As potential climatic changes take place on a global scale, the severity and rate of occurrence of a typhoon, which involves high-speed wind, heavy rainfall, and storm surge, are becoming worrisome and a significant danger to lives and property. The traditional methods of predicting the movement of a typhoon fail to perform effectively in the occurrence of large-scale and complicated data. Recent advances in deep learning have supported learning of temporal patterns from large-scale meteorological data more effectively, making them attractive for typhoon track forecasting applications. This study presents a comparative analysis and a framework for multi-horizon typhoon track forecasting using various machine learning regression and deep sequence models. It includes Linear Regression, Random Forest Regression, Recurrent Neural Network, Long Short-Term Memory and Gated Recurrent Unit models. The key typhoon parameters considered are Latitude, Longitude, Maximum Sustained Wind Speed (VMAX), Minimum Sea Level Pressure (MSLP), and Radius of Maximum Wind (MRD). Forecasts are generated for lead times of 6, 12, 24, 48, and 72 hours. The quality of the forecast by the models is quantified through error-based and trajectory-based measures, primarily using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Great Circle Distance (GCD). The proposed framework consists of data collection, preprocessing, and time-series sequence generation for forecasting horizons of up to 72 hours. Historical typhoon records over the Western North Pacific region obtained from the Joint Typhoon Warning Center (JTWC) for the period 1945–2024 are used for model development and evaluation. The experimental results showed that, compared with the LSTM model, the GRU model reduced the Great Circle Distance (GCD) errors by 4–20
Drought is a recurrent hydro-climatic hazard in semi-arid Mediterranean environments, where strong rainfall variability and increasing water stress threaten agricultural production and water resource sustainability. In this study, monthly precipitation records from seven rainfall stations covering the period 1967/68–2021/22 were analyzed in the Wadi Sly basin using four precipitation-based drought indices: the Standardized Precipitation Index (SPI), the empirical Standardized Precipitation Index (SPIe), the Rainfall Anomaly Index (RAI), and the China Z-Index (CZI). Drought events were identified through the Run Theory, and their main characteristics, including frequency, duration, and severity, were quantified. In addition, the Innovative Trend Analysis (ITA) method was applied to assess long-term changes in drought behavior. The results show a strong consistency among SPI, SPIe, and CZI, with very high correlations across all stations, indicating that these indices provide comparable representations of drought variability in the basin. In contrast, the RAI systematically produced higher estimates of drought frequency and severity, suggesting a greater sensitivity to precipitation anomalies. Spatial analysis revealed marked heterogeneity among stations, with some sites characterized by more frequent but shorter drought episodes, while others experienced fewer but more severe and persistent events. Trend analysis highlighted a progressive decline in wet extremes and a tendency toward increasing climatic dryness in the basin. These findings highlight that drought characterization is strongly influenced by the choice of index and underscore the importance of selecting robust, regionally appropriate indicators for effective drought monitoring and water resource management in Mediterranean North Africa.
Atmospheric Boundary Layer Height (Zi), a key characteristic of the boundary layer, is a crucial parameter in modelling cloud formation, weather, and air quality. Hence, its accuracy is vital for reliable modelling and forecasting. In this study, India’s first regional reanalysis dataset, the Indian Monsoon Data Assimilation and Analysis (IMDAA), is validated against multi-platform measurement datasets. It is found that Zi from IMDAA is in good agreement with radiosonde and satellite observations and performs better than the other reanalysis datasets. However, IMDAA significantly underestimates Zi (< 100 m) during December over humid subtropical climatic regions like Jammu and Gorakhpur, possibly due to high negative net radiation, overestimated cloud cover, and inaccuracies in surface energy partitioning in IMDAA. For the first time, a comprehensive characterization of the daily maximum Zi (referred to as Zi, max) is performed. The analysis of diurnal and seasonal variations of Zi, max revealed that the atmospheric boundary layer is fully developed (deepest) at 15:30 IST across all seasons over most of the Indian region. Spatially, Zi, max reached its highest value of 4500 m during the pre-monsoon season over central India. But during the monsoon, it reduces to 2500 m, and the peak location gets shifted to northwest India. In addition, we report that it is the sensible heat flux that primarily drives the Zi, max rather than the Bowen ratio over central India. These results have large implications for Zi dynamics in regional numerical weather, climate, and air quality simulations.
Understanding the propagation dynamics from meteorological to agricultural drought is essential for improving early warning systems and enhancing drought preparedness. This study investigates drought dynamics in the Palakkad district of Kerala, India, using the Standardized Precipitation Evapotranspiration Index (SPEI) and the Standardized Soil Moisture Index (SSI) as indicators of meteorological and agricultural droughts, respectively. A combination of methods, including correlation-based propagation time estimation, copula-based conditional probability analysis, and Event Coincidence Analysis (ECA) was employed. A two-month propagation time between meteorological and agricultural droughts was observed across the study area based on the correlation-based analysis. Conditional probability analysis using the Clayton copula revealed strong associations across drought categories, with moderate to extreme meteorological droughts frequently preceding agricultural droughts of similar or higher intensity. ECA results showed that trigger rates consistently exceeded precursor rates, highlighting the dominant role of drought initiation. A sharp increase in trigger coincidence rates at the 2-month time window reinforces the findings from the correlation-based propagation time analysis. However, slight variations in the optimal time window were observed across different drought severity classes, which may be useful for class-specific drought analysis and tailored mitigation strategies. Furthermore, ECA demonstrated that local climatic variables (rainfall, temperature, and humidity) and global climate indices particularly ENSO-related indices, significantly influence drought occurrence, with global influences becoming more prominent during severe droughts. This study enhances the understanding of drought propagation mechanisms and provides valuable insights for improving drought forecasting and management strategies under increasing climate variability.
Climate change has posed a severe threat to fragile deltaic ecosystems and densely populated regions. The study aims to examine long-term temporal changes in rainfall and temperature during 1972–2024 and forecast seasonal temperature and rainfall for 2025–2054 using data from India Meteorological Department. Modified Mann–Kendall, Sen’s slope estimator and false discovery rate were utilized to detect significant trends while random forest, long short-term memory and convolutional neural networks-bidirectional long short-term memory (CNN-BLSTM) models were employed for seasonal forecasting of temperature and rainfall. The comparative evaluation of these models confirmed the superior accuracy and robustness of the CNN-BLSTM model for seasonal forecasting of rainfall and temperature. Rainfall trends exhibited significant increases in January (0.36 mm year⁻1), June (1.53 mm year⁻1) and December (0.13 mm year⁻1) while April showed a significant decline of − 0.50 mm year⁻1, highlighting disruptions in seasonal freshwater availability. An increasing trend in maximum temperature was observed during monsoon and early winter months with July, August, September and December months while minimum temperature displayed widespread and consistent increasing trend across most months and seasons with an annual trend increasing by 0.0206 °C year⁻1. The forecasting indicated a decreasing summer and increasing monsoon rainfall with an increase in mean and minimum temperature during summer and monsoon seasons. These findings underscore serious implications of climate change on mangrove ecosystem resilience, agriculture, water resource management and socio-ecological conditions in SBR. The paper offers a novel reproducible methodological framework that can be applied to other geographical regions for analyzing and forecasting climate variability.
Accurate meteorological forcing is critical for modelling the atmospheric transport and dispersion of radionuclides. This study assessed WRF configuration choices for radionuclide applications over Ireland. First, FLEXPART-WRF sensitivity experiments were used to identify which aspects of WRF configuration produced the largest differences in simulated transport and dispersion, thereby reducing the shortlist. Second, six shortlisted WRF configurations were evaluated against hourly observations from 28 Irish meteorological stations using wind speed, wind direction, precipitation, and 2 m air temperature for April 2015 and October 2015. Wind speed, wind direction, and precipitation were used as configuration-ranking variables, while 2 m temperature served as a near-surface thermodynamic diagnostic relevant to boundary-layer interpretation. Boundary-layer behaviour was examined using modelled PBL height across the station network and a single-site comparison with ceilometer-derived mixing-layer depth at Mace Head. Case-study analyses diagnosed differences in precipitation structure and near-surface wind fields, and the formal configuration ranking was based on the month-scale station evaluation. Different configurations led to the combined variable-specific evaluations: LGN-YSU-RG for wind speed, MKN-MYJ-CC for wind direction, and LGN-MYNN-RG for precipitation. When the three variables were weighted equally, LGN-YSU-RG and MKN-MYJ-CC were the joint highest overall. The results therefore identify two leading configurations with complementary strengths. For the periods and configurations tested, the results support an application-dependent choice between the leading configurations for radionuclide applications over Ireland. The station-based and single-site diagnostics provide evidence on near-surface meteorology and boundary-layer behaviour, but the validation of the full three-dimensional fields is outside the scope of this study.
The onset dates of extreme temperature events (ETEs) have become increasingly uncertain and complex under global warming, posing unprecedented challenges to society and natural ecosystems. Analyzing variations in the onset dates of ETEs can provide a reference for agricultural management, health protection, and natural disaster prediction. Reanalysis datasets have the advantages of extensive spatial coverage, long-term continuous records, and high spatiotemporal resolution, and are widely used in studies of extreme climate events. However, whether reanalysis datasets are suitable for analyzing changes in the onset dates of ETEs remains unverified. In this study, 12 indices were selected to quantify such changes based on daily temperature data from 595 meteorological stations in mainland China. Three reanalysis datasets were adopted in this study: the fifth-generation European Reanalysis (ERA5)-Land dataset, the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) dataset, and the Japanese Reanalysis (JRA)-3Q dataset. Five evaluation indicators were then used to assess their ability to capture the onset dates of ETEs in China: normalized root mean square error (NRMSE), relative bias (RB), Pearson’s correlation coefficient (CC), normalized mean absolute error (NMAE), and the distance between the indices of simulation and observation (DISO). The results show the following. (1) All three reanalysis datasets can capture the annual variations in the onset dates of ETEs, but their simulation accuracy differs. The NCEP/NCAR dataset exhibits particularly large deviations in representing these annual variations. (2) The applicability of the reanalysis datasets varies across regions. Large deviations were observed for all datasets in Southwest and South China, as well as for frost and growing season onset dates in the Qinghai–Tibet region. (3) Overall, the ERA5-Land dataset performs best (DISO = 0.65), followed by JRA-3Q (DISO = 0.75) and NCEP/NCAR (DISO = 0.99). These findings are helpful for understanding changes in ETEs in mainland China and selecting suitable reanalysis datasets for determining the onset dates of ETEs in subsequent studies.
Global Climate Models (GCMs), as physically-based models (PBMs), are robust tools for climate-adaptive planning, yet when time or expertise is limited, stochastic modeling can offer a simpler alternative. Few studies have systematically compared stochastic models with PBMs for forecasting precipitation and temperature. This study evaluates ARMA, ARIMA, NSTF, and an improved NSTF variant (INSTF), while testing the effects of Yeo–Johnson (YJ) and Inverse Hyperbolic Sine (IHS) transformations using limited observational data from Lashkenar Village, Iran (2007–2023), and PBM outputs for 2024–2054. The INSTF model introduces quasi-dynamic (induced) noise via Iterative Fourier Series (IFS) to mimic annual cycles and enforce regime-aware noise selection, replacing the pure noise structure used in NSTF. During the simulation period (2019–2054), the standard uncertainty of the mean—a desirable performance metric—for the profound models was nearly zero for ARMA (2,3) and 0.03 for INSTF in annual precipitation modeling, while for annual temperature modeling, it was − 0.08 for both NSTF and INSTF. The YJ transformation generally outperformed IHS, although in some cases, results were better without any transformation. The De Martonne aridity index, which initially indicated a semi-arid climate, declined post-2024 across all models, indicating intensifying dryness. While stochastic models aligned well with PBMs trends in climate change (CC) and can therefore be recommended for CC forecasting, they remain less reliable for real-time precipitation and temperature event forecasting, particularly in design-phase applications.
This study statistically evaluates convective downburst environments over the Tehran region using the Downburst Precursor Parameter (DPP) framework, based on ERA5 reanalysis data and surface observations from Mehrabad and Imam Khomeini airports. The analysis spans five warm-season months (April to August) across the years 2014, 2021, and 2025. Downburst events were identified from METAR and SPECI reports using a multi-criteria filtering algorithm, capturing severe events with actual recorded wind gusts reaching up to 50 knots. Two core components of the DPP—bulk wind shear between 500 and 875 hPa and the equivalent potential temperature difference (Δθe) between 875 and 700 hPa—were extracted from reanalysis data. These specific pressure levels were tailored to capture the deep, dry planetary boundary layer (PBL) characteristic of Tehran’s semi-arid climate. Using logistic regression with class balancing and out-of-sample validation, regionally optimized DPP coefficients were derived for Tehran. Additional classification models, including Random Forest and XGBoost with SMOTE and weighting strategies, were used to benchmark the predictive skill of the DPP framework. ROC AUC scores ranged from 0.586 to 0.624, confirming the index’s discriminatory power despite the class imbalance inherent in downburst detection. Finally, to compensate for the lack of operational radar data, a well-documented convective event on 27 April 2025 was analyzed using surface observations, Meteosat Dust RGB imagery, lightning data, and ERA5-derived DPP maps. The results demonstrate that DPP captures the combined thermodynamic and dynamic conditions favoring dry downbursts and offers value as an operational diagnostic tool in urban forecasting contexts.
This study provides a meteorological‑statistical explanation of the way in which the duration of seasonal thermal periods can be annually determined, as well as the possibility of drawing conclusions related to climate change from their analysis. To the best of our knowledge, this is the first study in North Macedonia to analyze seasonal dynamics using such a statistical-meteorological approach. The study used data on daily temperature series and seasonal durations, obtained for the period 1951–2024, from the Main Meteorological Station (MMS) in Prilep (National Hydrometeorological Service, NHMS‑Skopje, North Macedonia). The methodological approach is based on defining indicative temperature benchmarks, obtained from the characteristic average temperatures for the months in which the seasons onset: March, June, September, and December. Based on indicative temperature markers and the quartile distribution of the mean daily temperature, thermal thresholds were defined to determine the duration of seasons. During the annual analysis the onset of the seasons were determined through the criterion of six consecutive days exceeding the quartile threshold for spring and summer, and six consecutive days with the mean daily temperature falling below the threshold for autumn and winter. Furthermore, Student’s t-test and Fisher’s F-test were performed to compare selected historical 19-year intervals (1951–1969, 1969–1987, 1979–1997, and 1987–2005) with the most recent 19-year interval (2006–2024), in order to assess whether seasonal-duration distributions have changed over time. Significant differences in the mean values and variances of summer duration were observed. In winter, the statistical analysis showed a decrease in winter duration in the recent period, although interannual variability remained relatively stable. These results indicate a structural change in seasonal dynamics, expressed mainly through longer and more stable summer conditions and shorter winter conditions. The observed changes are consistent with broader evidence of recent warming and altered seasonal temperature regimes, but they should be interpreted as station-level evidence for the Prilep MMS rather than as a national-scale conclusion. These findings align with global evidence of a reduction in temperature variability coupled with an increase in the occurrence and severity of extreme heat events in recent decades (Zhou et al. 2024).
The West African Monsoon (WAM) is a fundamental driver of regional climate and socio-economic stability; however, its variability—particularly regarding aerosol-mediated influences—remains insufficiently understood. This study investigates the impacts of primary regional aerosol types—mineral dust, black carbon (BC), organic carbon (OC), and sea salt—on WAM precipitation dynamics over West Africa from 1980 to 2025. Utilizing high-resolution reanalysis datasets (ERA5 and MERRA-2), we conducted precipitation anomaly assessments, Modified Mann–Kendall trend testing, and Simple Daily Intensity Index (SDII) analysis, alongside bivariate and canonical correlation analyses (CCA). Our results reveal pronounced spatio-temporal variability in precipitation, characterized by a distinct coastal–Sahelian dipole and a notable shift toward wetter conditions post-2000, particularly during the monsoon withdrawal phase (September–November, SON). While June–August (JJA) rainfall is characterized by high-intensity events, SON rainfall is driven by increased frequency. Aerosol–precipitation interactions exhibit strong regional contrasts: carbonaceous aerosols (BC and OC) show significant negative correlations with precipitation depth over the Guinea Coast, attributed to radiative heating and cloud-suppressing effects. Conversely, sea salt displays positive correlations, likely due to enhanced cloud condensation nuclei (CCN) activity. Mineral dust exerts a comparatively weak and spatially heterogeneous influence. CCA underscores the dominant role of mixed aerosols, with the leading mode explaining the majority of the covariance between aerosol loading and precipitation dynamics. Elevated carbonaceous aerosol concentrations, coupled with reduced sea salt levels, are strongly associated with suppressed rainfall, highlighting the critical role of aerosol radiative and microphysical processes. These findings provide essential insights for refining regional climate models and improving precipitation forecasting, with direct implications for water resource management and agricultural planning in West Africa.
Climate variability driven by large-scale oceanic and atmospheric phenomena such as El Niño- Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD), influences rainfall variability in India. Unlike previous studies that primarily focused on broad scale analyses, this study investigates the independent impacts of pure ENSO and pure IOD events on rainfall and drought variability at regional scale. By integrating rainfall magnitude and frequency, Standardized Precipitation Index (SPI), and Palmer Drought Severity Index (PDSI), it provides a comprehensive regional scale assessment of hydroclimatic variability. The present study focuses on the climate-sensitive, predominantly rainfed Vidarbha region of Central India, where agriculture is highly vulnerable to climate variability. Accordingly, it assesses the impacts of pure ENSO (El Niño and La Niña) and pure IOD (positive and negative) events on rainfall over a 44-year period (1981–2024), providing a basis for future finer-scale socioeconomic assessments. Pure events refer to years when one climate mode is active while the other remains neutral, categorized using the Oceanic Niño Index (ONI) for ENSO events and the Dipole Mode Index (DMI) threshold values for IOD events. The study employed the optimized hotspot analysis (OHA), SPI, PDSI, and number of rainy days (NRD) to identify the rainfall pattern and drought condition. Pure El Niño events in the Vidarbha region are associated with reduced rainfall and drier conditions (lower SPI, PDSI, and heavy rainy days), whereas pure positive IOD events are associated with wetter conditions. Pure El Niño events recorded 2.44 rainy days (> 50 mm) and a mean PDSI of − 1.30, compared with 5.24 rainy days (> 50 mm) and a mean PDSI of 1.82 during pure positive IOD events. Pure events of negative IOD and La Niña have a mixed impact on the rainfall in the region. The study concludes that ENSO and IOD are among the major drivers of rainfall variability in the Vidarbha region, exerting a significant influence on monsoon rainfall and drought variability, although other climatic and regional atmospheric factors also play important roles.
To develop approximate initial perturbation schemes for a convection-permitting ensemble prediction system (CPEPS), we studied the characteristics of multiscale singular vector (SV) downscaling perturbations and observed perturbations and revealed the advantages of combined perturbations over single perturbations. The results indicate that multiscale SV downscaling perturbations exhibit flow-dependent features with larger magnitudes across various wavelengths. In contrast, the observed perturbations are distributed throughout the entire domain, with circular patterns around the sounding observation stations and smaller magnitudes. The combination of multiscale SV perturbations and observed perturbations, referred to as the combined perturbations, effectively integrates the characteristics of both, thereby producing the largest energy spectra and initial perturbation patterns with broader spatial distributions. Baroclinic instability and moist convection are the main mechanisms driving the development of these perturbations. Multiscale SV downscaling perturbations show the highest perturbation energy at both initial and forecast times, as well as the best spread–skill relationships and probabilistic prediction ability for all variables. Although the observed perturbations have overall lower prediction skills, they still positively impact certain precipitation cases in southern China, highlighting their essential role. Compared with the best-performing multiscale SV perturbations, the combined perturbations further enhance the perturbation energy, probabilistic prediction scores, as well as spread–skill relationships for all the variables over the first 24 h. Overall, this study demonstrates the distinct impacts of various initial perturbation methods in CPEPS and provides insights for developing improved initial perturbation schemes.
This study documents the lifecycle and large-scale environment of a large-hail producing supercell that occurred in southern Brazil (SB) on 6 October 2015. The supercell underwent a merger with a nascent supercell near a Doppler weather radar and rapidly intensified shortly after the interaction. The original (premerger) supercell developed in a high-shear, high-instability pre-cyclogenesis environment and remained isolated until the merger due to the presence of a relatively rare elevated mixed layer that helped suppress widespread convection and provided steep lapse rates for robust updrafts. A small cell developed 30 km to the north of the original supercell and merged in between its inflow and forward-flank regions, resulting in a new (postmerger) supercell that inherited the main features of the intensifying nascent cell. Radar-derived quantities related to supercell intensity reveal that the updraft and mesocyclone intensified significantly within 30 min of the merger, resulting in a large episode of hailfall with 6 + cm hail reported at the surface. After reaching peak intensity, the postmerger supercell and its attendant low-level mesocyclone rapidly decayed, likely as a result of a hailfall-induced outflow surge and large convective inhibition. This case illustrates the importance of mergers between mature and nascent supercells to the short-term forecasting of severe weather, a topic that received little attention in studies and prediction of storms in SB.
Each category of tropical cyclones (TCs) poses distinct features at different stages of its life cycle. Although general features of rapid intensification (RI) have been studied, the distinct nature of RI among different categories of TCs over the North Indian Ocean (NIO) remains less understood. The present study aims to explore the distinct physical characteristics across different categories of RI and non-RI (NRI) TCs over the NIO. It is observed that with increasing intensity, cyclonic systems are more prone to RI. The Extreme severe cyclonic storm (ESCS) category is the most frequent (52.11
Accurate prediction of precipitable water vapor (PWV) is important for weather forecasting, hydrological modeling, and climate diagnostics. This need is particularly relevant in regions with complex topography and limited observational coverage. Existing models often struggle to capture temporal dependencies, maintain predictive accuracy across heterogeneous regions, and provide interpretable forecasts. To address these limitations, this study proposes an explainable hybrid deep learning framework in which each module targets a specific forecasting challenge. One-dimensional convolutional neural networks (1D-CNNs) extract localized month-to-month patterns from the meteorological sequences. The adaptive scaled dot-product attention (ASDPA) mechanism uses learnable scaling and temperature parameters to refine the weighting of informative temporal features. Bidirectional recurrent modules capture dependencies across the 12-month input window. Three model variants were developed using the bidirectional gated recurrent unit (BiGRU), bidirectional long short-term memory (BiLSTM), and a hybrid BiGRU–BiLSTM module. To improve robustness and spatial generalization, a stacking ensemble combines the proposed deep model with four machine learning regressors. Shapley additive explanations (SHAP) and gradient-weighted class activation mapping (Grad-CAM) provide model interpretation by identifying influential predictors and relevant temporal segments, respectively. The models were trained using monthly ERA5-derived data from eight representative cities in Türkiye. Their performance was evaluated on a national-scale dataset covering 81 cities. Among the individual architectures, CNNASDPA-BiGRU provided the best balance between predictive accuracy and computational efficiency at the reference stations. It achieved R^2=0.973 and an RMSE of 1.058 kg m ^-2 . At the national scale, the stacking ensemble achieved the best overall performance, with an RMSE of 1.358 kg m ^-2 . SHAP identified air temperature and dew point temperature as the dominant predictors. Grad-CAM highlighted temporal segments associated with pronounced moisture variability and dry-wet transitions. Overall, the proposed framework provides an accurate, interpretable, and scalable approach for monthly PWV forecasting over Türkiye.
Based on the data from the Ningxia and Guangxi Fast Antenna Lightning Mapping Arrays (FALMA), the time-reversal (TR) method is applied to locate discharge pulses of intracloud flashes. Monte Carlo simulations show that horizontal errors exhibit little dependence on source altitude, whereas vertical errors decrease with increasing altitude. Higher signal-to-noise ratio systematically improves both horizontal and vertical accuracy. The preliminary breakdown (PB) with temporally separated pulses and recoil leader (RL) with clustered pulse sequences are selected to compare the locating performance between TR and TOA. The results indicate that the average 3D difference of two locating methods is much larger for RL than for PB (X: 1.14 vs. 0.13 km; Y: 2.16 vs. 0.05 km; Z: 4.36 vs. 1.76 km), suggesting that localization performance depends on the characteristics of the lightning discharge process. Additionally, by analyzing the stacked amplitude, defined as the peak amplitude of the waveform obtained by coherently stacking the time-reversed signals after back-propagation to a candidate source position, it is found that 62.5
This study examines the impact of land surface processes on the initiation and development of atmospheric convection during a severe pre-monsoon thunderstorm event that occurred over eastern India on March 17, 2019. Utilising the Weather Research and Forecasting (WRF) model at a high spatial resolution of 1 km × 1 km, we examined the performance of two land surface schemes Noah and Noah-MP in simulating key land–atmosphere interactions, surface energy fluxes, and precipitation. Results indicated that soil moisture plays a pivotal role in modulating surface energy partitioning and near-surface atmospheric conditions. The Noah scheme, with its enhanced representation of soil moisture, produced higher latent heat flux and near-surface humidity, along with lower land surface temperatures. These conditions facilitated more realistic convection and rainfall, closely aligning with observations. On the other hand, the dry bias of Noah-MP led to higher surface temperatures, elevated sensible heat flux, and a deeper planetary boundary layer. An overall increase in instability, accompanied by a drop in evaporative cooling and relative humidity, which suppressed convection, resulted in an underestimated intensity of the storm and precipitation. An analysis of error metrics, such as bias, MAE, and RMSE, reveals that the Noah scheme generally surpassed Noah-MP in performance across the assessed land surface variables. Specifically, for soil temperature, Noah consistently achieved lower MAE and RMSE values at both Alipore (4.29 and 5.53) and Jamshedpur (5.11 and 6.66) compared to Noah-MP. A similar enhancement was observed in sensible heat flux, where Noah significantly reduced the RMSE to 105.42 W m⁻2, much lower than the 159.24 W m⁻2 recorded with Noah-MP at Alipore. Regarding latent heat flux, Noah-MP showed the greatest underestimation, whereas Noah demonstrated a relatively smaller bias (−132.78 W m⁻2) and RMSE (215.86 W m⁻2), indicating a generally improved depiction of surface energy distribution. It also more accurately captured the diurnal evolution of the boundary layer, particularly its post-convection descent. This study underscores the sensitivity of convective storm simulations to land surface parameterisations, emphasising the importance of accurate soil moisture and energy flux representation in high-resolution weather modelling. Improved storm forecasts can enhance early warning systems, reducing economic losses and safeguarding vulnerable communities.
Continuous precipitation data are essential for time-series analysis and accurate prediction of hydro-meteorological disasters. However, data gaps introduce significant uncertainty in the assessment of extreme hydro-meteorological events. This study addresses this issue by evaluating statistical methods and gridded precipitation datasets (GPDs) for infilling missing daily rainfall at five target stations in Himachal Pradesh, located in the Himalayan region of India. Statistical approaches utilized observed data from 27 neighboring stations to improve spatial estimation. The accuracy of imputation was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Nash-Sutcliffe Efficiency (NSE), followed by the computation of extreme precipitation indices (EPIs) from the infilled datasets to examine their reliability in representing precipitation extremes. Results indicate that statistical approaches generally outperformed GPDs in reconstructing daily rainfall, with Multiple Linear Regression (MLR) showing the highest overall accuracy across stations. Among gridded datasets, IMD showed the closest agreement with observations, while ERA5-Land and PERSIANN exhibited moderate performance and CHIRPS consistently showed the lowest agreement for daily rainfall reconstruction. MLR also demonstrated relatively better performance in reproducing intensity-based EPIs, while neighbour-based methods performed reasonably for frequency and duration indices. Performance varied with elevation, inter-station distance, and terrain complexity, highlighting the strong influence of orographic processes on precipitation variability in the Himalayas. These findings provide important hydro-meteorological insights and emphasize the need to consider terrain characteristics and station relationships when addressing data gaps and analyzing extreme precipitation in mountainous environments.
Rainfall forecasting plays a vital role in water resource management, agriculture planning, and disaster preparedness. Holt Winters method relies solely on past values of a single variable. However, rainfall is influenced by multiple climatic factors, including temperature and humidity. In this study, we introduce two novel multivariate extensions of the Holt Winters method—Multivariate Additive Holt Winters (MAHW) and Multivariate Modified Holt Winters (MMOHW)—to enhance the accuracy of long range rainfall prediction. Using monthly data on rainfall, temperature, and humidity from Mizoram (1986–2023), the proposed models were evaluated against traditional univariate Holt Winters, Multiple Linear Regression (MLR), and Vector Auto Regression (VAR) methods by splitting the data into 80
