Climate change threatens cooling costs in Middle East office buildings, study shows

A research team has modeled how rising temperatures will reshape energy demands inside a modern office building in Riyadh, Saudi Arabia, over the next few years—and the picture is complicated. The building, a contemporary structure in one of the world's hottest cities, sits at the center of a study that combines a decade of real electricity bills with machine learning algorithms and physics-based building simulations to forecast what climate change means for cooling costs and indoor comfort.

The researchers chose this particular building because it offered something rare: ten years of actual monthly electricity consumption data, from 2009 through 2019, paired with detailed architectural information—floor plans, wall materials, window specifications, occupancy patterns, and the mechanical systems that keep the interior habitable. That foundation of measured reality is what made the modeling possible. They fed historical climate variables—temperature, solar radiation, humidity—into two different machine learning approaches: one called extreme gradient boosting and another called long short-term memory networks. These algorithms learned from the past to predict the future. They also ran a calibrated building energy simulation called EnergyPlus, which uses the laws of physics to model how heat moves through walls and how air conditioning systems respond.

The team then asked: what if the climate shifts according to two different warming scenarios—a moderate one and a more severe one—between 2020 and 2024? The models produced an unexpected result. Energy consumption did not simply climb in a straight line as temperatures rose. Instead, the projections showed unpredictable swings, sometimes up, sometimes down, depending on which scenario and which model was doing the predicting. The thermal comfort indices—measures of how pleasant or unpleasant the indoor environment would feel to an occupant—shifted toward worse conditions across the board. The building's cooling system would face a more hostile climate outside, and the people inside would experience less comfortable conditions, at least according to the simulations.

When the researchers compared the two machine learning models head-to-head, the long short-term memory network performed slightly better at predicting both historical energy use and future thermal comfort metrics. Its error margins were tighter. But when they applied rigorous statistical tests to see if this difference was genuinely meaningful or just noise in the data, the tests came back negative. The models were close enough that no clear winner emerged. The researchers were careful to note this limitation: they had only one building, one decade of data, and no actual measurements of how comfortable people felt inside. The thermal comfort predictions came entirely from computer simulations, not from surveys or sensors placed in occupied spaces.

This restraint matters. The study's authors explicitly cautioned that their findings apply to this specific building in Riyadh and should not be stretched to describe what will happen across the city's entire office building stock. Other buildings have different shapes, materials, orientations, and cooling systems. They may respond to climate change in different ways. To truly understand the threat to Riyadh's commercial real estate, researchers would need to test the same approach on multiple buildings, validate the models against independent data, and see whether the patterns hold. What this study does show is that the interaction between climate change, building physics, and energy consumption is intricate enough that simple predictions fail. It shows that the tools exist to investigate these questions building by building. And it shows that in a city where air conditioning is not a luxury but a requirement for survival, the stakes of getting this right are high.