When you buy through our articles links, Future and its syndication partners can earn commissions.
USDA technician Dan Palic supports at the weather station. ; | Credit: USDA Photo Scott Bauer. Video number K7688-7
This article was originally published The conversation. The publication contributed to the Space.com article Expert voices: OP-ED and insights;
Farmers are at risk for each planting solution, and many of these risks are increasing due to climate change. One of the consequences is the air, which can harm the harvest and livelihood of the harvest. For example, a delayed monsoon can force a rice grower in South Asia to completely transplant or switch crops, lose time and income.
Access to reliable, timely weather forecasts can help farmers prepare for the coming weeks, find the best time to plant or determine how much fertilizer will be needed, resulting in increased yields and lower costs.
However, in many low and middle -income countries, accurate weather forecasts remain inaccessible, thus limiting the high technology costs and infrastructure requirements for traditional forecasting models.
A new wave of AI -powered weather forecasting models can change it.
Using artificial intelligence, these models can provide accurate, local forecasts through part of the cost of calculating the usual physics models. This allows developing countries to provide national meteorological agencies to farmers timely, local information about the changing precipitation models needed by farmers.
The challenge is to get this technology where it is needed.
Why AI Forecast is important right now
The physics -based weather forecasting models, which are used worldwide, are used by major meteorological centers, but expensive. They mimic atmospheric physics to predict weather conditions, but they need expensive calculation infrastructure. The cost of most developing countries does not reach them.
In addition, these models were mainly designed and optimized by the Nordic countries. They tend to focus on temperate climate, high -income regions and pay less attention to the tropics, where there are many low and middle -income countries.
The main change in the weather models began in 2022, when industrial and university researchers developed deep learning patterns that could create accurate short and mid -range forecasts worldwide for up to two weeks.
These models operated at speeds at a few degrees faster than physically based models, and they could operate on laptops rather than supercomputers. Newer models such as Pangu-Weather and Graphcast have coordinated or even outperformed leading physics-based systems for certain forecasts such as temperature.
AI -powered models require dramatically lower calculation power than traditional systems.
While physics systems may require thousands of processors to perform one cycle of prognosis, modern AI models can do this using one GPU in minutes when the model is trained. This is because the intensive part of the AI model training, which learns the data from the data, can use those learned relationships to create a forecast without further calculation – this is the main link. On the contrary, physics -based models must calculate each variable in each variable in each location and in every forecast.
Better forecasts for the weather can be ensured by safer measures for natural disasters such as hurricanes. | Credit: NOAA via Wikimedia Commons
While teaching these models from physics-based models data, significant pre-investments are needed when teaching, the model can generate large ensemble forecasts-farm forecasts for testing costs of physics models.
Even the expensive step of the AI weather model training shows a lot of calculation savings. One study found that the early model Fourcastnet model can be trained in a supercomput of about an hour. This allowed the time to provide the predicted thousands of times faster than the state -of -the -art, physically based models.
The result of all these advances is the high resolution forecasts worldwide in a few seconds on one laptop or a desktop computer.
Studies are also rapidly improving to expand the use of AI for forecast weeks before the next few months, which helps farmers choose planting. PG models are already investigated to improve extreme weather forecasting, such as extraratropic cyclones and abnormal precipitation.
Forecasts related to real -world solutions
While Ai weather models offer impressive technical capabilities, they are not solutions for Plug and Play. Their effect depends on how well they are calibrated due to local weather compared to real world agricultural conditions and compatible with the factual decisions that farmers have to make, such as what and when to plant, or when drought is likely.
In order to exploit all its potential, the forecasting must be related to people whose decisions are intended to lead.
That is why groups such as Scale goal, cooperation, with whom we work as researchers of public policy and sustainability, help governments create AI tools that meet the real world needs, including training users and sewing forecasts for farmers’ needs. International development institutions and a global meteorological organization also try to expand access to AI forecast models in low and medium -income countries.
PG predictions can be applied in the light of the specific contextual agricultural needs such as setting up optimal planting windows, anticipating dry spells or planning of pests. By spreading these forecasts through text messages, radio, extension agents or programs for mobile, then can help access farmers who can be useful. This is especially true when the reports themselves are constantly inspected and improved to meet the needs of farmers.
A recent study in India found that when farmers received more accurate monsoon forecasts, they made more reasonable decisions on what and how much to plant – whether to plant at all – which resulted in better investment results and reduced risks.
Indian farmers are the main investigation of the case where the weather forecast is needed. | Credit: Michael Gäble via Wikimedia Commons
A new era of climate adaptation
Pg weather forecasting reached the main moment. Tools that were experimental just five years ago are now integrated into government weather forecasting systems. But technology alone will not change life.
Based on low and medium -income countries, it can create opportunities to generate, evaluate and act independently, providing valuable information to farmers who have long been lacking in air services.
This article is repeatedly published from The conversation According to the Creative Commons license. Read Original article;