Forecasting Model Research Articles

Humans vs. large language models: Judgmental forecasting in an era of advanced AI

This study investigates the forecasting accuracy of human experts versus large language models (LLMs) in the retail sector, particularly during standard and promotional sales periods. Utilizing a controlled experimental setup with 123 human forecasters and five LLMs—namely, ChatGPT-4, ChatGPT3.5, Bard, Bing, and Llama2—we evaluated forecasting precision through the absolute percentage error. Our analysis centered on the effect of the following factors on forecasters’ performance: the supporting statistical model (baseline and advanced), whether the product was on promotion, and the nature of external impact. The findings indicate that LLMs do not consistently outperform humans in forecasting accuracy and that advanced statistical forecasting models do not uniformly enhance the performance of either human forecasters or LLMs. Both human and LLM forecasters exhibited increased forecasting errors, particularly during promotional periods. Our findings call for careful consideration when integrating LLMs into practical forecasting processes.

An Integrated Artificial Intelligence Approach for Building Energy Demand Forecasting

Buildings are complex assets, characterized by environments and uses that change over time, variable occupancies, and long life cycles. They have high operational costs, mostly due to their energy requirements, and account for 30% to 40% of global greenhouse gas emissions. Consequently, substantial effort has been made to forecast their energy needs, with the scope of optimizing their economic and environmental impact. In this regard, the available literature focuses mainly on short-term modeling through the implementation of sets of physics-based equations (i.e., white-box), functional relationships between input and output variables (i.e., black-box), or a combination of both (i.e., grey-box). On the other hand, more research is required on long-term forecast models with the aim of reducing the energy needs. Within this context, this article presents an original automatic procedure for forecasting the energy needs of buildings in short- and long-term time horizons. This is accomplished by scaling an unknown facility from a similar facility that is already known and by executing a black-box approach based on machine learning algorithms. The proposed method is implemented in real case studies in Italy, predicting the energy needs (i.e., heating, cooling, and electricity) of Sant’Anna Hospital in Ferrara using the historical data of Ca’ Foncello Hospital in Treviso. The results show an adjusted coefficient of determination above 0.7 and an average error below 10% for all the energy vectors, demonstrating a feasible forecast performance with a low training set-to-test set ratio.

A Comparative Study of the Effect of Weather Conditions on Solar Irradiance Forecasting Using Various Methods in Yogyakarta, Indonesia

Abstract The government’s plan to build utility-scale solar power plants for the next decade will impact the high penetration of solar energy into the grid. This problem is associated with the intermittent nature of solar electricity, which can disrupt the grid system’s stability and reliability if it is not tackled. Furthermore, the complicated weather conditions in Indonesia make this variable challenging to predict. Therefore, one of the solutions offered is the solar irradiance forecasting technique that helps the operator during planning. This paper will look at the differences in these weather characteristics in predicting solar irradiance using three forecasting methods: LSTM, RNN, and ARIMA. The data is taken from direct measurements installed in the center of the island of Java, namely Yogyakarta. The result of this work shows that among LSTM, RNN, and ARIMA forecasting models, the highest forecast accuracy has been performed by LSTM with RMSE, MAPE, MAE, and R2 are 33.01 W/m2, 7.72%, 24.51 W/m2 and 0.96 respectively. Subsequently, partially cloudy, cloudy, or rainy weather has a worse predicting performance than weather with a sunny or clear sky. However, they are still regarded as reasonable forecasts.

Smart material selection strategies for sustainable and cost-effective high-performance concrete production using deep learning

The creation of high-performing concrete (HPC) is greatly influenced by the selection of materials, with cost and sustainability factors playing a bigger part in contemporary building techniques. To overcome these limitations, we developed a Multi-Objective Ant Colony Adaptive Dense Convolutional Neural Network (MOAC-ADenseNet) with 5-K-Fold cross validation, a dependable and precise forecasting model for the cost-effective selection of HPC material. First, we collect a concrete material dataset for evaluating the suggested method. MOAC-ADenseNet utilized Dense convolutional neural networks and ant colony optimization for complex material data analysis, which makes it easier to choose expensive and sustainable materials for high-performance concrete manufacturing operations. The experimental findings of the suggested approach are evaluated for the relative measure such as Pearson's Linear Correlation Coefficient (R) is 0.93, the Root Mean Square Error (RMSE) is 91.38, Mean Absolute Error (MAE) of 58.15, and Mean Absolute Percentage Error (MAPE) is 8.79. The outcomes demonstrated that the material cost of HPC was correctly predicted by the MOAC-ADenseNet. The actual measured value and the MOAC-ADenseNet model predictions, following 5-K-fold cross-validation and input feature improvement, shows its effectiveness. A The MOAC-ADenseNet approach provides feasible method for enhancing material selection in HPC manufacturing accomplishing sustainability and cost-effectiveness goals.

A Shared Economy Data Prediction Model Based on Deep Learning

In the dynamic and intricate shared economy, efficient resource management and forecasting are still crucial. In this research, a new prediction model is presented that aims to improve the operational efficiency of several shared economy services. Our method creatively combines Long Short-Term Memory (LSTM) networks with Genetic Algorithms (GA) to assess user demand and optimize resource deployment. We apply this methodology especially to e-scooter sharing services, but the underlying ideas and methods can be applied to other shared economy platforms as well, like peer-to-peer lending services, car sharing, and vacation rentals. The model starts with a GA to adjust the hyperparameters of the LSTM network, making the network better suited to handle specific characteristics of common economic data. Capturing the complex temporal and spatial patterns of user behavior and demand on these platforms requires this optimization. The LSTM element then predicts changes in service demand due to its capacity to analyze sequential records. Further to being useful for analyzing sequential data and predicting destiny wishes, this predictive functionality is important for a shared economy platform to correctly manage stock, allocate assets, and predict personal wishes. We use a large dataset to check our technique, demonstrating the predictive accuracy of the model and demand and its potential to aid strategic choice-making. Compared to traditional fashions, the consequences show a large development in forecast accuracy and resource allocation efficiency. Our methodology creates a robust basis for statistics-driven insights that decorate customer happiness and decorate the long-term increase of the shared financial system. This work highlights the blended ability of GA and LSTM inside the shared economy and paves the manner for future enhancements in using modern-day gadget mastering techniques to optimize and alter various shared services. In quick, effective useful resource control and forecasting in the shared economic system is tough, however our specific forecasting model combining GA-LSTM gives a manner ahead. Our technique, as it should be, predicts fluctuations in provider demand; it was first refined the usage of GA to regulate the LSTM hyperparameters. The consequences show how correct and powerful our version is and spotlight how it can enhance customer pleasure and operational performance. This research paves the manner for future trends within the software of gadgets, gaining knowledge of methods and supports the continuing enlargement and development of shared economic system offerings.

A hybrid deep learning-based short-term forecast model for ionospheric foF2 in East Asia region

The short-term forecast of the critical frequency of the ionospheric F2 layer (foF2) is particularly challenging due to its nonlinear and non-stationary characteristics. To improve the short-term forecast accuracy of foF2, we propose a short-term hybrid deep learning model that integrates the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) and long short-term memory (LSTM) algorithms. The proposed model leverages the ICEEMDAN decomposition algorithm to expand foF2 time series data into multi-dimensional space and utilizes the LSTM algorithm to preserve long-term data information, capturing detailed changes in foF2 parameters and obtaining internal information about foF2. The 2014 and 2017 foF2 data from four observation stations located in Mohe, Beijing, Jeju, and Kokubunji in the mid-latitudes of East Asia were utilized for modeling and 1-hour forecast analysis. The forecast values of the proposed model align closely with the measured data, exhibiting minimal fluctuation in forecast error and effectively tracking the changing trend of foF2. The average root mean square error (RMSE) across the four stations in 2014 is 0.40 MHz, with a relative RMSE (RRMSE) of 7.11 % and a coefficient of determination (R2) of 0.98. The average RMSE in 2017 is 0.43 MHz, RRMSE is 7.68 %, and R2 is 0.92. The proposed model reveals significant improvements in the forecast accuracy of the comparison model compared with the CCIR, URSI, NN, and LSTM models. Especially compared with the CCIR and URSI models in the International Reference Ionosphere (IRI), demonstrating superior real-time tracking capability for foF2 and effectively addressing the challenge of poor short-term forecast accuracy observed in the IRI model.

Sales Forecasting Models: Comparison between ARIMA, LSTM and Prophet

Sales Forecasting Models: Comparison between ARIMA, LSTM and Prophet

A novel hybrid model for multi-step-ahead forecasting of wind speed based on univariate data feature enhancement

Reliable multistep ahead wind speed forecasting (MAWSF) is critical for the energy management of wind farms and the long-term maintenance of wind power systems. However, relying on inherent meteorological features such as temperature and atmospheric pressure often fails to meet the deep learning model’s feature requirements for accurate wind speed forecasting (WSF). This paper introduces a hybrid multistep forecasting model that constructs a univariate wind speed feature enhancement framework, combining random forest (RF) and Transformer models for WSF. Initially, the hybrid enhancement framework decomposes the univariate wind speed data and extracts time-series features, effectively mining the latent features information. Subsequently, the RF feature selector filters out significant features contributing to WSF and eliminates redundant features to provide stable features data. Finally, Transformer model is utilized for both short-term and long-term MAWSF. This study conducted MAWSF on data with sampling intervals of 20 min, 30 min and 1 h. The results indicate that, compared to existing state-of-the-art models, the hybrid model in MAWSF tasks reduces the dependency of models on inherent meteorological features, achieving more accurate forecasting and faster computation speeds. Ultimately, the proposed model can provide reliable technical support for energy management and maintenance guidance in real-world wind farms.

Enhancing Weather Forecast Accuracy Through the Integration of WRF and BP Neural Networks: A Novel Approach

AbstractIn the past century, scholars from both domestic and international communities have delved into the study of numerical weather prediction models to promptly understand meteorological factors and mitigate the impacts of extreme weather events on humanity. Effective and precise prediction models enable the forecasting of meteorological conditions in the upcoming days, empowering individuals to implement proactive measures to minimize the adverse effects of extreme weather (Liang et al., 2021). The WRF (Weather Research and Forecasting) modeling system is commonly used for forecasting meteorological elements. However, uncertainties terribly hamper the correctness of the forecasting results. To this end, the present study was conducted to build a secondary model on the basis of the WRF forecast model. The WRF‐BPNN model was proposed for verification after constructing the network, the temperature vertical profile and the mixing ratio vertical profile were predicted, and the results on the validation set were tested. The results showed that the WRF‐BPNN model could effectively predict the temperature profile and mixing ratio profile, presenting better performance than the traditional WRF model.

Current progress in subseasonal-to-decadal prediction based on machine learning

The application of machine learning (ML) techniques to climate science has received significant attention, particularly in the field of climate predictions, ranging from sub-seasonal to decadal time scales. This paper reviews recent progress of ML techniques employed in climate phenomena prediction and the enhancement of dynamic forecast models, which provide valuable insights into the great potentials of ML techniques to improve climate prediction capabilities with reduced computational time and resource consumption. This paper also discusses several major challenges in the application of ML to climate prediction, including the scarcity of datasets, physical inconsistency, and lack of model transparency and interpretability. Additionally, this paper sheds light on how climate change impacts ML model training and prediction, and explores three key areas with potential breakthroughs: large-scale climate models, knowledge discovery driven by ML, and hybrid dynamical-statistical models, underscoring the important role of the integration of “ML and dynamical models” in building a bridge between the artificial intelligence and climate science.

Carbon Peak Prediction in China Based on Bagging-integrated GA-BiLSTM Model under Provincial Perspective

Many studies have revealed potential carbon peaking pathways in China. However, to minimize the uncertainty inherent in forecasting studies, more accurate forecasting models as well as more segmented forecasting subjects are always necessary. Based on panel data of 30 Chinese provinces from 2000 to 2020, this study applies the STIRPAT model and ridge regression to examine the key influencing factors of carbon emissions in each province in China, further constructs an innovative Bagging-integrated GA-BiLSTM model for forecasting, and finally explores China's peak carbon pathways in conjunction with scenario analysis. The results show that: (1) increasing urbanization and population growth are the main reasons driving the increase of carbon emissions in most provinces; (2) there is significant variability in the carbon peaking dynamics of China's regions, which can be classified into three categories: autonomous earlier peaking, policy-guided peaking, and difficulty in peaking; (3) For most provinces, the green scenario is more effective than the development scenario in reducing the peak carbon level rather than advancing the peak time; (4) under the development and the green scenario, China's overall carbon peak will be reached in 2032 and 2028, with the peak carbon levels of 11.88 billion tons and 11.42 billion tons, respectively.

FAFTransformer: Multivariate time series prediction method based on multi‐period feature recombination

AbstractMultivariate time series forecasting is widely used in various fields in real life. Many time series prediction models have been proposed. The current forecasting model lacks the mining of correlation between sequences based on different periods and correlation of periodical features between different periods when dealing with data. In this paper, we propose a multivariate data prediction model FAFTransformer based on the reorganization of multi‐periodic features, which first extracts the multi‐periodic information of the time series using the method of frequency domain analysis. The temporal dependencies within sequences are then captured using convolution based on different periods, and the correlations between sequences are learned by combining the multivariate attention mechanism to obtain the intra‐sequence and inter‐sequence correlations under the same period. Finally, period fusion is proposed to capture the correlation of period characteristics between different periods. The experimental results show that the model achieves the best results on multiple datasets compared to the latest seven predictive models for time‐series data.

Evaluating Machine Learning–Based Probabilistic Convective Hazard Forecasts Using The HRRR: Quantifying Hazard Predictability and Sensitivity to Training Choices

Abstract The High-Resolution Rapid Refresh (HRRR) model provides hourly updating forecasts of convective-scale phenomena, which can be used to infer the potential for convective hazards (e.g., tornadoes, hail, and wind gusts), across the United States. We used deterministic 2019–20 HRRR, version 4 (HRRRv4), forecasts to train neural networks (NNs) to generate 4-hourly probabilistic convective hazard forecasts [neural network probability forecasts (NNPFs)] for HRRRv4 initializations in 2021, using storm reports as ground truth. The NNPFs were compared to the skill of a smoothed updraft helicity (UH) baseline to quantify the benefit of the NNs. NNPF skill varied by initialization time and time of day but was all superior to the UH forecast. NNPFs valid at hours between 1800 and 0000 UTC were most skillful in aggregate, significantly exceeding the baseline forecast skill. Overnight NNPFs (i.e., valid 0600–1200 UTC) were least skillful, indicating a diurnal cycle in hazard predictability that was present across all HRRRv4 initializations. We explored the sensitivity of HRRRv4 NNPF skill to NN training choices. Including an additional year of 2021 HRRRv4 forecasts for training slightly improved skill for 2022 HRRRv4 NNPFs, while reducing the training dataset size by 40% using only forecasts with storm reports was not detrimental to forecast skill. Finally, NNs trained with 2018–20 HRRRv3 forecasts led to a reduction in NNPF skill when applied to 2021 HRRRv4 forecasts. In addition to documenting practical predictability challenges with convective hazard prediction, these findings reinforce the need for a consistent model configuration for optimal results when training NNs and provide best practices when constructing a training dataset with operational convection-allowing model forecasts. Significance Statement Convective hazards, such as hail and tornadoes, are often challenging to predict. To improve hazard predictions, we used machine learning (ML) to generate forecasts of convective hazards across the United States leveraging forecasts of prior events. The ML hazard forecasts were consistently better than a non-ML approach and varied in skill based on the time of day, with nighttime forecasts being particularly challenging. ML forecasts of wind gusts and hail were more skillful than tornadoes. Different strategies of constructing the dataset of prior events led to differences in forecast performance; thus, this work provides recommendations for how to assemble these datasets and train ML models to generate improved forecasts of severe weather events.

Time-Dependent Aftershock Probabilistic Forecasting and Ensemble Model Implementation: Insights from the 2021 Ms 7.4 Maduo Earthquake

Abstract On 22 May 2021, an Ms 7.4 earthquake with a focal depth of 17 km struck the Maduo region of Qinghai province, breaking a 3.8-year quiescence of strong earthquakes (magnitude >7.0) in mainland China. This event has increased stress on the Maqin–Maqu segment of the Kunlunshan fault, heightening the potential for a large earthquake in the region. In this study, we employed the epidemic-type aftershock sequence (ETAS) model and Reasenberg–Jones (R–J) model to fit the aftershock sequence following the mainshock, analyzing the temporal response and its ability to influence future generations. Concurrently, ensemble models were explored to leverage the strengths of both the ETAS and R–J models. Short-term forecasts of the probability and occurrence rate of aftershock events with varying magnitudes were conducted for the next three days. Then statistical methods, that is, receiver-operating characteristic diagrams and information gain, were used to evaluate the unconditional and relative performance. Our findings include that the ETAS model indicates a high decay rate with many aftershocks triggered by previous events, whereas the R–J model shows a normal decay rate with a higher proportion of strong aftershocks. The ETAS and R–J models perform better than random guesses for different aftershock magnitudes, but the ETAS model is somewhat affected by the problem of missing small earthquakes for a short period of time after the mainshock. The ensemble model that chooses the minimum strategy shows promise, especially in the early period, suggesting a reasonable and cautious decision approach should be chosen during the unstable stage. This study highlights the importance of short-term aftershock probability estimation for seismic research and decision-making. In the absence of more accurate models, current analytical approaches within the operational earthquake forecasting framework remain valuable. Continuous testing, feedback, and refinement of forecasting models, along with the development of ensemble models, are essential for enhancing seismic risk assessment and mitigation strategies.

Time Series Forecasting of Air Quality Index in Lahore : A Machine Learning Perspective with Facebook Prophet Model

This research investigates the rising problem of air pollution, which is a serious issue affecting people's health and the environment on a global scale. Focusing on Lahore, Pakistan, a city facing severe pollution, the research examines the Air Quality Index (AQI) from May 2019 to December 2023. Using the Facebook Prophet model for time-series forecasting, the analysis reveals trends at monthly, weekly, and hourly intervals. Peak AQI concentrations are identified in winter months, notably December and January, aligning with hazardous smog levels. Weekly trends show minimum AQI on weekends, particularly Saturday and Sunday. The hourly trend points to peak AQI concentrations between 7:00 and 09:00 AM, correlating with morning traffic and emphasizing the role of vehicular emissions in air quality degradation. These forecasts are crucial for anticipating and mitigating the adverse impact of air pollution on human health. This research underscores the critical need for comprehensive strategies to address air pollution, acknowledging the Lahore High Court and Government's initiatives, such as artificial rain experimentation, closure of commercial markets by 10 PM, ban on cutting trees and smoke-emitting vehicles. In summary, this research significantly contributes to understanding and tackling the increasingly urgent challenges associated with air quality.

Predictive Patterns and Market Efficiency: A Deep Learning Approach to Financial Time Series Forecasting

This study explores market efficiency and behavior by integrating key theories such as the Efficient Market Hypothesis (EMH), Adaptive Market Hypothesis (AMH), Informational Efficiency and Random Walk theory. Using LSTM enhanced by optimizers like Stochastic Gradient Descent (SGD), Adam, AdaGrad, and RMSprop, we analyze market inefficiencies in the Standard and Poor’s (SPX) index over a 22-year period. Our results reveal “pockets in time” that challenge EMH predictions, particularly with the AdaGrad optimizer at a size of the hidden layer (HS) of 64. Beyond forecasting, we apply the Dominguez–Lobato (DL) and General Spectral (GS) tests as part of the Martingale Difference Hypothesis to assess statistical inefficiencies and deviations from the Random Walk model. By emphasizing “informational efficiency”, we examine how quickly new information is reflected in stock prices. We argue that market inefficiencies are transient phenomena influenced by structural shifts and information flow, challenging the notion that forecasting alone can refute EMH. Additionally, we compare LSTM with ARIMA with Exponential Smoothing, and LightGBM to highlight the strengths and limitations of these models in financial forecasting. The LSTM model excels at capturing temporal dependencies, while LightGBM demonstrates its effectiveness in detecting non-linear relationships. Our comprehensive approach offers a nuanced understanding of market dynamics and inefficiencies.

Comparative analysis of the predictive power of statistical models of macroeconomic indicators in conditions of permanent crises

Subject. The article discusses prediction of the state of the economy, the accuracy of forecasts of traditional models during crises, the need to find more effective model specifications to predict macro indicators. Objectives. The purpose is to carry out a comparative analysis of the predictive ability of ensemble methods in comparison with a set of models, including traditional statistical algorithms and machine learning algorithms. Methods. The comparative analysis of predictive ability of the models and interpretation of results obtained were performed using a dynamic factor model (DFM), a neural network with long-term short-term memory (LSTM), and integrated gradient methods (IG). Results. We performed the analysis of predictive ability of the ensemble model to forecast GDP, which combines DFM and LSTM to account for both linear and nonlinear dependencies in the data; the analysis of predictive power of various indicators, which showed that an increase in forecast error is observed for all models except DFM, the ensemble model with an error correction structure, and ARMAX. The obtained results can be used to build models of macroeconomic indicators in order to make strategic decisions by enterprises of various industries operating in a highly uncertain environment. Conclusions. The combination of DFM and LSTM in the ensemble provides higher accuracy forecasts than LSTM and competitor models, however, with less predictive power than DFM.

Tracing of phosphorus in river sediments using the structural equation modeling

ABSTRACT Phosphorus (P) is transported into the water resources mainly due to soil erosion. Using the structural equation modeling, this study investigated the relative contribution of different P-producing sources in the Pasikhan River watershed, south of the Caspian Sea, Northern Iran. For this purpose, 79 surface soil samples and 14 suspended sediments were taken from the potential P sources and the river. These sources included undisturbed and degraded rangelands, forests, rice fields, tea gardens, and gullies. The built-in structural equations in PLS software were used for modeling purposes. The overall model fitting index (goodness of fit (GOF) = 0.591) showed the model's strong forecasting capability. Degraded rangelands, gullies, rice fields, and tea gardens significantly contribute to P in the river sediments. In particular, the highest contribution was related to degraded rangelands and gullies (intensity of the effect = 0.63 and 0.47). Finally, the results showed that gullies' contribution was 28.26% to the P production in river sediments, while the other sources had a relatively equal contribution. In conclusion, this study elucidates that undisturbed rangelands and forests exhibit negligible contributions to phosphorus levels in river sediments. Overall, the results confirmed that structural equation modeling is a robust and efficient approach to identifying P sources.

Navigating OpenDOAR Distribution: A Deep Dive into Predictive Modeling and Trend Forecasting

Effective navigation and strategic foresight are imperative in the dynamic realm of academic repositories. This study aims to comprehensively understand global scholarly communication dynamics by analyzing repository distribution and forecasting future trends using predictive modeling methodologies. Employing a mixed-methods approach, this research blends quantitative analysis and predictive modeling techniques. It encompasses phases such as distribution analysis of repositories across continents, examination of repository counts across countries using GeoPandas and Matplotlib, optimization of Long short-term memory (LSTM) networks, application of Random Forest Regressor models, and time series analysis using Autoregressive Integrated Moving Average (ARIMA) models. The study reveals profound disparities in repository counts across continents, reflecting variations in research infrastructure. Visualizations vividly depict geographical patterns and disparities in repository distribution. Predictive modeling techniques offer insights into future trends in repository development, empowering decision-makers with strategic insights. While this study provides valuable insights, limitations exist, such as the scope limited to OpenDOAR repositories and the need for further refinement in predictive modeling techniques; future research should explore additional factors influencing repository development.

HIGH-RESOLUTION SATELLITE ESTIMATION OF SNOW COVER FOR FLOOD ANALYSIS IN EAST KAZAKHSTAN REGION

The increasing frequency of extreme weather events linked to climate change has made flood forecasting an important issue, particularly in mountainous regions where snowmelt is a major driver of seasonal flooding. This study explores the application of snow cover estimation techniques to assess snowmelt dynamics and their potential impact on flood risks in the Ulba and Uba basins in East Kazakhstan. To achieve this, high-resolution multispectral satellite imagery from the Sentinel-2 Surface Reflectance dataset is used, focusing on images collected between March and October for the years 2021 to 2024. The images are processed in Google Earth engine platform with strict filtering based on spatial intersection with the basins and cloud cover pixels percentage, ensuring high-quality data for snow cover analysis. The study utilizes multiple remote sensing indices for snow cover estimation. The normalized difference snow index is calculated using the green and shortwave infrared bands to detect snow-covered pixels. Fractional snow-covered area is derived from the NDSI using the 'FRA6T' relationship, offering a more nuanced estimate of snow distribution across the basins. Additionally, a near-infrared to shortwave infrared ratio threshold is employed to minimize confusion between snow and water, improving the detection of snow cover, particularly in regions near water bodies or during melt periods. The resulting snow cover maps and fSCA estimates provide a detailed picture of snow distribution and melt dynamics, contributing to the assessment of snowmelt’s role in flood risk development. The obtained insights can assist in refining flood forecasting models, improving early warning systems, and supporting informed water resource management in vulnerable regions.