Search Results (143)

Incomplete station meteorological records constrain sustainable water-demand planning and climate-adaptation decisions in data-limited regions. Open-data reference evapotranspiration (ETo) products can provide continuous hydroclimatic information for atmospheric-demand screening, but they do not directly represent realized water demand and their use as station substitutes may introduce bias because gridded and station-derived ETo represent related but non-identical calculation and spatial domains. This study develops a station-anchored open-data ETo screening workflow for assessing local product compatibility under incomplete meteorological records. The workflow combines four official Turkish State Meteorological Service (MGM) stations in the eastern Marmara region, station-derived Hargreaves–Samani (HS) ETo for 1990–2020, TerraClimate v1.1 reference ETo, calendar-month and anomaly diagnostics, blocked historical gap-transfer tests in which omitted periods are excluded from correction fitting and tuning, and measured-radiation FAO-56 Penman–Monteith (PM) sensitivity subsets. TerraClimate reproduced the first-order annual ETo cycle strongly (r = 0.972–0.978), but mean TerraClimate-minus-HS bias remained station-dependent (+1.78 to +11.55 mm month⁻¹) and strongly calendar-month-dependent. After removing each series’ calendar-month climatology, Pearson correlation decreased to 0.559–0.849, indicating that the common annual cycle explains a substantial part of the raw agreement. Train-only monthly-bias and nested L2 residual corrections reduced pooled blocked out-of-sample MAE, although no correction method was universally superior across stations. Measured-radiation PM subsets at Kocaeli and Yalova showed PM-minus-HS mean offsets of −4.42% and +3.37%, respectively. The blocked tests evaluate transfer to omitted historical periods rather than prospective forecasting. The results support local, season-aware compatibility screening before gridded ETo is used for climate-resilient water-resource planning; they should not be interpreted as validation of actual ET, crop water use, irrigation withdrawals, or realized water demand. Full article

Freshwater scarcity can weaken water governance when hydrological pressure interacts with intensive agricultural demand, regulatory weakness, and political conflict. This research evaluates whether Irrigation District 005 (IR 005) in Chihuahua, northern Mexico, demonstrates local water-governance fragility across three domains: public security, the rule of law, and the ability to sustain water access and food production. A mixed-methods approach integrates legal and human rights documentation, institutional records, published studies, and a structured media review with hydrological, agricultural, climatic, and reservoir data. Water balances were analysed for 1998–2023, precipitation trends for 1980–2020, and crop water requirements were estimated using the Food and Agriculture Organization’s Irrigation and Drainage Paper No. 56 (FAO-56) Penman–Monteith framework, the crop coefficient (Kc), the water-stress coefficient (Ks), the United States Soil Conservation Service (SCS) Curve Number method, and application-efficiency assumptions. The 2020 water conflict resulted in fatalities, injuries, arrests, and documented human rights violations. Rule-of-law capacity was further diminished by unauthorised withdrawals, cultivation beyond authorised irrigation plans, and limited enforcement. The annual water balance shifted to persistent deficits after 2016, reaching an estimated deficit of 2268 cubic hectometres (hm³) in 2020. Annual precipitation did not exhibit a statistically significant monotonic decline during 1980–2020 (Mann–Kendall Z = −0.79, τ = −0.0878, p = 0.4251; Sen’s slope = −1.1628 mm yr⁻¹; Mann–Whitney p = 0.5313), indicating that recent stress is more closely linked to production scale, crop mix, governance conditions, and irrigation efficiency than to a long-term reduction in rainfall. Sensitivity analysis revealed that ±15% changes in Kc and Ks altered gross water requirements by approximately ±16–17%, while equivalent changes in effective precipitation produced changes of only 1–3%. These results demonstrate heightened water-governance fragility resulting from mutually reinforcing hydrological, institutional, and conflict-related pressures. Future research should refine locally calibrated water-demand parameters and develop reproducible monitoring systems that combine hydrological, institutional, satellite, and participatory data to support anticipatory, transparent, and rights-based water governance. Full article

Water resources are essential for sustainable agricultural development, and accurate crop water requirement prediction is important for improving irrigation efficiency and optimizing water allocation in irrigation districts. This study focused on the Chushandian Irrigation District in Henan Province, China. Reference evapotranspiration (ET₀) was calculated using the FAO Penman–Monteith equation, and the monthly crop water requirements (ET_C) of wheat, peanut, rapeseed, corn, rice, and vegetables were estimated using crop coefficients (K_c). XGBoost feature importance, Pearson correlation, Mantel, and SHAP analyses were used to examine the meteorological drivers of crop water requirement. Atmospheric pressure showed high nonlinear predictive importance, whereas mean air temperature, relative humidity, and sunshine duration exhibited more consistent physical and statistical relationships with crop water requirement. A process-informed TCN–Transformer framework was then developed for joint and crop-specific prediction. The TCN module extracted local temporal variations, while the Transformer module captured long-term dependencies. In the joint prediction task, the proposed model achieved an R² of 0.9487 and an RMSE of 33.24 mm, outperforming the LSTM, GRU, and CNN–LSTM baselines. The crop-specific results further demonstrated that the model effectively represented seasonal variations and periods of relatively high water requirement across the six crops. The proposed framework can support monthly water-allocation planning and seasonal irrigation scheduling in multi-cropping irrigation districts. Full article

The Penman–Monteith model is the standard method for computing reference evapotranspiration (ET_o–PM), requiring full meteorological data including solar radiation (R_s), actual vapor pressure (e_a), and wind speed (u₂). However, not all of these data are always available. This study incorporated temperature-based methods recently developed for estimating R_s (R_{s_dev}) and e_a (e_{a_dev}) into the ET_o–PM model (PMT) and compared them with the alternative procedures developed by Paredes and Pereira for estimating R_s (R_{s_PP}) and e_a (e_{a_PP}). In cases where u₂ is missing, a global mean value of 2 m s⁻¹ (u_{2_def}) or a local long-term average (u_{2_ave}) was employed. All models were evaluated based on the comparison with ET_o–PM, calculated with complete data from 1993 to 2016 at 96 radiation stations. The results showed that using R_{s_dev}, e_{a_dev}, or u_{2_ave} (compared to R_{s_PP}, e_{a_PP}, or u_{2_def}) in ET_o–PM estimation was associated with better agreement with the full-data benchmark within the evaluated dataset when one or two variables are missing, with RMSE decreasing by 7–51%, although the improvements were less pronounced in arid regions. When all three variables are unavailable, combining R_{s_dev}, e_{a_dev}, and u_{2_ave} yields comparable or better agreement with the full-data benchmark without requiring additional humidity data. Overall, the updated PMT framework provides a practical pathway for reproducing FAO-56 Penman–Monteith ET_o under incomplete meteorological conditions. Full article

Coordinated management of drip irrigation water demand and lateral canal supply is a critical strategy for improving water use efficiency in arid irrigation districts; however, under water-deficit conditions, the efficient and equitable allocation of limited canal water among multiple drip irrigation systems remains largely unresolved. This study developed a bi-objective cooperative water allocation and scheduling model for a lateral canal serving 11 subordinate drip irrigation systems. Subject to canal diversion flow balance and total deficit constraints, the model simultaneously minimized (i) the mean coefficient of variation (CV) of water allocation duration within rotation irrigation groups, targeting temporal uniformity, and (ii) the sum of squared deviations of the water supply satisfaction rate across systems, targeting distributional equity. Water demand inputs were derived from a localized FAO-56 Penman–Monteith irrigation schedule for Jinghe County with stage-specific crop coefficients. A hybrid binary–continuous NSGA-II encoding with a dynamic intra-group flow allocation mechanism was employed. For the baseline deficit scenario (Early May, supply-to-demand ratio β = 67.76%), the model partitioned the 11 systems into four rotation groups with a mean CV of 3.15 × 10⁻³, while the sum of squared deviations of the satisfaction rate decreased from 4.366 under the empirical scheme to 1.50 × 10⁻⁴, confining all systems to 67.38–68.45% and eliminating the coexistence of over-supply and complete deprivation (Wilcoxon signed-rank test, p < 0.001; Cohen’s d = −1.698). The Pareto front revealed a significant efficiency–equity trade-off (Spearman’s ρ = −0.9999), and NSGA-II outperformed SPEA2 by approximately 29-fold and 22-fold in the two objectives. Robustness was confirmed across three deficit scenarios and algorithm parameter sensitivity analyses (CV < 2%). The study offers methodological support for refined water allocation management of terminal canal systems in arid regions. Full article

Reference evapotranspiration (ET_o) is essential for irrigation scheduling, water resources management, and climate-related applications, but the FAO Penman–Monteith (FAO-PM) method is often constrained by limited meteorological data availability. This study evaluated four machine learning (ML) algorithms, Kernel Approximation Regression (KAR), Multilayer Perceptron (MLP), Extreme Gradient Boosting (XGB), and Random Forest (RF), for reproducing FAO-PM ET_o under different levels of geographical and meteorological information availability in the Czech Republic. Daily observations from 59 meteorological stations (1980–2024) were used to develop eight input scenarios. Model performance was evaluated using a station-wise chronological train–test framework and station-based analyses. The results showed that predictor availability had a greater influence on model performance than model selection. The geographical-information scenario produced the lowest performance, whereas substantial improvements were achieved when meteorological variables were incorporated. Among the single-variable meteorological scenarios, relative humidity provided the greatest improvement in agreement with the FAO-PM ET_o benchmark. Across all input scenarios and ML algorithms, testing performance ranged from R² = 0.683 to 0.998 and RMSE = 0.076 to 0.939 mm d⁻¹, indicating progressively improved agreement with FAO-PM ET_o as additional meteorological information became available. The reduced-input scenarios therefore provide a practical approach for approximating FAO-PM ET_o at stations represented during model development when some meteorological inputs are unavailable. Full article

This study investigates hydroclimatic variability and water balance dynamics in the Akmola region during 2003–2023 using observations from 16 meteorological stations. The study evaluates changes in air temperature, precipitation, reference evapotranspiration (ET₀), climatic water balance, and drought conditions. Reference evapotranspiration was calculated using the FAO-56 Penman–Monteith method, while drought variability was assessed using the 12-month Standardized Precipitation–Evapotranspiration Index (SPEI-12). Temporal trends were analyzed using the Mann–Kendall test and Sen’s slope estimator. The results revealed significant spatial heterogeneity in hydroclimatic conditions across the region. Air temperature showed a consistent increasing trend at most stations, accompanied by increasing atmospheric evaporative demand. All stations were characterized by a persistently negative climatic water balance, with mean annual values of approximately −750 mm, indicating a regional moisture deficit. Reference evapotranspiration exhibited significant spatial variability, with the highest values observed in the southern and central parts of the region. Precipitation remained the dominant control of water balance variability (r = 0.79–0.96), while increasing temperature intensified moisture deficits through increased evapotranspiration. SPEI-12 indicated recurrent drought episodes and increasing drought vulnerability associated with warming-induced atmospheric water demand. The study demonstrates that increasing evapotranspiration is becoming a major driver of water stress in the Akmola region and highlights the importance of integrating climatic water balance and SPEI indicators for drought monitoring and climate adaptation in semi-arid steppe regions. Full article

Reference evapotranspiration (ET₀) is essential for irrigation management, but its spatial estimation is limited by sparse meteorological observations. Satellite products such as MOD16 potential evapotranspiration (PET) provide spatial continuity, yet they are not directly comparable to FAO-56 ET₀ due to structural differences in model parameterization. This study proposes a topography-constrained framework to convert MOD16 PET into FAO-consistent ET₀. The approach was tested in two heterogeneous regions of northern Italy (Piemonte and Veneto) using ground-based ET₀ derived from the FAO Penman–Monteith equation (2010–2022). PET–ET₀ transformation coefficients, estimated via station-wise linear regression, showed no significant temporal drift over the study period and strong spatial structure. Among the tested topographic predictors, elevation was retained as the main topographic proxy for modelling the spatial variability of the correction coefficients. The locally calibrated correction reduced the systematic overestimation of raw MOD16 PET and improved agreement with station-based ET₀ in both regions. Its performance was comparable to an IDW interpolation benchmark, although IDW slightly outperformed the topography-based model in Piemonte. A cross-region test showed that the correction reduced MOD16 PET errors when transferred between Piemonte and Veneto, but residual bias remained. The proposed framework should therefore be interpreted as a parsimonious regional topographic correction approach that requires local calibration and validation before application to other areas. Full article

Accurate crop evapotranspiration (ET_c) estimation at high spatial and temporal resolution remains a major challenge for precision irrigation. This study presents a multi-sensor data fusion framework combining daily MODIS (250 m), Sentinel-2 (10 m), and Landsat 8/9 (30 m) imagery with FAO-56 Penman–Monteith reference evapotranspiration (ET₀) to generate pixel-wise daily ET_c maps for irrigated maize (Zea mays L.) near Nyírbátor, Hungary, over six growing seasons (2020–2025). The proposed Median Time Series Model exploits field-scale MODIS NDVI as a temporal backbone and derives pixel-wise linear transfer functions to reconstruct daily NDVI at 10–30 m resolution. Three gap-filling strategies were compared; the median approach yielded the highest agreement (NDVI reconstruction R² = 0.81; RMSE = 0.19 (NDVI units); pixel-wise correlation 0.70–0.85) and effectively suppressed sub-pixel spectral mixture artefacts. Sentinel-2 consistently outperformed Landsat 8/9 (pixel-wise R² = 0.36–0.78 vs. 0.001–0.91). A nonlinear power crop coefficient model (K_c = a · NDVI^b) proved more robust than linear rescaling (mean validation R² of 0.80 (power) vs. 0.71 (rescale) across Sentinel-2 seasons; both methods were positive in all six seasons after correcting an unconstrained-fit artefact). Seasonal ET_c ranged from 313 to 545 mm, with cumulative water deficits reaching −334 mm during the 2021 drought. Six-year mean seasonal ET_c (428–483 mm for Sentinel-2) falls within the 400–600 mm range published for irrigated maize under comparable continental conditions, with season-integrated ET_c/ET₀ ratios (rescale method mean 0.86; power method mean 0.84) consistent with expected FAO-56 K_c trajectories. Cross-validation against an independent MATLAB implementation confirmed algorithmic consistency (reference ET₀ (R² = 0.88–0.91, Pearson r = 0.97–1.00)) and daily ET_c while identifying meteorological input as the dominant source of absolute ET_c uncertainty (estimated at ±15–30% through first-order error propagation). Plausibility assessment was limited to comparison with published seasonal benchmarks and an independent algorithmic implementation; no eddy covariance or lysimeter measurements were available for direct ET_c validation. Full article

This study evaluates the spatial variability of hydric stress in the State of Sinaloa, northwestern Mexico, through the integrated analysis of hydroclimatic variables, multispectral remote sensing indicators, and environmental factors. Historical hydroclimatic conditions were analyzed using meteorological records from 1961 to 2020, whereas Landsat 8 imagery acquired on 7 July 2025, was used to evaluate the spatial expression of hydric stress. Reference evapotranspiration (ETo) was estimated using the FAO-56 Penman–Monteith methodology, and hydrological deficit conditions were determined from the relationship between precipitation (P) and ETo. Spectral indicators including land surface temperature ($\overline{T}$_a), the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), and the NDWI/MNDWI relationship were used to evaluate vegetation response, surface moisture conditions, and thermal anomalies associated with hydric stress. The results revealed persistent conditions where ETo systematically exceeded P, with hydrological deficit values ranging from approximately −1600 mm·year⁻¹ to localized positive values near 50 mm·year⁻¹. The most severe deficits were concentrated within the northwestern and north-central agricultural valleys of Sinaloa. Statistical validation revealed significant negative relationships between hydrological deficit and all evaluated spectral indicators. The strongest association was observed for MNDWI (R² = 0.387), followed by NDWI/MNDWI (R² = 0.277), NDWI (R² = 0.220), and NDVI (R² = 0.134), confirming the sensitivity of vegetation and moisture-related indicators to long-term hydrological stress conditions. Spatial analyses revealed a strong correspondence among low NDVI, negative NDWI and MNDWI responses, elevated $\overline{T}$_a, and regions characterized by high atmospheric evaporative demand. Additional spatial validation integrating land-use and vegetation-cover changes (1993–2011), regional geology, topography, and the distribution of highly productive agricultural valleys demonstrated that the most severe hydrological deficits coincided with areas affected by vegetation-cover loss, agricultural expansion, and intensive land use. Although these datasets correspond to different observation periods, they collectively reflect the cumulative environmental effects associated with persistent hydrological stress across the region. The combined effects of hydrological imbalance, forest-cover reduction, and agricultural intensification have progressively reduced ecosystem resilience and increased environmental vulnerability throughout one of the most productive agricultural regions of northwestern Mexico. These findings provide a scientific basis for water-resource management, territorial planning, ecosystem restoration, and climate-adaptation strategies under increasing water-scarcity conditions. Full article

Field-based phenotyping of water-related traits for precision irrigation in tropical agroecosystems poses a persistent methodological challenge, driven by high climatic variability and the complex water-use physiology of Crassulacean Acid Metabolism (CAM) crops such as pineapple (Ananas comosus var. MD2). We developed and validated a Physics-Informed Machine Learning (PIML) framework that integrates high-resolution UAV multispectral imagery, IoT-based microclimatic records, and a mechanistic soil water balance based on the FAO-56 Penman–Monteith standard to predict plot-scale soil moisture depletion as a proxy of plant water status. A six-month field campaign (March–August 2022) across 25 georeferenced commercial pineapple plots in the Colombian Orinoquia piedmont yielded a spatiotemporally balanced dataset of $N=150$ observations. Soil-adjusted vegetation indices (OSAVI, MSAVI) outperformed standard NDVI for capturing water-related canopy traits, effectively decoupling spectral responses from substrate noise. A Gradient Boosting regressor achieved $R^2=0.842$ and $RMSE=0.0705$ on a normalized target scale, corresponding to a $7.05\%$ error over the prediction range, while the traffic-light Decision Support System (DSS) for irrigation scheduling reached $91.1\%$ accuracy (Cohen’s Kappa $=0.91$). Incorporating daily soil moisture depletion as a mechanistic feature improved predictive accuracy over a spectral-only baseline ($\mathrm{\Delta }{R}^2=+0.052$) and anchored predictions within a physically consistent framework based on the FAO-56 water balance, with no false negatives observed for water deficit detection in the hold-out validation set. This framework advances high-throughput, population-scale phenotyping of water-related traits in open-canopy CAM crops, establishing a transferable methodology for operational precision irrigation under tropical savanna conditions. Full article

In this study, daily reference evapotranspiration (ET₀) values for the period 2025–2030 for Lake Burdur, located in the Mediterranean climate zone and within the Burdur closed basin, were estimated using nested architecture focused on high accuracy. The ET₀ target corresponds to the FAO-56 Penman–Monteith reference evapotranspiration variable provided by the Open-Meteo Historical Weather API, and it is treated throughout as a standardized measure of atmospheric evaporative demand rather than as actual lake-surface evaporation or basin water loss. For this purpose, daily mean air temperature, relative humidity, shortwave surface radiation, and evapotranspiration data for the period 1984–2024 were obtained from the Open-Meteo platform. In the first stage of the study (Model 1), separate SARIMAX (statistical), XGBoost (machine learning), and LSTM (deep learning) models were applied for temperature, relative humidity, and radiation series; the model with the highest validation mean for each variable was selected. Accordingly, LSTM (Mean $R^2$ = 0.967) was determined to be the most successful model for temperature, SARIMA(X) (Mean $R^2$ = 0.812) for relative humidity, and XGBoost (Mean $R^2$ = 0.845) for the radiation variable, which is non-linear, has strong autocorrelation, and exhibits distinct seasonality. In the second stage (Model 2), these best climate predictions were used as independent variables for evapotranspiration, and LSTM provided the highest success for evapotranspiration (Mean $R^2$ = 0.941). Trend analyses revealed that the increase in temperature and evapotranspiration and the decrease in relative humidity observed in the past period will continue in the near future. The uncertainty analysis conducted using the Monte Carlo/resampling approach on historical data showed that the 95% prediction intervals largely protected the upward trend in evapotranspiration against random fluctuations. These intervals reflect residual-based uncertainty under the fitted model rather than the full predictive uncertainty of future basin evapotranspiration. The findings indicate that designing model selection appropriate to the structure of the variables within a nested prediction framework significantly improves forecast accuracy and can provide a viable decision support input for sustainable water management in Mediterranean basins experiencing water scarcity. Full article

Drought is becoming more frequent and severe in many regions, particularly in Mediterranean climates, where water demand is increased by warming and changes in precipitation regimes. A long-term assessment of meteorological drought at the Lisbon climatological station is provided in this study using the Standardized Precipitation Index (SPI) and the Reconnaissance Drought Index (RDI) over the period 1864–2021. Monthly precipitation and temperature data are used to compute SPI and RDI at 3-, 6-, and 12-month time scales, so that short-, mid-, and long-term droughts and their temporal evolution can be characterized. RDI is evaluated with three widely used empirical potential evapotranspiration (PET) formulations—Hargreaves, Thornthwaite, and Blaney–Criddle—in order to examine how PET estimations influence drought classification. Given the absence of a physically based reference PET—such as FAO-56 Penman–Monteith—for this station, the focus is on the internal consistency of the PET methods. Furthermore, the Hargreaves formulation is retained as a representative empirical PET for subsequent SPI–RDI comparison. The results show broadly consistent standardized RDI behavior across PET methods; it is indicated that drought conditions are captured more comprehensively by RDI than by SPI because both precipitation deficits and enhanced evaporative demand are included. At the Lisbon station, the estimated average return periods for short-, mid-, and long-term droughts are 3.79, 7.31, and 7.92 years according to RDI, compared with 3.86, 5.69 and 10.88 years from SPI. Several severe drought episodes are identified, including the years 1907, 1922–1923, 1944–1945, 1976, 1981, 1992–1993, 2005, and 2018. While no formal attribution analysis is performed, the drought characteristics are interpreted in the context of observed long-term warming and documented rainfall variability in Lisbon. The findings provide a single-station benchmark of historical drought behavior, by which local water-resources management can be supported and which can serve as a basis for future multi-station and climate-projection-based studies in Portugal. Full article

Reference evapotranspiration (ET₀) represents the atmospheric demand for water from a well-watered vegetated surface and is a key component of the hydrological cycle and agricultural water management. This study evaluated the performance of seven machine learning (ML) models: linear regression (LR), Random Forest (RF), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), Decision Trees (DT), and Cubist, for predicting monthly FAO-56 Penman–Monteith ET₀ in Benin. The target variable was calculated from data collected at six synoptic stations over the 2017–2021 period. Ten remote-sensing and topographic predictors were used: MODIS Land Surface Temperature (LST), six Sentinel-2 optical vegetation indices (NDVI, EVI, NDMI, NDWI, MSI, NDRE), elevation, and cyclic month encoding. Models were trained on the 2017–2019 period and evaluated on an independent temporal test set (2020–2021). All models showed positive predictive performance, with the BMA ensemble achieving the highest accuracy (RMSE = 7.0% of mean ET₀, R² = 0.802), followed by Cubist (RMSE = 7.3%, R² = 0.787) and DT (RMSE = 7.5%, R² = 0.776). The seven models were combined via Bayesian Model Averaging (BMA) with posterior weights estimated by the EM algorithm to produce 1 km monthly ET₀ maps for Benin for 2025. BMA-derived inter-model standard deviation provided spatially explicit uncertainty estimates, revealing that prediction uncertainty is greatest in the northern Sudanian zone during the dry season. The ET₀ target variable was constructed as a hybrid product combining station temperature observations with solar radiation, wind speed, and vapor pressure deficit extracted from the TerraClimate gridded reanalysis dataset; this methodological choice is discussed as a study limitation. Full article

Accurate hourly crop evapotranspiration (ETc) estimation is important for data-driven irrigation management support in tropical horticulture, yet existing approaches are constrained by data requirements and an inability to capture multi-scale temporal dynamics. This study proposes a Transformer encoder model for one-step-ahead hourly FAO-56 PM-derived ETc estimation in a durian orchard in Chanthaburi Province, Eastern Thailand, using 36,528 hourly meteorological observations obtained from the Visual Crossing Weather API for the orchard location over four years, with ETc computed from these inputs using the FAO-56 Penman–Monteith equation. The model employs a 168-h (7-day) look-back window, three stacked encoder blocks with multi-head self-attention ($h=8$, $d_{\mathrm{model}}=128$), and five meteorological input features (air temperature, relative humidity, solar radiation, wind speed, and ETc). A SARIMA$(2,1,2){(1,0,0)}_{24}$ model trained on the same dataset served as the statistical baseline. The Transformer achieved an RMSE of 0.0308 mm/h, MAE of 0.0188 mm/h, and $R^2$ of 0.9018 on the 168-h test set, outperforming SARIMA (RMSE = 0.0717, MAE = 0.0593, $R^2$ = 0.4688), representing a 57.0% reduction in RMSE, a 68.3% reduction in MAE, and a 92.4% improvement in $R^2$. The Transformer also achieved a daytime-only RMSE of 0.0414 mm/h vs. 0.0791 mm/h for SARIMA, and a daily cumulative ETc MAE of 0.1599 mm/day vs. 0.5901 mm/day, demonstrating superior accuracy during agronomically critical periods. The Transformer accurately reproduced both the 24-h diurnal cycle and the 7-day weekly pattern of ETc, whereas SARIMA exhibited a damped amplitude response. A recursive 168-h heuristic simulation demonstrated that the model generates physically plausible ETc patterns under an approximated meteorological scenario, suggesting the approach warrants further investigation as a component of future irrigation decision-support research. These results highlight the potential of Transformer-based deep learning for site-specific, proof-of-concept ETc estimation from meteorological inputs in tropical fruit production, pending validation across diverse sites and seasons. Full article