Artificial intelligence predicts more solar energy in Andalusia by 2100, but not equally in all areas
A team of researchers from the University of Córdoba has developed an artificial intelligence model that estimates how much solar energy Andalusia will have by the end of the century based on temperature data. It is a study that combines climatology, energy science, and computer models at a time when Spain is rapidly expanding renewable sources, while the southern part of the country already has one of the most favorable solar positions in Europe. The value of the work lies not only in its scientific novelty, but also in the fact that it attempts to answer a very practical question: where, and to what extent, it will be worthwhile to plan new photovoltaic systems in the decades to come.
The research is signed by Juan Antonio Bellido-Jiménez, Javier Estévez, and Amanda P. García-Marín, affiliated with the Hydrology and Agricultural Hydraulics Research Group at the University of Córdoba. According to the published paper, their model estimates daily solar radiation and then calculates the so-called peak solar hours, an important indicator for the photovoltaic sector because it enables standardized measurement of the annual available energy. In practice, this indicator helps investors, designers, and public authorities understand how much energy a solar panel can receive at a specific location and during a specific period. When such estimates are extended to the period up to 2100, the result is a tool that can influence long-term decisions on energy infrastructure, spatial planning, and adaptation to climate change.
What peak solar hours are and why they matter
The term peak solar hour, also known as peak solar hour, is used in the photovoltaic industry to translate solar radiation into a standard and comparable energy measure. One peak solar hour corresponds to radiation of 1,000 watts per square meter during one hour. This does not mean that the Sun actually shines with the same intensity throughout the whole day at every location, but rather that the total amount of available energy can be expressed in a single format understandable to power plant designers, energy distributors, and public administrations. This is precisely why peak solar hour estimates are useful when selecting locations for solar power plants, comparing the production potential of different areas, and estimating the future yield of installations.
In Andalusia, a region that already today has a strong natural advantage for the development of solar energy, such data are particularly important. The Andalusian Energy Agency already has solar radiation maps that combine satellite and ground data, while the European Commission Joint Research Centre’s PVGIS system serves as an open tool for estimating solar radiation and photovoltaic system production in Europe and beyond. The new research from Córdoba goes a step further because it does not stop at estimating the current situation, but attempts to model the future availability of solar energy based on a variable that is relatively easy to measure and for which extensive historical and projection datasets exist: air temperature.
How the model works and why temperature was a sufficient starting point
According to the authors of the paper, the basic idea was to use temperature as an easily available and widely measured variable from which, with the help of a machine learning model, solar radiation and then peak solar hours can be estimated. This is important because direct measurements of solar radiation are not always equally available at all locations, whereas maximum and minimum daily temperatures, thermal range, and other temperature derivatives are much more often part of meteorological databases. Such an approach opens the possibility that similar models can also be used in areas with fewer resources or weaker infrastructure for the direct measurement of radiation.
The researchers compared four machine learning models and several input data configurations. They tested different combinations of temperature variables, including daily maxima, minima, and thermal amplitudes, in order to determine which approach provides the most reliable results. The best performer was the multilayer perceptron model, or MLP, specifically in the version that included the largest number of temperature variables. The published paper also states that all models outperformed the empirical Hargreaves-Samani method, which is often used in estimates related to climatic and hydrological calculations when more detailed data are lacking.
An important element of the whole approach is also that, after development, the model does not require specialized supercomputing infrastructure for everyday use. The authors state that the computational effort needed for development was significant, but that the finished model can then be run on a standard computer as well. This increases its practical value for other research groups, regional administrations, and energy planners.
Reliability verification at 122 meteorological stations
In order to verify whether the model can indeed reliably convert temperature data into estimates of solar radiation, the researchers validated it against real measurements. For this purpose, they used data from 122 meteorological stations in Andalusia that measured solar radiation between 2000 and 2022. Based on the comparison between estimated and measured values, they concluded that the model performs well enough to be applied to future climate scenarios as well.
It is precisely this validation step that distinguishes a scientifically useful projection from a mere theoretical assumption. Energy planning, especially when discussing investments that last for decades, cannot rely only on an elegant mathematical model. It is necessary to show that estimates derived from temperature have a real basis in measured solar radiation data. In this case, the researchers claim that the model succeeded in reproducing patterns faithfully enough that it can also be used to estimate future changes in the spatial distribution of solar potential.
What the projections say about the period until the end of the century
The main conclusion of the research is that peak solar hours are increasing across most of Andalusia in all analyzed climate scenarios. In the more moderate emissions scenario, according to the paper, average annual values rise from approximately 1,850 kilowatt-hours per square meter annually in the period 2024–2030 to about 1,950 kilowatt-hours per square meter annually by 2100. In the higher-emissions scenario, the increase is more pronounced and exceeds 2,000 kilowatt-hours per square meter annually. The authors also found a generally positive and statistically significant trend for most of Andalusia.
At first glance, such a finding sounds like unequivocally good news for solar energy, but the research also opens the more complex question of the relationship between useful energy potential and climate warming. The increase in available solar energy, namely, does not occur in a vacuum, but together with rising temperatures and changes in climate patterns. In other words, the same processes that can increase available solar energy simultaneously also bring broader risks for agriculture, water resources, public health, and the resilience of infrastructure to heatwaves. That is why the results cannot be read in isolation as a simple announcement of “more sun, more benefits”, but as part of the broader picture of the climate transformation of southern Europe.
Not all parts of Andalusia will have the same trend
One of the more important details of the paper is that the increase is not spatially uniform. Although most of Andalusia shows a positive trend, certain coastal areas and parts of the Sierra Nevada record very weak growth, and in some places even a negative trend in peak solar hours. This means that even in a region often perceived as homogeneously “sunny”, there is no single answer to the question of where future solar yield will be best.
For planners and investors, this is an exceptionally important message. Large regional averages are easy to turn into a political slogan, but decisions on the location of new installations are made at a much finer spatial level. Coastal areas may be influenced by different regimes of cloud cover, humidity, and local air circulation, while mountainous areas have their own microclimatic patterns. This is precisely why a peak solar hour map, produced at high spatial resolution, can be more useful than general regional assessments that do not distinguish local specificities.
Why this matters right now
This scientific result comes at a time when Spain is rapidly increasing its solar and wind power capacities. According to Red Eléctrica data published in March 2026, during 2025 almost 10 new gigawatts of installed wind and solar photovoltaic capacity were brought online, while solar photovoltaic technology alone increased by 8.8 gigawatts. When self-consumption is also included, the country’s total photovoltaic capacity approached 50 gigawatts, making solar the technology with the largest share in the installed capacity of the Spanish power system. At the same time, renewable sources, including the contribution of self-production, accounted for 56.6 percent of electricity generation.
Such development is not accidental. Spain’s updated National Integrated Energy and Climate Plan to 2030 sets the target that as much as 81 percent of electricity should come from renewable sources. In that context, every new method that helps with more precise site selection, better yield modeling, and more efficient alignment of the grid with future generation has direct operational value. Andalusia, because of its climatic conditions, available space, and existing sector development, is naturally one of the key areas of that transition.
From scientific paper to a tool for public policy and investment
The authors of the paper claim that the model is intended not only as an academic experiment, but also as a tool open to other researchers and managers. That openness is important for at least two reasons. First, it enables the verification and adaptation of the model to other areas, which increases scientific credibility and the possibility of comparison. Second, it facilitates the inclusion of the results in real planning processes, from regional energy strategies to local decisions on infrastructure development.
In practice, such projections could be used to compare the future energy performance of different locations, to assess the profitability of new solar fields, and to better understand the relationship between climate and the energy system. They could also be useful when planning grid reinforcements, energy storage, and industrial investments seeking to be located near future renewable sources. In a country striving to accelerate the electrification of transport, heating, and industry, the quality of such spatially detailed energy estimates becomes a matter of economic competitiveness, not just scientific curiosity.
What the results do not mean
It is important, however, to avoid oversimplification. This paper does not mean that every solar investment in Andalusia will automatically be more successful simply because the projections show an increase in peak solar hours. Solar yield also depends on a range of other factors: grid access, topography, module temperature, seasonal production patterns, regulation, cost of capital, energy storage, social acceptance, and environmental impact. An increase in available solar energy does not remove those constraints, but only improves one part of the equation.
Moreover, projections up to 2100 necessarily carry a certain level of uncertainty. They depend on climate scenarios, the quality of input data, and assumptions embedded in the models. This does not reduce their usefulness, but it means that they should be interpreted as an instrument for planning under uncertainty, not as a precise map of the future down to the last decimal place. In that sense, the greatest value of the paper may not be in a single number, but in its ability to show the direction of change, spatial differences, and the relative advantages of individual areas.
Andalusia as a laboratory of the energy transition
Andalusia has long been profiling itself as one of the most important Spanish areas for the development of solar energy. Regional institutions and the national energy framework have for years been developing radiation maps, databases, and regulatory instruments intended to expand renewable sources. Now this is being joined by a layer of predictive analytics that attempts to answer not only the question of how much solar energy the region has today, but also how that potential could change in the decades ahead.
That is precisely why the work of the Córdoba researchers goes beyond a narrowly academic topic. It speaks about the way in which climate change, energy policy, and the development of artificial intelligence will increasingly intertwine in public decision-making. For southern Spain, this may mean more precise targeting of investments and more efficient use of the natural advantage it already has. But at the same time, it is a reminder that the growth of solar potential is not separate from the broader climate reality, but is its direct part. It is precisely the ability to read those two processes together that will determine how successful and sustainable the energy transition will be.
Sources:- ScienceDirect / Applied Energy – summary of the published paper on the projection of peak solar hours in southern Spain until 2100 using a machine learning model (link)- Red Eléctrica – official overview of the state of the Spanish power system and the growth of solar photovoltaic capacity in 2025 (link)- MITECO / Government of Spain – National Integrated Energy and Climate Plan 2023–2030 with the goal of 81 percent renewable electricity by 2030 (link)- Agencia Andaluza de la Energía – description of the regional solar radiation map for Andalusia and the methodological foundations of the data (link)- Joint Research Centre of the European Commission – PVGIS, an open European tool for estimating solar radiation and photovoltaic potential (link)
Find accommodation nearby
Creation time: 2 hours ago