The research team attacked the problem with the Weather Research and Forecasting model, version 4.4, a non-hydrostatic mesoscale simulation system that has become a workhorse of regional climate research. Crucially, they coupled the standard Noah land surface model with an advanced photovoltaic parameterization scheme that explicitly treats solar farms as a distinct land-use type. The scheme accounts for the radiative, thermal, and aerodynamic fingerprints of PV arrays: it modifies the surface energy balance, adjusts aerodynamic roughness to 0.09 meters based on eddy covariance observations, sets seasonal effective albedo values of 0.15 in summer and 0.13 in winter, and even represents the temperature-dependent conversion of incoming shortwave radiation into electricity. Because panel conversion efficiency falls as ambient temperature and irradiance rise, the model captures the subtle feedback between heating and power output that simpler approaches miss entirely.

To isolate the climate consequences of deployment scale, the team designed four scenarios. The control run, S0, contained no photovoltaic capacity at all. S1 imposed the current actual footprint of 134.4 square kilometers, mapped from 2024 Sentinel-2 satellite imagery. S2 projected a future expansion to 1,452 square kilometers, grounded in Inner Mongolia’s energy policy targeting 120 gigawatts of installed PV capacity by 2030. S3 pushed to the theoretical maximum of 5,166.5 square kilometers, using every parcel of land suitable for solar development. Each scenario ran with a seven-day spin-up followed by a 54-day analysis window covering January through early March 2018 for winter and July through August 2018 for summer, at a fine 3-kilometer resolution over two-way nested domains centered on the Kubuqi Desert and initialized with ERA5 reanalysis data.

Before trusting the scenario differences, the team validated the model against observations from six meteorological stations within the nested domain, with data drawn from NOAA’s National Centers for Environmental Information. Two-meter air temperature correlations reached 0.95 to 0.96, indicating the model reliably captures regional temperature variations. Ten-meter wind speed validation showed moderate-to-strong correlations of 0.49 to 0.77 with root-mean-square errors between 0.63 and 1.59 meters per second. The authors are careful to note that this station-based evaluation constrains the background meteorological fields; the PV-induced anomalies themselves are interpreted from the observation-constrained parameterization and matched scenario-difference experiments, a distinction that matters for how the results should be read.

The headline thermal finding is a daytime heat island that grows with deployment scale. In the maximum scenario, mean daytime warming over the PV area reached 0.41 degrees Celsius in summer and intensified to 0.63 degrees Celsius in winter. The mechanism is fundamentally radiative: PV panels reflect less sunlight than the surrounding desert, whose albedo typically ranges from 0.25 to 0.30, so the arrays absorb more solar radiation and dissipate it primarily as sensible heat. In the S3 run, peak summer sensible heat flux exceeded the control by roughly 130 watts per square meter during the day. The stronger winter anomaly suggests that panels absorb solar radiation far more efficiently than nearby high-albedo sand or snow-covered terrain, altering the surface energy balance most dramatically in the cold season.

Nighttime told a strikingly different story, and one of the study’s most intriguing results is this seasonal asymmetry. In summer, the PV area cooled slightly at night, by 0.10 degrees Celsius in the maximum scenario, because the modules have low heat retention capacity and radiate their stored energy away rapidly after sunset. In winter, however, the PV area stayed warm at night, with a positive deviation of 0.31 degrees Celsius. The explanation involves radiative shielding: the arrays act as a physical barrier between the ground and the sky, intercepting upward longwave radiation from the soil and re-radiating part of that energy back to the surface. Combined with the suppressed turbulent mixing that the panel structure induces under winter’s stable atmospheric stratification, this insulation effect prevents the rapid radiative cooling that normally strips heat from open desert nights.

The mechanical impact proved just as consequential as the thermal one. PV arrays raise the aerodynamic roughness length, generating surface friction that attenuates near-surface winds. In the maximum scenario, the spatially averaged reduction in ten-meter wind speed approached 1.0 meter per second, with summer values of 0.86 meters per second during the day and 0.92 at night, and winter reductions of 1.02 and 0.92 meters per second respectively. The finding aligns with a recent global meta-analysis reporting average wind reductions of more than 63 percent inside PV fields. For the Kubuqi Desert, this stilling effect carries profound ecological implications: calmer near-surface winds mean less wind erosion and reduced sand mobility, directly supporting the region’s ecological restoration goals even as the panels warm the daytime air.

Perhaps the most novel contribution is the diagnosis of what the authors call a heat-moisture coupling pump. Vertical cross-sections of water vapor mixing ratio revealed a wetting signal in the mid-lower troposphere that intensified with deployment scale, with considerable increases exceeding 0.15 grams per kilogram concentrated between the 600 and 800 hPa pressure levels over the deployment region in the maximum scenario. Turbulence diagnostics showed that turbulent kinetic energy averaged over 400 to 800 hPa increased significantly across all scenarios, from 0.19 to 0.31 square meters per square second from S1 to S3, while top-layer soil moisture slightly decreased. The interpretation is an active exchange process rather than passive moisture retention: reduced albedo and enhanced sensible heating provide the thermodynamic forcing, increased roughness strengthens turbulent mixing that consumes near-surface soil water, and reduced wind speed weakens horizontal ventilation, allowing water vapor to be redistributed upward within the PV region.

The spatial footprint of these effects turned out to be far more confined than coarse global models had suggested. Using a multi-buffer analysis that aggregated grid cells into distance bands from the PV areas, the team found that the thermal signal decays exponentially, becoming statistically indistinguishable from model variability beyond roughly 60 to 90 kilometers, with the strongest warming confined to the 0 to 10 kilometer source zone and an estimated thermal influence range of about 30 kilometers. The mechanical wind response was even tighter, essentially within 10 kilometers of the plant boundary. Previous coarse-resolution studies had hinted that large-scale albedo modifications could extend over hundreds of kilometers, and Earth system models have proposed that Sahara-scale solar farms could disturb global circulation and cloud cover as far away as North America and East Asia. The high-resolution results do not contradict those long-range feedbacks, but they do show that the directly attributable mesoscale response is locally concentrated.

The authors frame their findings as a scientific foundation for climate-adaptive planning of large-scale solar bases, and they are candid about the limitations. The simulations rely on static coupling without dynamic vegetation feedbacks, meaning the albedo-precipitation-vegetation feedback proposed for Sahara-scale deployment is not captured, even though the microclimate created by panels may in reality facilitate vegetation recovery that further modifies roughness and transpiration. The analysis covers only selected 2018 seasonal windows rather than multi-year climatological states, and no targeted observations from PV plants were available to validate the perturbations themselves. The moisture diagnostics, while consistent, do not constitute a closed water vapor budget. Still, the core message stands: desert solar farms do more than generate clean electricity. They reshape the local energy budget, calm the winds that drive sand movement, and pump heat and moisture into the desert atmosphere in ways that scale with deployment. As China and other nations push toward carbon neutrality with ever-larger solar bases, the Kubuqi simulations suggest that balancing energy generation with ecological stability will require laying out these installations with the same care applied to any other major land transformation.

Subject of Research: Regional climate effects of desert photovoltaic power plants

Article Title: Regional climate effects of photovoltaic power plant in desert revealed by high-resolution numerical simulation

Article References: Li, S., Cong, Z., Lei, H., Pan, B., Wang, L., & Yang, D. (2026). Regional climate effects of photovoltaic power plant in desert revealed by high-resolution numerical simulation. Frontiers in Earth Science, 14, Article 1894407. https://doi.org/10.3389/feart.2026.1894407

Image Credits: AI Generated

DOI: 10.3389/feart.2026.1894407

Keywords: Regional, climate, effects, photovoltaic, power, plant, desert, revealed, high-resolution, numerical, simulation, scientific research