Abstract: Accurate precipitation prediction is critical for water security and disaster mitigation, yet remains challenging due to atmospheric complexity and class imbalance in rainfall data. This study introduces an integrated "architecture-feature-augmentation" framework to address these limitations. Through systematic comparison of CNN-LSTM and Trans-former architectures, we identify a fundamental trade-off: CNN-LSTM demonstrates higher enhanceability, achieving 80% recall for heavy rainfall when combined with phys-ics-informed augmentation, while Transformer shows superior inherent sensitivity (75% recall) but greater vulnerability to data distribution shifts. Feature engineering benefits are model-specific, significantly improving CNN-LSTM but often introducing redundancy for Transformer. Notably, oversampling techniques like SMOTE achieve peak F1 scores but with substantial generalization gap (ΔF1 > 0.47), indicating overfitting risks, whereas physics-informed augmentation proves more reliable. We establish a principled decision framework: for robust predictions, use CNN-LSTM with physics-informed augmentation; for peak performance where risks are tolerable, employ CNN-LSTM with SMOTE. These findings provide scientific guidance for extreme weather preparedness and water resource management.

Abstract: A harmonic analysis of Antarctic ice core proxy temperature, CO₂ and CH₄ data is presented spanning 350,000 years. Using a greedy algorithm to select periodic components, the analysis initially obtained 59 periods for temperature but subsequently refined this to 55 periods after removing four components (22,150, 9,000, 8,000, and 4,540 years) that exhibited high correlations in the normalized covariance matrix. This refinement ensures stable, well- conditioned parameter estimates while maintaining excellent fits: R² = 0.952 for temperature and R² = 0.964 for CO₂ (truncated at 1850 CE), R² = 0.873 for CH₄ . The algorithm independently recovers the canonical Milankovitch orbital periods (approximately 100,000, 41,000, and 23,000 years) without prior specification, validating both the methodology and the orbital pacing of ice ages (Milankovitch, 1941). Phase analysis reveals that CO₂ consistently lags temperature by 600–4,000 years at orbital timescales, supporting the hypothesis that temperature drives CO₂ through ocean degassing rather than the reverse. Examination of the Last Interglacial (Eemian) reveals a striking asymmetry: CO₂ remained elevated at 275–280 ppm for approximately 13,000 years while temperature declined by 7°C. An R² analysis clearly reveals the Mid-Pleistocene Transition and justifies limiting the input to the fit. The modern CO₂ spike, which departs dramatically from the 350,000-year orbital envelope, is clearly anomalous relative to the harmonic structure of the paleoclimate record.

Abstract: In regions lacking sufficient data, remote sensing (RS) offers a reliable alternative for precipitation estimation, enabling more effective drought management. This study comprehensively evaluates four commonly used RS datasets-Climate Hazards Center InfraRed Precipitation with Station data (CHIRPS), Tropical Applications of Meteorology using Satellite data (TAMSAT), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks-Climate Data Record (PERSIANN-CDR), and Multi-Source Weighted-Ensemble Precipitation (MSWEP) against ground-based data-with respect to their performance in detecting precipitation and drought patterns in the Great Ruaha River Basin (GRRB), Tanzania (1983-2020). Statistical metrics including the Pearson correlation coefficient (r), Mean Error (ME), Root Mean Square Error (RMSE), and Bias were employed to assess the performance at daily, monthly, seasonal (wet/dry), and annual timescales. Most of the RS products exhibited lower correlations (r<0.5) at daily timestep and low RMSE, Bias and ME. Monthly performance improved substantially (r > 0.8 at most stations) particularly during wet season (r = 0.52-0.82) while annual and dry-season performance declined (r < 0.5 and r < 0.3, respectively). Performance under RMSE, Bias, and ME declined at higher timescales, particularly during wet season and annually. CHIRPS, MSWEP, and PERSIANN generally overestimated precipitation while TAMSAT consistently underestimated it. CHIRPS and MSWEP showed superior performance especially in capturing monthly precipitation patterns and major drought events in the basin. Most products struggled to detect extreme dry conditions with exception of CHIRPS and MSWEP at certain stations and periods. Based on these findings, CHIRPS and MSWEP are recommended for drought monitoring and water resource planning in the GRRB. Their appropriate use can help water managers make informed decisions, promote sustainable resource use and strengthen resilience to extreme weather events.

Abstract: This study investigates the interactions between surface atmospheric variables such as temperature, relative humidity, dew point, solar radiation, wind speed, and pressure to understand how thermodynamic and dynamic processes effect local weather conditions. Four diagnostic analyses were performed, viz. (i) the inverse relationship between temperature and relative humidity, (ii) the positive coupling between wind speed and pressure variability, (iii) the association between temperature and dew point during warm and moist conditions, and (iv) the multivariate correlations between all variables. The results show that cooler temperatures correspond to higher relative humidity, while higher temperatures follow with higher dew point values, which indicates improved heat–moisture interaction during warm periods. Wind speed increases with decreasing pressure, reflecting dynamic instability during disturbed weather. The correlation structure reveals two coherent clusters, such as a thermodynamic cluster (temperature, dew point, humidity, solar radiation) and a dynamic cluster (pressure and wind). These findings provide a foundational understanding of weather behavior and offer valuable perceptions for climate modelling, forecasting, and risk assessment.

Abstract: The snowpack is the interface between the atmosphere and the earth’s surface when snow is present. The snowpack energy balance is dictated in part by the nature of the snow surface. The roughness of the snow surface can be quite dynamic. At the Sierra Nevada ski resort in Spain, we measured several snow surface forms: natural, with the presence of dust, with the presence of sun cups, and groomed snow (tracked and between tracks). The snow surface was assessed in 2-D from snow roughness boards and in 3-D from iPad surface scanning to measure across resolutions. Both data collection methods provided similar roughness estimates via the random roughness (RR) and variogram analysis (scale break, SB and fractal dimension, D) for each distinct surface. The geometry-based aerodynamic roughness length (z₀) was computed for the iPad-scanned surfaces yielding an order of magnitude variability in z₀. This produced substantial differences in modelled sublimation.

Abstract:

Wind energy has the potential to become an important source of energy for remote Arctic regions. However, there are risks associated with the exposure of coastal wind parks to extremely strong winds caused by storms and polar lows. Extreme winds can either enhance or reduce wind energy production. The outcomes largely depend on the coastal landscape surrounding the wind park. To address these questions, we conducted a series of simulations using the Weather Research and Forecasting (WRF) model. This study focuses on one of the strongest wind events along the western Norwegian coast - the landfall of the storm “Ylva” (November 24–27, 2017). The study employs terrain-resolving downscaling by zooming in on the area of the Kvitfjell-Raudfjell wind park, Norway. The terrain-resolving WRF simulations reveal stronger winds at turbine hub height (80 m to 100 m above the ground level) in the coastal area. However, it was previously overlooked that the landfall of an Atlantic storm, which approaches this area from the southwest, brings the strongest winds from southeast directions, i.e., from the land. This creates geographically extensive and vertically deep wind-sheltered areas along the coast. Wind speeds at hub height in these sheltered areas are reduced, while they remain extreme over wind-channeling sea fjords. The study demonstrates that optimal wind park siting can take advantage of both sustained westerly winds during normal weather conditions and wind sheltering during extreme storm conditions. We found that the Kvitfjell-Raudfjell location is nearly optimal with respect to the extreme winds of “Ylva.”

Abstract: Flooding episodes caused by a heavy rainfall event has become more frequent especially during the rainfall season in Botswana, this poses some socio-economic and environmental risks. This study investigates the capability of Weather Research and Forecasting (WRF) in simulating a heavy rainfall event which occurred on the 26th of December 2023 in Mahalapye District, Botswana. This event is one amongst many which has had negatively impacted the lives and infrastructures in Botswana. The WRF model was configured using the tropical-suite physics schemes, i.e. (Rapid Radiative Transfer Model, Yonsei University planetary boundary layer scheme, Unified Noah land surface model, New Tiedtke, Weather Research and Forecasting Single-Moment 6-class) on a two-way nested domain (9 km and 3 km grid-spacing) and was initialized with GFS dataset. Gauged station data was used for validation alongside synoptic charts from GFS and ECMWF ERA5 dataset. The results show that the WRF model simulation using the Tropical-Suite physics Schemes is able to reproduce the spatial and temporal patterns of the observed rainfall but with some notable biases. Performance metrics including RMSE, correlation coefficient and KGE showed moderate to good agreement highlighting the model’s sensitivity to physical parameterization and resolution. The results of this study concludes that the WRF model demonstrates promising potential in forecasting extreme rainfall events in Botswana but more sensitivity tests to different parameterization schemes are needed in order to integrate the model into the early warning systems to enhance disaster preparedness and response.

Abstract: This paper examines advancements in climate modeling, emphasizing integrated, physics-grounded and process-oriented approaches to enhance predictive reliability. It underscores the critical role of multiphase phenomena in atmospheric systems, including interfacial heat and mass transfers, and the integration of empirical data and high-resolution observational networks. Coupled with targeted laboratory and numerical experiments, these elements refine the physical basis of climate models. Efforts focus on addressing model limitations, including feedback uncertainties and challenges in AI/ML integration. A central focus is placed on “Statistical-Induced Uncertainties” (systemic biases introduced by spatio-temporal averaging, data interpolation, and ensemble processing) which propagate across modeling stages and may obscure physical interpretations. By embedding empirical rigor and prioritizing transparency, the study advocates for interdisciplinary collaboration to fill observational gaps, especially in under-observed regions, with statistical approaches aligned with physical interpretability. The paper highlights the value of ensemble modeling and AI, not as substitutes but as complements to physics-driven frameworks, supported by clear interpretive methods that anchor models in fundamental process closure. This integrated approach is essential for advancing climate projections and informing effective responses to global climate challenges.

Abstract: On-road vehicles are a primary source of urban air pollution. It is known that high-emitting vehicles represent a fraction of the fleet but contribute significantly to the total emissions. Usually, road transportation emission inventories do not capture the impact of these types of vehicles, underestimating emissions. This study introduces a simple method to refine vehicle emission inventories by incorporating data from Ecuador's Inspection and Maintenance (I/M) program. We analyzed I/M data from Quito to develop a correction factor for the Vehicular Emissions INventory (VEIN) model, accounting for the higher emissions from vehicles that fail inspection. Our analysis showed that while less than 10% of gasoline and 20% of diesel vehicles failed inspection, their emissions were substantially higher; for instance, reproved vehicles produced 3.9 times more CO and 6.2 times more HC on average. Applying our correction factor increased total emission estimates by an average of 33%, with CO emissions rising by 65%. These findings demonstrate that incorporating I/M data is crucial for accurately quantifying vehicular pollution. The proposed methodology offers a way to create more realistic emission estimates, providing a better tool for policymakers to manage air quality.

Abstract: This study investigates the mechanisms of topographic steering and the resultant track morphology of typhoon-like vortices over complex terrain. Leveraging a dynamic model based on potential vorticity (PV) conservation, we conducted a comprehensive sensitivity analysis over both an idealized bell-shaped mountain and the realistic topography of Taiwan. Results indicate that a triad of controls governs track evolution: vortex intensity (α), terrain geometry (dhB*/dt*), and interaction time (impinging angle γ). To quantify predictability, we introduce the Track Divergence Percentage (td), which partitions the phase space into distinct Track Diverging (TDZ) and Converging (TCZ) Zones. While idealized simulations established this fundamental structure, realistic experiments incorporating vortex decay revealed that orographic complexity and shallow approach angles (&lt;145°) lead to regimes of hyper-sensitivity and "terrain capture." These findings establish a unified quantitative framework for understanding track bifurcation and looping, offering crucial insights for assessing forecast uncertainty in mountainous regions.

Abstract: Baroclinic wave resonance, particularly Rossby waves, has attracted great interest in ocean and atmospheric physics since the 1950s. Research on Rossby wave resonance covers a wide variety of phenomena that can be unified when focusing on quasi-stationary Rossby waves traveling at the interface of two stratified fluids. This assumes a clear differentiation of the pycnocline where the density varies strongly vertically. In the atmosphere, such stationary Rossby waves are observable at the tropopause, at the interface between the polar jet and the ascending air column at the meeting of the polar and Ferrel cell circulation or between the subtropical jet and the descending air column at the meeting of the Ferrel and Hadley cell circulation. The movement of these air columns varies according to the declination of the Sun. In the oceans, quasi-stationary Rossby waves are observable in the tropics, at mid-latitudes, and around the subtropical gyres (i.e. the gyral Rossby waves GRWs) due to the buoyant properties of warm waters originating from tropical oceans, transported to high latitudes by western boundary currents. The thermocline oscillation results from solar irradiance variations induced by the Sun's declination, as well as solar and orbital cycles. It is governed by the forced, linear, inviscid shallow water equations on the β-plane (or β-cone for GRWs), namely the momentum, continuity and potential vorticity equations. The coupling of multi-frequency wave systems occurs in exchange zones. The geostrophic forces reflecting the superposition of zonal/polar and meridional/radial currents of the waves perturb the geostrophic balance of the basin. Here, it is shown that the ubiquity of resonant forcing in (sub)harmonic modes of Rossby waves in stratified media results from two properties: 1) the natural period of Rossby wave systems tunes to the forcing period - 2) the restoring forces between the different multi-frequency Rossby waves assimilated to inertial Caldirola-Kanai (CK) oscillators are all the stronger as the perturbations of the geostrophic balance in the exchange zones are more significant. According to the CK equations, this resonance mode ensures the sustainability of the wave systems despite the variability of the forcing periods. The resonant forcing of quasi-stationary Rossby waves is at the origin of climate variations as well-known as El Niño, glacial-interglacial cycles or extreme events generated by cold drops or, conversely, heat waves. This approach attempts to provide some new avenues for addressing climate and weather issues.

Abstract: Atmospheric rivers (ARs) bring much of the winter precipitation to the western United States and are important for water supply and flood hazards. This study analyzes the clustering of ARs and evaluates how seasonal cycles and climate modes influence their frequency and spacing. Daily precipitation records and reanalysis data from 1980–2021 were used to detect ARs with an integrated vapor transport threshold, and events were grouped when two or more ARs occurred within a 7-day window. The results show that models capture the general occurrence of ARs but underestimate clustering by about 20–30%, with larger errors in winters affected by El Niño and La Niña. Sensitivity tests show that cluster measures depend on threshold and time window choices, while regional differences are linked to topography and circulation. The results suggest that persistence and spacing should be studied as separate measures, and that higher-resolution models are needed to represent narrow moisture plumes and rapid sequences. Although limited by dataset length and resolution, this study provides guidance for improving seasonal forecasts and for supporting water and flood management in AR-prone regions.

Abstract:

The last atmospheric emission inventory for Cuenca, a city located in the Andean region of southern Ecuador, was developed for the year 2021 (EI 2021), encompassing both primary pollutants (NO_x, CO, VOC, SO₂, PM₁₀, and PM_2.5) and greenhouse gases (CO₂, CH₄, and N₂O). We formally assessed the quality of the emission inventory by modeling air quality levels during October 2021 using the Weather Research and Forecasting with Chemistry (WRF-Chem 3.2) model at a high spatial resolution (1 km). We activated the direct effects for modeling the feedback between aerosols and atmospheric variables. The metrics indicated that both meteorological and air quality variables were modeled acceptably, suggesting the quality of the emission inventory and the ability of WRF-Chem 3.2 to perform atmospheric modeling in this complex region, using the “one atmosphere” approach. The results and spatial distribution of the EI 2021 were compared to the emission data coming from the last version of the Edgar Emissions Dataset (spatial resolution of 11.1 km) one of the most used global emission data, which suggested that for the Equatorial Andean region, the Edgar Dataset results require improvement, at least for some primary pollutants (CO, VOC, SO₂) in terms of magnitude, and of the spatial configuration of all the pollutants, before they can be used for atmospheric modeling. We also identified future research activities to improve the emission inventories and atmospheric modeling performance in the Equatorial Andean region.

Abstract: This study introduced a pre-training method for machine learning-based climate prediction models. The method leverages the advantage of climate events (some theoretical knowledge) to address their limitation (small sample size). It consists of the following steps: generating artificial samples via composite analysis of high and low anomaly events, pre-training predictive models with these samples, and selecting an optimal pre-trained model that most closely matches the observational training set from numerous repeated experiments with only the model’s random number seeds being varied. Sensitivity experiments demonstrate that this pre-training method not only substantially improves predictive skills but also significantly reduces prediction instability. This simple and practical pre-training method is applicable not only to the climate prediction events in this study but also to all climate events for which composite analysis is applicable.

Abstract: Large eddy simulation (LES) within a weather research and forecasting (WRF) model coupled with an active scalar transport equation was used to simulate Atmospheric Boundary Layer (ABL) conditions during the Mosquito wildland fire, the largest wildland fire in California during September of 2022. The simulations were conducted with realistic boundary conditions derived from the NOAA High Resolution Rapid Refresh (HRRR) model, with the aim of better understanding the two-way coupling between atmospheric boundary layer (ABL) and plume dynamics. The terrain was extremely inhomogeneous and the topography varied quite significantly within the numerical domain. Initially, LES of the smoke-free ABL were conducted on nested domains, and detailed ABL data were gathered from 08-09 September, 2022. LES simulations were validated using 4 ASOS stations and NOAA-MET Twin Otter measurements and the desired accuracy has been established. The smoke plume was then released into the ABL at noon on 09 September, 2022 and the plume simulations were conducted for a period of one hour from the release. During this period, the ABL transitioned from convective to buoyancy-shear-driven regimes. Late-night and early-morning conditions are influenced by the complex topography and low-level jet, whereas buoyancy and shear control the ABL dynamics during the morning and afternoon hours. The plume vertical transport is influenced by the ABL-depth and the size of the vertical turbulence structures during that time, whereas, the wind conditions and turbulent kinetic energy within the ABL dictate the horizontal transport scales of the plume. In addition, the results demonstrate that the plume modifies the microclimate along its path.

Abstract: Dry climate regions face heightened risks of flooding and infrastructure damage even with minimal rainfall. Climate change is intensifying this vulnerability by increasing the duration, frequency, and intensity of precipitation events in areas that have historically experienced arid conditions. As a result, accurate precipitation estimation in these regions is critical for effective planning, risk mitigation, and infrastructure resilience. This study evaluates the performance of five satellite- and model-based precipitation products by comparing them against in-situ rain gauge observations in a dry-climate region: The fifth generation European Centre for Medium-Range Weather Forecasts Reanalysis (ERA5) (analyzing maximum and minimum precipitation rates separately), the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA2), the Western Land Data Assimilation System (WLDAS), and the Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG). The analysis focuses on both average daily rainfall and extreme precipitation events, with particular attention to precipitation magnitude and the accuracy of event detection, using a combination of statistical met-rics—including bias ratio, mean error, and correlation coefficient—as well as contingency statistics such as probability of detection, false alarm rate, missed precipitation fraction, and false precipitation fraction. The study area is Palm Desert, a mountainous, arid, and urban region in Southern California, which exemplifies the challenges faced by dry re-gions under changing climate conditions. Among the products assessed, WLDAS ranked highest in measuring total precipitation and extreme rainfall amounts but performed the worst in detecting the occurrence of both average and extreme rainfall events. In contrast, IMERG and ERA5-MIN demonstrated the strongest ability to detect the timing of pre-cipitation, though they were less accurate in estimating the magnitude of rainfall per event. Overall, this study provides valuable insights into the reliability and limitations of different precipitation estimation products in dry regions, where even small amounts of rainfall can have disproportionately large impacts on infrastructure and public safety.

Abstract: Due to the increasing concern over climate change and its environmental impacts, effective greenhouse gases (GHG) control strategies are of pivotal importance. The use of models such as STILT (Stochastic Time Inverted Lagrangian Transport), statistical techniques, and experimental data analysis provide valuable tools to quantify emissions and identify GHG tendencies. The Mediterranean basin is considered a global hotspot for air-quality and climate change: here, we combine experimental datasets of atmospheric methane (CH4) and carbon dioxide (CO2) with atmospheric transport models to present an atmospherically-based framework for monitoring GHG emissions. We applied methodologies, i.e., the Smoothed Minima (SM) and STILT, to extract background concentration data from the time series of atmospheric gases and identify measurements deemed representative of atmospheric background (GRD) levels. At the Lamezia Terme (Global Atmosphere Watch, GAW code: LMT), Capo Granitola (GAW code: CGR), and Lampedusa (GAW code: LMP) observation sites, GHG measurements were performed with specific calibration routines carried out using primary standards of calibration from the National Oceanic and Atmospheric Administration – Global Monitoring Laboratory (NOAA–GML), with secondary standards used to evaluate possible drifts and calibration factors stability. The first two are coastal stations and the third is an island station. At these sites, atmospheric CH4 and CO2 mole fractions can be evaluated at local and continental scales, in locations with specific Mediterranean climatic characteristics. This paper presents the variability of CH4 and CO2 in the central Mediterranean basin by analyzing hourly GHG concentrations over a 9-year period (2015-2023) for LMT, a 8-year period (2015-2022) for CGR, and a 19-year period (2006-2024) for CO2 and 5-year period (2020-2024) for CH4 at LMP. STILT provides 3-hourly results for methane and carbon dioxide concentrations that correlate well with surface measurements at LMT, CGR, and LMP. These analyses are aimed at relevant long-term datasets of GHG over southern Italy. This work would provide a useful contribution to comparing the observed concentrations of gases measured at three sites in the central Mediterranean with those predicted by models such as STILT.

Abstract: The COVID-19 lockdown in 2020 caused a marked decline in industrial and traffic emissions, which lowered aerosol levels in many regions. This study examined how these changes affected rainfall in the Western United States. Data from 95 meteorological stations and MODIS satellite products were used for March–December 2020 and compared with the same months in 2018, 2019, and 2021. A linear regression model and a rainfall anomaly index were applied to analyze the relation between aerosol optical depth and rainfall. Results show that aerosol optical depth and nitrogen dioxide fell by 15–25% in large urban areas, while rainfall changes were small and uneven. Valleys with warm-rain processes recorded increases of up to 5%, while mountain regions with mixed-phase clouds showed changes within ±3%. Most aerosol–rainfall relations were weak and not significant at the 95% level. These results indicate that reduced aerosols may slightly increase warm-rain efficiency, but circulation and terrain remain the main drivers. The lockdown acted as a short-term case to study air quality and rainfall, and the findings are useful for climate studies, water management, and rainfall forecasting.

Abstract: Machine Learning (ML) models are powerful tools for meteorological applications but often operate as "black boxes", hindering scientific understanding. This study addresses this challenge by implementing an interpretable ML approach Rulefit for Quantitative Precipitation Estimation (QPE). The algorithm was applied to a multisource dataset from eastern China, incorporating composite radar reflectivity from China’s New Generation DopplerWeather Radar, Digital Elevation Model data, and Himawari-8 satellite bands 7-10. A geographical mask based on China’s national borders was applied to exclude data points outside the land area. Additionally, standard scaling normalization was applied to all input features to account for different units and value ranges across datasets. The model’s performance was evaluated against six baseline models and traditional Z-R relationship. The RuleFit model demonstrated strong predictive performance, achieving a Critical Success Index (CSI) of 0.6015 and a Probability of Detection (POD) of 0.9359 for 10-minute precipitation events exceeding a 2mm threshold. This accuracy was comparable to other ML models and significantly surpassed traditional Z-R relationship methods. Crucially, the generated decision rules and Partial Dependence Plots(PDP) provided transparent insights into the model’s logic, revealing key non-linear interactions between radar and satellite features and showing an emphasis on predicting heavy rainfall. Our findings show that RuleFit is not only an accurate QPE tool but also a framework for uncovering meteorological relationships, thereby building researcher confidence and addressing the critical need for interpretability in ML-based atmospheric science.

Abstract: Accurate short-term temperature forecasts are important in snow-dominated mountain regions because small errors can strongly affect snowmelt, runoff, and water storage. This study applied analog ensemble post-processing to improve temperature forecasts in the Sierra Nevada, which provides major water supply for the western United States. Historical forecast–observation pairs were used to build analog ensembles, and the corrected forecasts were applied to a degree-day snowmelt model and a hydrological model. The corrected forecasts reduced mean bias from +1.2 °C to +0.5 °C and lowered root mean square error by about 18 % compared with raw forecasts. Snowmelt onset was predicted within two days of observation, while the control forecasts were five days early, and peak melt overestimation fell from 20 % to 7 %. Streamflow simulations using corrected forecasts also improved, with Nash–Sutcliffe efficiency increasing from 0.61 to 0.77 and peak flow error decreasing from 22 % to 8 %. Reservoir storage curves showed better agreement with observed filling and release cycles. These results show that correcting temperature forecasts improves hydrological prediction during the spring melt season. The study highlights the value of analog ensemble methods for water management, including reservoir operation, flood control, and drought planning, but also notes that further tests are needed under different climate conditions and in other regions.