance; meaningful, but less-authentic, forecasts also could be obtained as far as three days in advance. "Ideally, a seasonal forecast would be wonderful, only in terms of saving lives, even a few hours' accelerate notice can exist critical," says Kimball.

The team is now working on expanding the forecasting model to other regions. They also are developing a Web-based platform that could be accessed via smartphone, so their model can be used by people on the basis, such equally local agencies helping to relocate people before a potential inundation. "Fifty-fifty the full general public could access maps to see what areas are flooding," says Kimball.

Other data also could be integrated into their organization to provide more detailed flood forecasts. While their current model is able to predict how water will spread in a region over time, for case, it does not indicate the depth of any flooding. "We're working with partners within the region to incorporate other kinds of data, for case stream flow simulations, to make a more than comprehensive assessment of inundation danger," says Kimball.

Although machine learning is promising for predicting floods, information technology too faces challenges. In a recent newspaper, Dennis Wagenaar, a research beau at Nanyang Technological University in Singapore, and his colleagues delved into electric current opportunities for automobile learning, also as concerns. Organizations involved with flood take chances management, for example, often are aware that machine learning is an emerging tool they should utilise, only they may not have in-depth knowledge about it, or its use. Wagenaar said the paper "was aimed at these kinds of people who want to know where information technology is useful and where it is non."

The increasing availability of public information should prove an advantage for machine learning methods. The resolution of available satellite images is constantly improving, for example, while views of streets, such equally 360-degree panoramas provided by Google Streetview, cover increasingly more locations and could become a valuable source of data. Wagenaar thinks algorithms could use street imagery to gauge the height of homes and therefore potential inundation impact, an aspect of predictions that is currently defective. "We tin can predict a flood, only we (often) don't know whether the h2o will enter homes or not," says Wagenaar.

On the other hand, Wagenaar and his colleagues believe automobile learning systems should not completely replace existing methods. Physics-based models, for example, apply well-established formulas to capture sure aspects of flood predictions, such as how h2o flows, and therefore probably can't be improved on using machine learning. Wagenaar thinks the focus should be on components of alluvion forecasts that tin't exist described by existing formulas, such as the bear upon of water depth on people or buildings. "I see more than of a role for machine learning in some parts of the chain (of models) that are fairly complex," he says.

The interpretability of motorcar learning models is a business organisation, besides. If a car learning organization predicts high overflowing hazard in a specific expanse, for example, it could accept serious implications, such as reducing the value of homes, or forcing people to evacuate/relocate. However, predictions tin can be based on relationships identified by algorithms that are not always obvious. "Alluvion models have an impact on communities, and you want to be able to explain those decisions," says Wagenaar.

Machine learning models could be useful to "future-proof" urban areas prone to floods. In new work, Wagenaar and his team are planning to train machine learning models on historical data of urban center growth to extrapolate how a city might expand in fifty years' time, initially using Manila in the Philippines as a instance study. If agricultural fields will be turned into roads or pavement, for case, less water would be absorbed into the ground during extreme rainfalls, which would impact how h2o spreads out over land. And if such an area becomes more than populated, more than people would exist at risk in the event of a flood.

"You might want to invest much more than money actually (in a alluvion-protection wall), if yous take into business relationship that in the futurity, it needs to protect many more people," says Wagenaar. "I think that's very of import for long-term planning."

Sandrine Ceurstemont is a freelance science author based in London, U.Thousand.

No entries found

kendrickinctureniou86.blogspot.com

Source: https://m-cacm.acm.org/news/255673-here-comes-the-rain-again/fulltext?mobile=true