From The Eye Of The Storm, AI Weather Prediction Is Impacting Retail And Logistics Decision-Making

A heat dome over Europe, fatal flash floods in Texas, wildfires in Arizona and Colorado national parks. That is just the past few weeks in the life of planet Earth as extreme weather events become more common and unpredictable.

An average of 3.4% of retail sales are directly affected annually by changes in the weather, according to research published by the American Meteorological Society. Applying this globally means the weather alone influences about $1T in retail sales every year. And that, in turn, affects real estate decision-making.

Planning for weather-driven demand, climate sensitivity and weather volatility across stores and distribution networks has become especially crucial for retailers and logistics operators in not only meeting consumer demand but also helping their inventory teams have the right product in the right place at the right time to avoid empty shelves.

Now, these firms and the real estate owners they lease from are turning to increasingly sophisticated artificial intelligence tools to help them plan when and where to put stock and goods — and thus the type of real estate they need, and where. 

“Advanced modelling tools are something we are seeing more requests for on a regular basis,” said Michael Carson, Cushman & Wakefield head of supply chain and logistics advisory for Europe, the Middle East and Africa.

These technologies help simulate disruptions and identify optimal responses across an end-to-end distribution network, Carson said, although he warned there are no silver bullets when it comes to organising a supply chain. 

“Rather than holding inventory for broad markets, such as the entire UK, retailers can tailor stock levels to specific cities,” he said. “This granular approach not only improves operational efficiency but also supports more resilient and responsive supply chains.”

This has implications for logistics real estate and last-mile delivery, with more small fulfilment centres providing more flexible warehouse formats in urban and high-demand areas, supporting rapid delivery and meeting seasonal or event-driven demand spikes.

“Retailers are also segmenting their store networks not just by geography but by typical causal factors, such as weather sensitivity or local event cycles,” Carson said.

Predicting the weather has for decades been a labour-intensive job and provided varying degrees of accuracy. Computer technology has provided consistent improvements, but AI is set to create an exponential leap in the speed and accuracy of forecasting. 

Traditional models rely on physics-based simulations requiring huge computational resources and taking hours to generate forecasts. By contrast, AI enables the creation of data-driven models that can process vast amounts of meteorological data more efficiently.

Nvidia has been a key player, using its AI frameworks to train models that can predict weather patterns in minutes rather than hours. In June, it announced its Earth-2 project, which is building a digital twin of the Earth to simulate global climate conditions, collaborating with weather agencies to develop an AI model that delivers forecasts much faster.

Forecasting specialist AccuWeather uses a hybrid AI approach, combining machine learning with the expertise of its meteorologists, using inputs from 190 sources and collaborating with national meteorological and hydrological services worldwide to create what it says is the largest repository of weather information on Earth.

“From severe weather disruptions in the next week to seasonal trends such as answering questions on how much snow will fall this winter, AccuWeather’s database allows retailers to best align supply chains to future weather patterns and predict shopper behaviour,” AccuWeather Vice President of Forecast Operations Dan DePodwin said.

One approach being taken by retailers is regionalisation within a global framework, although this is not always feasible. Advanced algorithms enable highly accurate, hyperlocalised predictions, allowing stock to be positioned strategically across a network.

Walmart has worked with technology partners to build AI models that help determine where to hold excess stock ahead of predicted climate disruptions. Before the 2024 U.S. hurricane season, Walmart redistributed inventory from coastal distribution centers in Florida to inland facilities in Georgia and Tennessee.

And in February this year, Walmart announced a partnership with tech business Helios to improve its evaluation of medium- and long-term climate risks for the global agricultural supply chain, which predicts the price and availability of agricultural commodities using climate risk and AI.

Climate events that impact the consumer also have a huge influence on the products that shoppers need in a hurry, which is becoming an increasing focus for many retailers as they race to try and access emergency stock, especially food and drink essentials.

Extreme weather events are critical moments for customers and communities, said H-E-B Senior Product Manager Randy Bermea. He said at the National Retail Federation Big Show earlier this year that the Texas-based grocery chain has leveraged demand analytics to better understand behaviour “before, during and after” a climate event.

“We can use historical data and analytics to more accurately predict demand for specific items, helping us to better prepare in advance, which is important,” he said. “It also means we can inform our supply chain teams so they can prepare for spikes.

“We understand that weather doesn’t just influence shopping in extreme weather events. Weather can impact daily shopping, and at a macro level, we can use data to correlate how rainy days might depress demand across specific categories.”

To manage the growing complexity of climate impacts, supply chain operators and logistics real estate firms are increasingly deploying AI and machine learning to predict disruptions, assess asset vulnerability and guide investment.

Prologis uses a proprietary AI-driven model to simulate localised flooding, wildfire and heat risks across its 1.2B SF logistics portfolio. The results help influence whether it decides to retrofit buildings with flood defences, install solar roofs or even offload vulnerable assets.

Prologis has also been working with Munich Re.

“Munich Re data points to an upward trend in insured losses that is closely linked to increasingly exposed assets,” Munich Re spokesperson Uta Apel said. “Major catastrophes, such as extreme hurricanes and earthquakes, are clearly responsible for the significant volatility in the scale of losses.

“The trend toward increasing insured losses, on the other hand, is being driven particularly by nonpeak perils such as severe thunderstorms with hail, tornadoes, flooding and forest fires.”