Reference evapotranspiration, abbreviated ETo, is the backbone of irrigation planning worldwide. It represents the rate at which water would be lost from an idealized, uniformly green surface with ample moisture, and it is the quantity that farmers, water managers, and hydrologists multiply by crop coefficients to estimate real irrigation demand. The internationally accepted method for computing it is the FAO Penman–Monteith equation, a physically based formula that combines temperature, humidity, wind, and solar radiation into a single estimate. Because the equation draws on so many inputs simultaneously, a long-standing puzzle has been which of those inputs matters most, and whether the answer shifts from place to place and decade to decade. That question has become urgent as climate change alters the individual variables at different rates.

Zarei and Mahmoudi assembled a dataset spanning 1967 to 2024, one of the longest windows ever applied to this problem in the Middle East. From the records of 34 synoptic stations, they extracted nine candidate drivers: maximum, minimum, and mean temperature; maximum, minimum, and mean relative humidity; wind speed; sunshine hours; and precipitation. For thirteen representative years distributed across the nearly six decades, they computed ETo at every station using the FAO Penman–Monteith equation and then interpolated the results into continuous spatial maps using the Inverse Distance Weighting method, a technique that estimates values at unsampled locations by weighting nearby observations according to their distance.

To move beyond station-by-station analysis, the researchers overlaid 2,000 randomly distributed points across Iran’s territory and extracted both climatic data and interpolated ETo values at each one. This created a vast statistical sample capturing the country’s extraordinary climatic diversity, from the humid Caspian coast to the arid central deserts and the mountainous northwest. The random-point approach meant that the subsequent modeling reflected the spatial character of the whole country rather than the arbitrary placement of weather stations, which tend to cluster in populated areas.

The heart of the study lies in its dual statistical framework, and this is where the methodological story becomes genuinely interesting for anyone who follows climate science. The first tool, the Corrplot approach, examines simple linear correlations between each climatic variable and ETo, identifying which variable shows the strongest individual association in each year. The second, the Lindeman, Merenda, and Gold method, usually abbreviated LMG, belongs to a family of techniques called variance decomposition. Rather than looking at each variable in isolation, LMG averages the contribution of each predictor across all possible orderings in which it could enter a regression model, thereby partitioning the explained variance fairly among predictors that are correlated with one another.

Why does that distinction matter so much? Climatic variables are notoriously interdependent. Maximum temperature and sunshine hours rise and fall together; humidity tracks inversely with temperature; precipitation suppresses both sunshine and temperature. In statistical language, this is multicollinearity, and it is the bane of attribution studies. A naive correlation analysis can assign importance to a variable that is merely riding on the coattails of another. The LMG method, by decomposing R-squared across all orderings of predictors, corrects for this shared variance and reveals which variable carries genuine, independent explanatory power. The contrast between the two methods in this study is a textbook demonstration of why the correction matters.

According to the Corrplot results, maximum temperature was the most frequently dominant variable, leading in 46.16 percent of the years examined, followed by sunshine hours at 23.08 percent, with precipitation and wind speed each dominant in 15.38 percent of years. That picture is broadly consistent with intuition in a warming country: hotter days, more evaporative demand. But when the LMG variance decomposition was applied, the hierarchy flipped dramatically. Sunshine hours emerged as the most frequently dominant factor in 92.31 percent of the study period, with maximum temperature leading in only 15.38 percent of cases. In other words, much of the apparent importance of temperature in the simple correlation analysis appears to reflect its correlation with sunshine duration rather than an independent physical influence of that magnitude.

The authors validated both the spatial interpolations and the statistical models rigorously. The quality of the IDW-generated maps was assessed using the Nash–Sutcliffe Efficiency coefficient, a standard hydrological goodness-of-fit measure, and Willmott’s Index of Agreement, which evaluates how closely predicted values track observed ones. Both metrics confirmed the reliability of the spatial products at the 1 percent significance level, and the LMG model’s performance was likewise statistically verified at the 1 percent level for every year investigated. These are demanding thresholds, and their satisfaction across nearly six decades of data lends considerable weight to the central finding.

The practical implications for Iran are substantial. The country is already one of the most water-stressed in the world, with shrinking aquifers, recurrent drought, and intense competition between agriculture and urban demand. If sunshine hours, a proxy for the solar radiation that supplies the energy for evaporation, are the true dominant lever on ETo, then projections of future water demand need to track cloud cover, aerosol loading, and radiation trends as closely as they track temperature. It also suggests that the well-documented global phenomenon of dimming and brightening, in which the solar radiation reaching the surface has changed over decades due to aerosols and cloudiness, may have played an underappreciated role in Iran’s evolving water balance. Irrigation scheduling tools, drought indices, and crop water models calibrated around temperature sensitivity alone could be systematically misattributing the drivers of demand.

The study also offers a methodological lesson that extends well beyond Iran. Sensitivity analyses of the Penman–Monteith equation have been published for regions around the world, from the Yangtze basin to Australia’s climatic zones to tropical savannas, and their conclusions have often varied. Some of that variation is surely real, reflecting genuinely different climates. But this study demonstrates that part of the disagreement may be an artifact of method: analyses relying on simple correlations can rank drivers differently than analyses that properly account for multicollinearity. As climate change reshapes the individual variables at unequal speeds, knowing which lever actually moves evapotranspiration will determine whether water managers in arid lands anticipate the future accurately or are surprised by it. For Iran, at least, the answer now points upward, toward the sun itself.

Subject of Research: Spatiotemporal drivers of reference evapotranspiration in Iran over 58 years

Article Title: A multi-decadal spatiotemporal evaluation of reference evapotranspiration drivers in Iran: Leveraging an integrated LMG and Corrplot statistical framework

Article References: Zarei, A. R., & Mahmoudi, M. R. (2026). A multi-decadal spatiotemporal evaluation of reference evapotranspiration drivers in Iran: Leveraging an integrated LMG and Corrplot statistical framework. Theoretical and Applied Climatology, 157(11), Article 703. https://doi.org/10.1007/s00704-026-06635-8

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06635-8

Keywords: reference evapotranspiration, Iran, FAO Penman-Monteith, sunshine hours, LMG variance decomposition, multicollinearity, Corrplot, Inverse Distance Weighting, climate variables, water resources, irrigation planning, Theoretical and Applied Climatology

The research tackles a problem that has quietly plagued evapotranspiration modeling for decades. The gold standard for computing ETo, the FAO-56 Penman-Monteith equation, requires a full suite of meteorological measurements including air temperature, solar radiation, wind speed, and humidity. Many weather stations around the world, particularly in developing regions, simply do not record all of these variables reliably. Machine learning models have emerged as powerful alternatives because they can learn the complex, non-linear relationships between whatever meteorological inputs are available and the evapotranspiration rates observed at a site. Yet feeding a model every possible variable is not always wise: redundant or irrelevant inputs inflate model complexity, slow computation, and can degrade accuracy through noise. The question of which variables to keep, and which to discard, is known as feature selection, and it is precisely where the new study makes its mark.

The researchers built a hybrid framework that couples seven distinct feature selection methods with an Extreme Gradient Boosting model, better known as XGBoost. XGBoost was chosen deliberately for its proven track record with tabular meteorological datasets, its built-in regularization that helps prevent overfitting, its native ability to capture non-linear interactions between variables, and its computational efficiency on moderately sized station records. The seven selection techniques span the two major families of feature selection. Filter-based methods, which evaluate variables independently of any predictive model, included the Pearson correlation coefficient, the minimum redundancy-maximum relevance criterion known as mRMR, the ReliefF algorithm, the Fisher score, and the Stability method. Embedded methods, which perform selection as part of model training, included LASSO regression and SHAP, the SHapley Additive exPlanations approach borrowed from the field of interpretable machine learning.

Each of these techniques interrogates the data in a fundamentally different way. Pearson correlation measures simple linear association between each meteorological variable and ETo. The mRMR criterion, rooted in information theory, seeks variables that are highly relevant to the target while being minimally redundant with one another, avoiding the trap of selecting several measurements that all encode the same information. ReliefF evaluates how well a variable distinguishes between neighboring instances that are similar or dissimilar in the target value, making it sensitive to feature interactions. The Fisher score assesses class separability in the feature space, while the Stability method quantifies how consistently a selection algorithm picks the same variables across resampled datasets, a property that matters enormously when models must be trusted on new data. LASSO, by contrast, shrinks regression coefficients toward zero and effectively eliminates weak predictors through regularization, and SHAP derives feature importance from game-theoretic principles by computing each variable’s marginal contribution to model predictions.

To test the framework under genuinely different atmospheric regimes, the team used data from two contrasting climatic conditions, an arid region and a humid region, with records obtained from the Islamic Republic of Iran Meteorological Organization. The study was structured in two modeling stages. In the first stage, the three most important features identified by each selection technique were fed into XGBoost as inputs. In the second, more aggressive stage, only the two most important features were used. This staged design functions as a feature-based domain adaptation procedure, systematically evaluating the relevance and redundancy of key meteorological variables while minimizing the input space, which in turn reduces model complexity and computational cost. The practical payoff is a set of modeling tools suited to data-scarce conditions, where reliable measurements are limited and every sensor counts.

The results are striking. In the arid region, the best-performing combinations achieved a coefficient of determination of at least 0.980, a root mean square error no greater than 0.344 millimeters per day, and a mean absolute error of at most 0.270 millimeters per day. The humid region performed even better, with R-squared values of at least 0.982, RMSE as low as 0.188 millimeters per day, and MAE down to 0.142 millimeters per day. For a variable that typically ranges across several millimeters per day seasonally, sub-0.2 millimeter daily errors represent an extraordinary level of precision, comparable to the uncertainty of the reference equation itself.

Perhaps the most actionable finding is that the optimal feature selection method depends on climate. In the arid region, ReliefF and LASSO delivered the lowest error indices when only the two most important features were used. In the humid region, the Pearson, SHAP, Fisher, and Stability methods produced the best results with the same two-feature approach. This divergence makes physical sense. In arid environments, where atmospheric demand is driven largely by energy availability and vapor pressure gradients, instance-based and regularization-based selectors appear to isolate the dominant drivers most effectively. In humid conditions, where radiation and temperature interplay differently and air moisture is less limiting, correlation-based and interpretability-driven methods capture the signal more cleanly. The lesson for practitioners is that feature selection is not a one-size-fits-all exercise; the choice of selector should be matched to the climatic context of the station being modeled.

The authors did not stop at raw accuracy figures. They subjected their models to bootstrap analysis, effect size assessment, and decision boundary analysis to evaluate robustness and statistical significance rather than relying on single-point error metrics alone. Bootstrap resampling tests whether performance holds across repeated random draws from the data, guarding against lucky splits. Effect size analysis quantifies the magnitude of differences between competing models, ensuring that reported improvements are meaningful rather than trivially small. Decision boundary analysis probes how models behave across the full range of input conditions, revealing whether accuracy is uniform or concentrated in particular regimes. Together, these diagnostics lend the headline numbers a level of statistical credibility that many applied machine learning studies lack.

As a final benchmark, the XGBoost models were compared against Random Forest and a feed-forward Artificial Neural Network, two of the most widely used alternatives in hydrological machine learning. XGBoost came out on top, confirming that its gradient-boosted tree architecture, with its regularization and efficient handling of non-linear interactions, is particularly well suited to the structure of daily meteorological records. This finding adds to a growing body of evidence that gradient boosting is a formidable default choice for tabular environmental data, often outperforming deep networks on datasets of modest size.

The implications reach well beyond Iranian weather stations. As climate change intensifies pressure on freshwater supplies, water managers in arid and semi-arid regions worldwide face the twin challenges of rising crop water demand and sparse monitoring infrastructure. A modeling recipe that achieves R-squared values above 0.98 using only two meteorological inputs means that stations lacking full instrumentation can still produce trustworthy evapotranspiration estimates for irrigation scheduling and drought assessment. The framework’s emphasis on stability and interpretability, through methods like SHAP and the Stability criterion, also aligns with the broader push for transparent artificial intelligence in environmental science, where black-box predictions alone are no longer sufficient for policy decisions. By demonstrating that the marriage of thoughtful feature selection and gradient boosting can squeeze remarkable accuracy from minimal data, the study offers a practical template for the next generation of evapotranspiration models, one that could help stretch every drop of information as far as every drop of water.

Subject of Research: Machine learning-based feature selection for estimating daily reference evapotranspiration from meteorological data

Article Title: Assessing machine learning-based feature selection methods for estimating daily reference evapotranspiration using meteorological data

Article References: Sadeghzadeh, M., Karimi, S., Shiri, J., & Sivakumar, B. (2026). Assessing machine learning-based feature selection methods for estimating daily reference evapotranspiration using meteorological data. Theoretical and Applied Climatology, 157(10), Article 629. https://doi.org/10.1007/s00704-026-06568-2

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06568-2

Keywords: reference evapotranspiration, feature selection, XGBoost, machine learning, meteorological data, ReliefF, LASSO, SHAP, arid climate, humid climate, irrigation scheduling, Theoretical and Applied Climatology

Reference evapotranspiration, abbreviated ETo, represents the water loss from a standardized grass surface under given weather conditions. Because directly measuring evapotranspiration with lysimeters or micrometeorological towers is expensive and technically demanding, hydrologists and agronomists rely on ETo as a proxy for atmospheric demand. The internationally accepted way to compute it is the Penman–Monteith equation as standardized by FAO-56, a physically based formula that combines an energy term, driven by net radiation, with an aerodynamic term, driven by wind speed and the vapor pressure deficit between the surface and the air. The equation’s multivariate structure is its strength, but it also raises a critical question: which of these meteorological ingredients matters most, and does the answer change from place to place and season to season?

That question motivated a team led by Lucas da Costa Santos of the Federal University of the Jequitinhonha and Mucuri Valleys. The researchers assembled thirty years of daily meteorological records, from 1990 to 2019, for six Brazilian municipalities chosen to represent sharply contrasting climate typologies under the Köppen–Geiger classification: Belém in the Amazon with an equatorial monsoon climate, Salvador with a tropical rainforest regime, Petrolina in the hot semi-arid northeast, Pirenópolis with a tropical savanna climate, and Franca and São Luiz Gonzaga in the humid subtropical south. The sites span a remarkable range, from mean annual temperatures of about 21.4 degrees Celsius in São Luiz Gonzaga to 27.7 degrees in Belém, and from mean solar radiation of 16.69 megajoules per square meter per day in the subtropical south to 23.19 in Petrolina.

To disentangle the influences, the team turned to path analysis, a statistical technique that decomposes simple correlations between weather variables and ETo into direct and indirect effects. Direct effects capture the independent contribution of each variable, while indirect effects reveal how variables interact with one another through their mutual correlations. The researchers first checked for multicollinearity among predictors using variance inflation factors, retaining mean air temperature as the thermal variable because maximum and minimum temperatures introduced redundancy that destabilized the regression models. They also ran Mann–Kendall trend tests to check whether the predictor variables themselves were changing over the study period, finding significant long-term trends in at least one variable at five of the six locations.

The results were striking in their consistency and their variability at the same time. The path models explained between 97.7 and 99.5 percent of the annual variability in ETo across all six locations, an extraordinarily high proportion. Yet the identity of the dominant driver shifted from site to site. In São Luiz Gonzaga, Franca, Salvador, and Belém, solar radiation carried the largest direct effect, with standardized coefficients reaching 0.812 in Franca, indicating that in these places evaporative demand is primarily limited by the energy available at the surface. In Pirenópolis and Petrolina, by contrast, wind speed took the lead, with a coefficient of 0.711 in the semi-arid Petrolina, pointing to a regime in which the aerodynamic removal of vapor, rather than energy supply, sets the pace of evaporation.

The decomposition of indirect effects added a layer of nuance that simple correlation studies miss entirely. In Franca, the seemingly dominant direct effects of radiation and wind were substantially offset by negative indirect pathways, cutting their total effects to a fraction of their direct coefficients. In humid Belém, the positive direct effect of wind speed was overwhelmed by a strong negative indirect effect mediated through solar radiation, so that wind’s net association with ETo actually turned negative. In Petrolina, the opposite occurred: positive indirect effects reinforced the direct dominance of wind, pushing its total effect to 0.837. In other words, the same meteorological variable can be amplified, muted, or even reversed in its apparent influence depending on the interaction structure of the local atmosphere.

Seasonality compounded the picture. When the team repeated the analysis for each meteorological season of the Southern Hemisphere, the hierarchy of controls reconfigured systematically through the year. Solar radiation ranked among the top drivers during the high-insolation summer and spring months, while wind speed gained prominence during autumn and especially winter, when energy availability declines but atmospheric ventilation remains strong, particularly at inland and semi-arid sites. Relative humidity, meanwhile, displayed negative direct effects at every location and in every season, a physically coherent signal of the vapor pressure deficit that governs the aerodynamic term of the Penman–Monteith equation. Drier air pulls water from the surface more aggressively, and the humidity coefficients, reaching −0.469 in São Luiz Gonzaga and −0.374 in Petrolina, quantify that pull.

Perhaps the most consequential finding concerns what did not dominate: temperature. Mean air temperature ranked among the three strongest drivers at only half of the locations at the annual scale, and even there its coefficients were systematically smaller than those of radiation or wind. This challenges the widespread reliance on temperature-based empirical methods such as Hargreaves–Samani, which implicitly assume that thermal proxies can capture evaporative demand. The authors argue that temperature functions mainly as a thermodynamic state variable, signaling the energy content of the atmosphere, while the actual fluxes of vapor depend on radiation balance, humidity gradients, and wind. In regions with strong ventilation or high radiative variability, temperature-only models risk systematic underestimation of water demand, with tangible consequences for irrigation scheduling and water resource planning.

The practical implications extend into a warming future. As global temperatures rise, the atmosphere’s water-holding capacity increases, and vapor pressure deficit tends to grow disproportionately, intensifying evaporative demand in ways that temperature records alone do not reveal. The study’s authors are careful to note the limitations of their approach: ETo is not an independently observed quantity but a deterministic product of the same variables used as predictors, so the path coefficients should be read as indicators of relative contributions within the FAO-56 framework rather than as evidence of independent causation. Solar radiation was estimated from sunshine duration rather than measured directly, each municipality was represented by a single weather station, and the significant long-term trends in several predictors mean the reported hierarchies describe dominant tendencies over the study period rather than fixed relationships.

Even with those caveats, the diagnosis is clear and actionable. There is no universal driver of evaporative demand; the balance between radiative and aerodynamic control is a property of each climate and each season. For water managers in aerodynamically dominated environments like Petrolina, monitoring wind speed and humidity is essential, and neglecting them can mean underestimating irrigation requirements during dry periods. In radiatively controlled regions such as Belém, Salvador, Franca, and São Luiz Gonzaga, cloud cover and seasonal insolation set the agenda, and models that ignore radiation do so at their peril. The study makes a strong case that the era of one-size-fits-all, temperature-based evaporation estimates should give way to multivariate, physically consistent approaches capable of tracking how the atmosphere’s thirst is truly regulated across a changing planet.

Subject of Research: Meteorological controls on reference evapotranspiration across contrasting Brazilian climate typologies

Article Title: Energy versus aerodynamic controls of reference evapotranspiration: a multi-regional assessment across Brazilian climate typologies

Article References: da Costa Santos, L., Capuchinho, F. F., do Patrocínio Figueiró, L. S., José, J. V., dos Reis, E. F., Bonfá, C. S., Araujo, J. E., & Nhongo, E. J. S. (2026). Energy versus aerodynamic controls of reference evapotranspiration: a multi-regional assessment across Brazilian climate typologies. Theoretical and Applied Climatology, 157(10), Article 646. https://doi.org/10.1007/s00704-026-06580-6

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06580-6

Keywords: reference evapotranspiration, Penman–Monteith, solar radiation, wind speed, relative humidity, vapor pressure deficit, path analysis, Brazil, climate typologies, irrigation management, FAO-56, hydrology

Reference evapotranspiration, usually abbreviated ET0, describes how much water would evaporate from a standardized reference surface and be transpired by an idealized crop under prevailing weather conditions. It is the backbone of irrigation planning: multiply ET0 by a crop-specific coefficient and you get the water demand of a real field. For decades, the gold-standard method has been the Penman-Monteith equation endorsed by the Food and Agriculture Organization, which combines temperature, humidity, wind speed and solar radiation measured at weather stations. The catch is that weather stations are unevenly distributed and nearly absent in mountains, deserts and other data-scarce regions where water management decisions are often most urgent. Traditional models therefore leave large gaps in our picture of where and how fast the atmosphere’s thirst for water is growing.

The Chinese team’s approach replaces some of that missing meteorological machinery with fluorescence from space. Earlier work had shown that SIF could be used to estimate ET0 in the arid and semi-arid regions of northern China, but the method’s reliability in humid southern climates and its ability to capture seasonal swings remained open questions. Zhang and colleagues refined the existing SIF-based model, which they call RET0-SIF, and validated it across multiple climatic zones spanning the country, from the deserts of Xinjiang to the monsoon-soaked far south. The refinement was designed specifically to improve the model’s handling of seasonal variation, a persistent weakness of earlier fluorescence-based evapotranspiration estimates.

The performance numbers are striking. At the national scale, the ET0 values estimated from satellite fluorescence showed high consistency with those computed by the Penman-Monteith equation, with most validation sites achieving a coefficient of determination above 0.8 and root-mean-square errors below 7 millimeters per 8-day period. In practical terms, the glowing signal from vegetation reproduced the water-demand estimates of a dense weather-station network across climates as different as alpine grassland and tropical forest. The researchers then used the improved model to generate spatially explicit maps of ET0 distribution across China and checked them against existing evapotranspiration products at several spatial scales, finding that the fluorescence-based maps held their own against established datasets.

With a validated two-decade map in hand, the team turned to the question of change. Between 2000 and 2019, overall reference evapotranspiration in China showed an upward trend, meaning the atmosphere’s evaporative demand has been intensifying as the climate warms. The increases were most significant in the North China Plain, the country’s agricultural heartland, and in the Mount Qinling region, a major east-west mountain barrier that shapes the boundary between China’s northern and southern climate regimes. Only 8.6 percent of the country showed significant decreases, concentrated in the Loess Plateau and the mountains of southern China. For a nation already juggling severe water scarcity in the north with flood-prone surplus in the south, a broadening gap in evaporative demand is a consequential signal.

The attribution analysis is where the study delivers its most surprising verdict. Using statistical methods capable of separating dominant, direct and indirect effects, the researchers found that changes in relative humidity explained 39.2 percent of the variation in ET0 across China, edging out temperature, which accounted for 35.6 percent. This is counterintuitive at first glance: public discussion of climate-driven water stress usually centers on rising temperatures. But the physics is straightforward. The drier the air, the steeper the vapor pressure gradient between the moist interior of a leaf and the surrounding atmosphere, and the faster water escapes. Declining relative humidity, a common companion of warming, acts as a direct throttle opening on evaporation, and in China over the past two decades it has been the single largest lever.

The picture shifts subtly when the analysis moves to the interannual scale. There, temperature and radiation, together with their correlations with ET0, dominate the variability, while relative humidity and wind speed exert mainly negative impacts, meaning that when these factors rise, evapotranspiration tends to fall. In other words, year-to-year swings in China’s water demand are choreographed primarily by the energy side of the energy-water balance, whereas the long-term trend is increasingly governed by how thirsty the air itself has become. This kind of scale-dependent attribution matters for forecasters: a model tuned to explain interannual variability may still misjudge the trajectory of the underlying trend, and vice versa.

What makes the study technically notable is the way fluorescence bridges two sides of a fundamental coupling. Photosynthesis and transpiration are physically linked through stomata, the adjustable pores on leaf surfaces through which plants take in carbon dioxide and lose water vapor. When stomata open to feed photosynthesis, water escapes; when the plant closes them under drought stress, both carbon uptake and fluorescence decline together. SIF therefore acts as a proxy for the biological conductance of the land surface, information that purely meteorological models must infer indirectly. By anchoring ET0 estimation in an observable signal of plant-atmosphere exchange, the RET0-SIF framework sidesteps the station-coverage problem while remaining grounded in the physical process it is trying to measure.

The implications extend well beyond China’s borders. Satellite missions carrying fluorescence-capable spectrometers now provide global coverage at increasingly fine spatial and temporal resolutions, and the strategy demonstrated here, refining a fluorescence-based model across diverse climate zones before deploying it for national-scale mapping, is directly transferable to other data-sparse regions of Central Asia, Africa and South America. For agricultural water managers, the study offers a way to forecast irrigation demand in places where weather stations are scarce but satellite data are abundant. For climate scientists, it adds an independent line of evidence on how evaporative demand is evolving under warming, one rooted in the physiology of living plants rather than in equations driven entirely by atmospheric measurements.

There are also cautionary notes embedded in the findings. The significant upward trends in the North China Plain, a region that already draws heavily on depleted groundwater aquifers, suggest that crop water requirements will keep climbing even where precipitation holds steady, tightening the squeeze on one of the world’s most intensively farmed landscapes. Meanwhile, the modest areas of significant decline in the Loess Plateau and southern mountains hint at regionally divergent trajectories that national averages conceal. As the authors note, the work provides valuable insights for large-scale ET0 estimation using SIF and supports water resource management decisions. In an era when every drop of irrigation water must be accounted for, the faint infrared glow of chlorophyll is proving to be more than a botanical curiosity; it is becoming a working instrument for measuring the planet’s thirst.

Subject of Research: Satellite solar-induced chlorophyll fluorescence estimation of reference evapotranspiration and its climatic drivers in China

Article Title: Spatiotemporal patterns and climatic drivers of reference evapotranspiration in China revealed by solar-induced chlorophyll fluorescence

Article References: Zhang, S., Zheng, J., Han, C., Liu, L., An, J., & Pei, W. (2026). Spatiotemporal patterns and climatic drivers of reference evapotranspiration in China revealed by solar-induced chlorophyll fluorescence. Theoretical and Applied Climatology, 157(10), Article 641. https://doi.org/10.1007/s00704-026-06566-4

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06566-4

Keywords: solar-induced chlorophyll fluorescence, reference evapotranspiration, China, climate change, remote sensing, Penman-Monteith, relative humidity, water resources, irrigation, North China Plain, drought, agricultural meteorology

Atmospheric evaporative demand sits at the hinge of the terrestrial water balance. It governs the moisture exchange between land and air, determines how much of a given rainfall episode is lost back to the sky, and in water-limited regions it is tightly coupled to the onset and intensification of drought. When demand rises faster than supply, soils dry, reservoirs shrink and vegetation stress compounds even if precipitation itself does not collapse. Yet despite this central role, the multi-decadal behaviour of ET0 and its structural changes have remained poorly characterised, particularly in semi-arid regions where the question of whether atmospheric conditions are statistically associated with their impact on water balance variability has not been systematically answered. The Tamil Nadu study set out to close that gap by examining long-term hydroclimatological behaviour through four lenses: spatial heterogeneity, temporal non-stationarity, atmospheric control and impacts on the climatic water balance across contrasting hydro-climatic regimes.

The evidence base was drawn from the NASA POWER database, a publicly accessible satellite-derived and modelled meteorological record that provided temperature, relative humidity, solar radiation, wind speed and precipitation for the period 1985 to 2024. Rather than relying on a single station, the researchers selected representative stations spanning the diverse hydro-climatic regimes of Tamil Nadu, from the wetter coastal and western zones to the drier interior plains. This design matters because a state-level average can easily mask the fact that neighbouring districts may be drifting in opposite directions. By treating each regime on its own terms, the analysis could reveal whether the drivers of evaporative demand behave uniformly or whether their dominance changes with the local climate setting.

Methodologically, the study leaned on a battery of robust non-parametric techniques chosen for their resilience to outliers and to the non-normal distributions typical of hydro-climatic data. Trend detection used the Mann–Kendall test paired with Sen’s slope estimator, a combination that has become a standard for identifying monotonic change in environmental series. The pivotal innovation, however, was the Pettitt test for regime shift analysis, which searches for a single abrupt change point in a time series rather than assuming gradual change. Extreme event analysis characterised the behaviour of ET0 at the tails of the distribution, where agricultural stress concentrates, and a SARIMA model, a seasonal autoregressive integrated moving average framework, was used to forecast short-term ET0 dynamics. Finally, a driver dominance analysis quantified the relative contribution of individual atmospheric factors to ET0 variability, allowing the team to rank the controls rather than merely list correlations.

The headline finding is that ET0 in Tamil Nadu does not follow a simple, monotonic trajectory. Instead, the records show statistically significant regime changes, meaning that the series jumps between quasi-stable states with different mean levels and different atmospheric controls. The interaction between temperature, radiation and humidity emerges as the governing mechanism, and the balance of power among these three variables shifts from one hydro-climatic regime to another. In practical terms, this means that a warming trend does not translate into a uniform increase in atmospheric thirst everywhere. Where humidity is high, rising temperatures may be partially offset by the suppression of evaporation; where the air is already dry, the same warming can push demand sharply upward. The study’s driver dominance analysis makes this regime dependence explicit, showing that the atmospheric control on ET0 varies systematically under different hydro-climatic conditions.

Extreme ET0 behaviour and the response of the climatic water balance also diverged across regimes, a pattern the authors interpret as the spatial fingerprint of climatic water stress. In the drier interior, where rainfall is marginal and evaporative demand is chronically high, shifts in ET0 translate almost directly into deeper water deficits, because there is little buffer between supply and demand. In wetter zones, the same shifts may be absorbed by soil moisture and surface storage, at least temporarily. This asymmetry has a sobering implication: identical large-scale climate signals can produce radically different drought outcomes depending on where they land. Water planners who rely on state-wide or basin-wide averages risk misjudging both the severity and the geography of emerging stress.

Perhaps the most consequential conclusion of the paper is a methodological warning. Atmospheric evaporative demand, the authors argue, cannot be understood on the basis of monotonic trends alone, because its spatial heterogeneity, temporal non-stationarity and regime dependence violate the assumptions underlying simple trend analysis. A Mann–Kendall test applied to a series that has jumped between two regimes may report a significant trend that is really an artefact of a step change, or conversely may miss genuine change that is concentrated in a short transition. For predictability, this reframing is critical. Forecast systems and drought early-warning schemes that extrapolate a linear trend will systematically misjudge the future if the underlying process is one of regime dynamics. Recognising the regime structure, by contrast, opens the door to forecasts conditioned on the current state, which is precisely what the SARIMA component of the study begins to explore for short-term ET0 dynamics.

The findings arrive amid a broader scientific reassessment of global evaporative demand. Recent work has documented that climate change has increased evaporative demand across most of the planet, with South Asia standing out as a notable exception in some global assessments, and researchers have begun naming prolonged episodes of extreme atmospheric demand, such as so-called thirstwaves, as a distinct class of agricultural hazard. The Tamil Nadu results add regional texture to this global picture, showing that even within a single Indian state the trajectory of atmospheric thirst is not uniform. They also echo a long-standing puzzle in hydrology, the so-called evaporation paradox, in which observed pan evaporation can decline even as temperatures rise, a reminder that humidity, radiation and wind interact with temperature in ways that defy single-variable intuition.

For a state where agriculture consumes the bulk of freshwater and where monsoon failures routinely trigger drinking-water emergencies, the practical stakes are considerable. Knowing which atmospheric variable dominates ET0 in each regime tells irrigation authorities what to watch: humidity and radiation in some zones, temperature in others. Knowing that regimes shift abruptly rather than drift smoothly suggests that water budgets should be revised at detected change points rather than on fixed assumptions. The availability of the underlying NASA POWER data, covering 1985 to 2024, and of the processed analytical outputs on reasonable request from the corresponding author, means the framework can be replicated and extended to other semi-arid regions facing similar questions. The research was supported by the Tamil Nadu Chief Minister’s Research Grant, reflecting state-level interest in the predictability problem.

Ultimately, the study reframes atmospheric evaporative demand from a passive consequence of warming into an active, regime-governed component of the climate system with its own predictability structure. In semi-arid Tamil Nadu, the atmosphere’s thirst arrives in steps, not slopes, and each step rewrites the local water balance in a different way. Capturing that step-like behaviour, the authors contend, is the key to anticipating drought before it takes hold, and to managing a water future in which the demand side of the equation may change faster than the supply side ever will.

Subject of Research: Multi-decadal variability and regime shifts in atmospheric evaporative demand and climatic water balance in Tamil Nadu, India

Article Title: Multi-decadal variability and regime dynamics of atmosphere evaporative demand and climatic water balance in Tamil Nadu, India: implications for predictability

Article References: Multi-decadal variability and regime dynamics of atmosphere evaporative demand and climatic water balance in Tamil Nadu, India: implications for predictability. (n.d.). https://doi.org/10.1007/s00704-026-06592-2

Image Credits: AI Generated

DOI: 10.1007/s00704-026-06592-2

Keywords: evaporative demand, reference evapotranspiration, ET0, Tamil Nadu, drought, regime shift, Pettitt test, Mann-Kendall trend analysis, SARIMA forecasting, climatic water balance, semi-arid climate, NASA POWER