Monsoon effects on surface solar radiation reliability assessed over Phuket

Phuket, the Thai island famous for its beaches, may appear on paper to be an ideal location for solar power, but a new study shows that the monsoon rhythm of the tropical atmosphere makes the island’s solar resource far less predictable than average irradiance figures alone would suggest. Werapong Koedsin of Prince of Songkla University’s Phuket Campus has developed an Earth-observation-based framework that quantifies not just how much sunlight reaches the ground in a tropical monsoon setting, but how reliable and stable that supply is from month to month. The research, published in Theoretical and Applied Climatology, draws on more than a decade of satellite and reanalysis data and offers a transferable method for photovoltaic planning across Southeast Asia.

The motivation for the study lies in a gap that has long troubled solar resource assessment. Most conventional assessments, Koedsin notes in the paper, emphasize the magnitude of the resource, the average amount of energy arriving at the surface, rather than its operational reliability. In the tropics, where deep convective clouds can develop and dissipate within hours, two locations with identical annual average irradiance can present radically different risks to a photovoltaic operator. A grid planner who knows only the mean daily energy figure has no way of judging how often the output will fall short, how drastically it will swing, or how dependable it will be during the seasons when demand for air conditioning peaks.

To close this gap, Koedsin assembled a daily dataset spanning 2015 to 2025 for Phuket Province, producing 4,018 daily observations. All of the underlying data came from freely available Earth observation products accessed through the Google Earth Engine platform, a cloud-based geospatial computing service that has transformed the accessibility of planetary-scale analysis. Daily all-sky Global Horizontal Irradiance, the total solar energy striking a horizontal surface under real atmospheric conditions, was derived from the ERA5 reanalysis produced by the European Centre for Medium-Range Weather Forecasts. Clear-sky GHI, the radiation that would reach the surface if the atmosphere were cloudless, was calculated using established clear-sky models. Cloud cover also came from ERA5, while precipitation estimates were drawn from the Integrated Multi-satellitE Retrievals for GPM, version 7, a NASA product that merges measurements from the Global Precipitation Measurement mission’s constellation of satellites.

The technical architecture of the study rests on four indicators, each capturing a different dimension of the solar resource. The first is cloud-induced radiation loss, defined as the fractional reduction of actual irradiance relative to the clear-sky potential. The second is the Solar Availability Index, or SAI, the ratio of all-sky to clear-sky GHI, which expresses how much of the theoretical maximum is actually delivered. The third is the Reliability Index, RI, which measures the frequency with which daily irradiance remains close to the clear-sky expectation, in effect quantifying how often the resource behaves as promised. The fourth is the Stability Index, SI, which characterizes the temporal variability of daily irradiance, drawing on concepts from the variability metrics developed at Sandia National Laboratories for quantifying irradiance fluctuation. Together, these indicators move the assessment from a single number, average irradiance, to a multidimensional profile of resource quality.

The headline results are striking. Mean daily all-sky GHI over Phuket was 5.38 kilowatt-hours per square meter per day, with a standard deviation of 1.29, compared with a clear-sky mean of 6.83 kilowatt-hours per square meter per day, with a standard deviation of only 0.47. The comparison reveals two important facts at once. First, roughly 21 percent of the island’s potential solar radiation is attenuated by the atmosphere before it reaches the surface, absorbed, scattered, or reflected by clouds and aerosols. Second, the difference in variability between the two figures, a standard deviation nearly three times larger under all-sky conditions, demonstrates that clouds, not the clear atmosphere, are the dominant source of day-to-day uncertainty in the resource.

The seasonal structure of this attenuation follows the monsoon calendar with remarkable precision. Cloud-induced radiation loss was at its lowest, 7.6 percent, in March, during the dry inter-monsoon period when skies over the Andaman coast are typically clearest. By September, deep in the southwest monsoon, the loss had climbed to 31.8 percent, nearly a third of the available solar energy never reaching the panels. The Solar Availability Index traced the same arc, declining from 0.925 in March to 0.686 in September. In practical terms, a photovoltaic array on Phuket that produces close to its theoretical maximum output in late winter delivers barely two-thirds of that potential at the height of the wet season.

Reliability deteriorated even more dramatically than availability. The Reliability Index fell from 98.0 percent in March to 44.2 percent in November, coinciding with the transition into the northeast monsoon, which brings persistent cloud and heavy rainfall to the Malay Peninsula’s west coast. The Stability Index declined in parallel, indicating that daily irradiance during the monsoon months was not only lower on average but also far more erratic. For grid operators, this combination of reduced availability and reduced stability is more troublesome than either alone, because it complicates both energy scheduling and the design of storage or backup capacity. The findings align with earlier theoretical work published in Nature Communications on how solar intermittency propagates into photovoltaic reliability, but they anchor that relationship in a concrete, decade-long tropical record.

To identify the physical driver of these patterns, Koedsin examined the statistical relationship between all-sky GHI and a set of candidate atmospheric variables. Cloud cover emerged as the strongest predictor, with a Spearman rank correlation coefficient of minus 0.69. This single statistic confirms that atmospheric attenuation by cloud is the primary control on photovoltaic resource availability in the region, dominating over other factors such as precipitation, which matters mainly as a proxy for deep convection, or the modest seasonal variation in solar geometry near the equator. Because Phuket sits at roughly eight degrees north latitude, the sun’s declination changes little across the year, so nearly all of the seasonal signal in surface radiation must be explained by the atmosphere rather than by astronomy.

The choice of data sources proved important for a region where ground-based pyranometer networks are sparse. ERA5 has been validated extensively for solar irradiance estimation, including in studies over Indonesia and in comparisons with the COSMO-REA6 reanalysis over Europe, and GPM IMERG version 7 has been evaluated at continental scales for sub-daily precipitation accuracy. By relying on these independently validated products, the study avoids the cost and logistical difficulty of maintaining ground stations in a marine tropical environment, and it demonstrates that a rigorous reliability assessment can be conducted entirely from publicly available data. The full processing pipeline was implemented in Google Earth Engine, and the derived daily dataset and scripts are available from the author upon reasonable request.

The broader implications extend well beyond one island. Thailand and its neighbors have committed to rapid expansion of renewable generation, and the International Energy Agency’s Renewables 2025 report documents accelerating solar deployment across emerging economies. Yet the tropical monsoon belt, which stretches from South Asia through Southeast Asia to northern Australia, remains underrepresented in the solar resource literature compared with the arid and temperate regions where most large photovoltaic markets developed. Earlier satellite-based solar mapping for Thailand, conducted by researchers at Silpakorn University in the 2000s and 2010s, established the feasibility of satellite irradiance retrieval in the tropics, but those efforts, like most resource maps, focused on spatial patterns of mean radiation. The reliability framework introduced here adds the temporal dimension that financiers and grid engineers actually need: not just how much energy a site receives in an average year, but how trustworthy that supply is in an average week.

Koedsin’s framework is also deliberately simple to reproduce. Because it requires nothing more than two public datasets and four derived indicators, it can be applied to any tropical monsoon region with an internet connection and a Google Earth Engine account, from coastal Vietnam to inland Myanmar to the Indonesian archipelago. For a province or country weighing investment in photovoltaic capacity against alternatives such as hydropower or wind, whose resource assessments have long included reliability analysis, the method provides a common language. The paper suggests that sites with high mean irradiance but low Reliability Index scores may warrant hybrid designs, combining solar with storage or complementary generation, that pure resource maps would never flag.

The study arrives at a moment when the economics of solar power in Southeast Asia hinge on precisely these second-order questions. As photovoltaic penetration rises, the marginal value of each new array depends increasingly on when it produces, not just how much. A monsoon-driven availability dip of thirty percent, concentrated in the very months when hydropower reservoirs may also be stressed, is a system-level risk, not a site-level curiosity. By quantifying that risk with eleven years of daily observations and a transparent, reproducible method, the Phuket study offers tropical energy planners something they have rarely had: a reliability forecast built into the resource assessment itself. The next step, the paper implies, is to extend the framework spatially, mapping reliability across entire monsoon-influenced regions rather than single provinces, so that the geography of solar dependability can inform siting decisions before the panels are ordered.