Rainfall Model Research Articles

A Comparative Study of Machine Learning Models for Daily and Weekly Rainfall Forecasting

AbstractAccurate rainfall forecasting is crucial for various sectors across diverse geographical regions, including Uttarakhand, Uttar Pradesh, Haryana, Punjab, Himachal Pradesh, Madhya Pradesh, Rajasthan, and the Union Territory of Delhi. This study addresses the need for precise rainfall predictions by bridging the gap between localized meteorological data and broader regional influences. It explores how rainfall patterns in neighboring states affect Delhi's precipitation, aiming to improve forecasting accuracy. Historical rainfall data from neighboring states over four decades (1980–2021) were collected and analyzed. The study employs a dual-model approach: a daily model for immediate rainfall triggers and a weekly model for longer-term trends. Several machine learning algorithms, including CatBoost, XGBoost, ElasticNet, Lasso, LGBM, Random Forest, Multilayer Perceptron, Ridge, Stochastic Gradient Descent, and Linear Regression, were used in the modeling process. These models were rigorously assessed based on performance metrics from training, validation, and testing datasets. For daily rainfall forecasting, CatBoost, XGBoost, and Random Forest emerged as top performers, showcasing exceptional accuracy and pattern-capturing capabilities. In weekly rainfall forecasting, XGBoost consistently achieved near-perfect accuracy with an R2 value of 0.99, with Random Forest and CatBoost also demonstrating strong performance. The study provides valuable insights into how climate patterns in neighboring states influence Delhi's weather, leading to more reliable and timely rainfall predictions.

An asymmetric tropical cyclone rainfall model in the Northern Vietnam coast

AbstractRainfall associated with landfalling tropical cyclones (TCs) along the Northern Vietnam coast is examined to develop an asymmetric parametric TC‐induced rainfall model starting from the axisymmetric Rain‐Climatology and Persistence (R‐CLIPER) model. We recalibrated the R‐CLIPER model (original R‐CLIPER denoted as NHC) against observed rainfall patterns of 14 landfalling TCs from 2001 to 2021 in the Northern Vietnam coast, while relaxing the model's underlying linear relationships. The recalibrated R‐CLIPER (denoted as Fit‐Ax), still axisymmetric, suggests that some parameters are better correlated with the normalized maximum wind speed using logarithmic and exponential relationships. Fit‐Ax reduces the 12‐hr total rainfall overall root‐mean‐square errors (RMSEs) and Bias magnitudes in the before‐ and after‐landfall periods from NHC for the entire 500‐km TC domain. We further redistribute the Fit‐Ax rainfall intensity across the four quadrants with respect to the TC forward motion to account for the observed large asymmetry in quadrant rainfall (version denoted as Fit‐As). The vertical wind shear (VWS) and landfall (before or after) are considered in this redistribution. Fit‐As generally outperforms Fit‐Ax and NHC in reproducing the observed rainfall distribution for the 14 TCs. At the quadrant level, both Fit‐Ax and Fit‐As show significant improvement in Bias over NHC. Fit‐As is further better overall in RMSE and Skill when weighted by quadrant rainfall volume. In pattern matching, Fit‐As produces the best grid‐averaged Pearson correlation coefficients for 11 TCs. In addition, its equitable threat scores (ETSs) are best beyond the 20‐mm rainfall threshold, with the maximum of 0.299 at the 90‐mm rainfall threshold. Thus, our locally fitted asymmetric rainfall model demonstrates improved capability in reproducing the historical TC‐induced rainfall along the Northern Vietnam coast.

Fractional Poisson and Gamma Models for Rainfall: Implications Climate Change and Marine Ecosystems

This research explores the benefits of using fractional Poisson and fractional Gamma models in rainfall modeling, highlighting their advantages in handling zero-inflated data, reducing overdispersion, and providing greater flexibility and accuracy. The second part of this study delves into the dynamic interplay between oceanic ecosystems and global climate change. It focuses on the role of phytoplankton in oxygen production and the impact of warming waters on this delicate balance. By employing mathematical models integrating differential equations and Brownian motion, the study offers a comprehensive framework for understanding how varying rates of oxygen production influence the sustainability of oceanic ecosystems. Finally, the research incorporates fractional Brownian motion into modeling plankton-oxygen dynamics, addressing the limitations of traditional Brownian motion. This approach captures the long-range dependencies and persistent effects critical for predicting the response of marine ecosystems to climate change. The findings underscore the need for nuanced mitigation strategies to address the imminent risks posed by global warming on marine life and atmospheric oxygen levels.

Ensemble of artificial intelligence and physically based models for rainfall–runoff modeling in the upper Blue Nile Basin

ABSTRACT This study investigated the performance of the adaptive neuro-fuzzy inference system (ANFIS), feed forward neural network (FFNN), Soil and Water Analysis Tool (SWAT), Hydrologic Engineering Center's Hydraulic Modeling System (HEC-HMS), Hydrologiska Byråns Vattenbalansavdelning (HBV), and support vector regression (SVR) models for rainfall–runoff modeling using gauged and satellite rainfall, and their fusions in the Gilgel Abay watershed, Ethiopia. Afterward, simple average ensemble (SAE), weighted average ensemble (WAE), and neural network ensemble (NNE) techniques were applied to combine the outputs of individual models under three scenarios. The performance of the models was evaluated using Nash–Sutcliffe efficiency (NSE) and root mean square error (RMSE). The results demonstrated that the ANFIS model outperformed all the other single models with validation stage NSE values of 0.864 and 0.875, and RMSE values of 23.58 and 21.84 m3/s for gauge and fusion rainfall data, respectively. Among the physical-based models, SWAT gave better modeling performance with the validation stage NSE values of 0.81 and 0.821 for gauge and fusion rainfall data, respectively. Moreover, an ensemble of artificial intelligence and physical-based models greatly improved the overall modeling performance. The NNE improved the performance of single models up to 15.7 and 21.2 5% for fusion and satellite-based rainfall modeling, respectively.

Novel optimized coupled rainfall model simulation based on stepwise decomposition technique.

Precipitation forecasting plays a pivotal role in guiding the effective management of regional water resources and providing crucial warnings for regional droughts and floods. Finding a monthly precipitation simulation model with robust fitting performance is a significant research endeavor in practical precipitation prediction. This paper introduces two modified African vulture optimization algorithms (MAVOA1 and MAVOA2). It provides hyperparameter optimization techniques for the least squares support vector machine (LSSVM), long short-term memory neural network (LSTM), and random forest (RF) models. These techniques are used to construct a monthly precipitation simulation model based on algorithmic optimization coupled with variational mode decomposition for full decomposition. The test results at five typical stations in the North China Plain reveal the following: (1) the LSSVM model demonstrates significantly better performance than the LSTM and RF models. (2) the MAVOA2-LSSVM model has the best-integrated effect: the average test fitting error is RMSE = 17.50 mm/month, MRE = 117.25%, NSE = 0.90, which shows its superiority in practical application and can significantly improve the accuracy of precipitation prediction; MAVOA2 is more suitable for machine learning models with more hyperparameters of its own, which provides a reference for hyperparameter optimization algorithms in the other fields.

Derivation of Rainfall Intensity-Duration- Frequency (IDF) Curves of Gainesville, Georgia, United States

This research presents a set of Intensity-Duration-Frequency (IDF) curves of Gainesville, Georgia, USA for duration of 15 minutes to 24 hours, and return periods of 2 to 100 years. The objective of this research focuses on the development of IDF curves for the Gainesville, Georgia, USA. Gumbel and Log Pearson Type III models were used to develop the IDF curves. The best rainfall model between Log Pearson III and Gumbel model for the predication of rainfall in the study area was determined for the different return periods and durations. The Gumbel and Log Pearson Type III distributions give maximum intensities of 138.17 mm/hr and 137.75 mm/hr at return period of 100 years with duration of 0.25 hours, respectively. The result of this study can be used by water resource planners and managers for the city of Gainesville, Georgia.

Research on landslide hazard assessment based on improved analytic hierarchy process optimizing multiple rainfall indicators

Rainfall is the key factor that leads to landslide, so this study introduces multiple rainfall indexes to optimize the rainfall model in view of the single evaluation index of the rainfall model for landslide hazard assessment. In this study, Xiangxiang City of Hunan Province was selected as the study area, and eight types of susceptibility assessment factors including slope, aspect, elevation, normalized vegetation index (NDVI), road, fault, lithology and land use were extracted. By analyzing the characteristics of local rainfall, six types of rainfall induced assessment factors were selected for hazard assessment of the study area. The two types of evaluation factors were substituted into the improved AHP and RF combined weighting models respectively to obtain the susceptibility zoning map and rainfall induced model of the study area, and finally superimposed to obtain the hazard zoning map of the study area.Using ROC curve and hazard zoning in the studied area test results, the results show that:The AUC value of the multi-rainfall index is 17.7% higher than that of the single rainfall index, and the AUC value of the improved AHP is 6% higher than that of the traditional AHP method. It is verified by the disaster points on the day of extreme rainfall in the study area, and the actual occurrence of the disaster points is basically consistent with the hazard evaluation and zoning of the multi-rainfall index. Therefore, the rainfall model of landslides is optimized by using multiple rainfall indexes, which significantly improves the rationality of landslide hazard assessment.The study of multiple rainfall induced indicators can fill the knowledge gap in the current field, provide new insights and understanding for the field, and provide assistance for predicting and preventing landslides in related areas.

Analysis of socioeconomic and environmental growth of Muzaffarpur city using a novel rainfall and flood forecasting model

Analysis of socioeconomic and environmental growth of Muzaffarpur city using a novel rainfall and flood forecasting model

Flood prediction through hydrological modeling of rainfall using Conv1D-SBiGRU algorithm and RDI estimation: A hybrid approach

Flood prediction through hydrological modeling of rainfall using Conv1D-SBiGRU algorithm and RDI estimation: A hybrid approach

Statistical modeling and mapping of rainfall in the endorheic basins of Northern Algeria: a comparison of spatial interpolation methods

Statistical modeling and mapping of rainfall in the endorheic basins of Northern Algeria: a comparison of spatial interpolation methods

STORM v.2: A simple, stochastic rainfall model for exploring the impacts of climate and climate change at and near the land surface in gauged watersheds

STORM v.2: A simple, stochastic rainfall model for exploring the impacts of climate and climate change at and near the land surface in gauged watersheds

Establishing resilience-targeted prediction models of rainfall for transportation infrastructures for three demonstration regions in China

Rainstorm is one of the global meteorological disasters that threaten the safety of transportation infrastructure and the connectivity of transportation system. Aiming to support the resilience assessment of transportation infrastructure in three representative regions: Sichuan–Chongqing, Yangtze River Delta, and Beijing-Tianjin-Hebei-Shandong, rainfall data over 40 years in the three regions are collected, and the temporal distribution of rainfall are analyzed. Prediction equations of rainfall are established. For the purpose of this, the probabilistic density function (PDF) is assigned to the rainfall by fitting the frequency distribution histogram. Using the assigned PDF, the rainfall data are transformed into standard normal space where regression of prediction equations is performed and the prediction accuracy is tested. The results show that: (1) The frequency of rainfall in the three regions follows a lognormal distribution based on which the prediction equations of rainfall can be established in standard normal space. The error of regression shows no remarkable dependence on self-variables, and the significance analysis indicates that the equations proposed in this paper are plausible for predicting rainfalls for the three regions. (2) The Yangtze River Delta region has a higher risk of rainstorm disaster compared to the other two regions according to the frequency of rainfall and the return period of precipitation concentration. (3) Over the period of 1980–2021, the Sichuan–Chongqing region witnessed an increase in yearly rainfall but a decrease in rainstorm disasters, whereas the other two regions experienced a consistent rise in both metrics.

Daily Rainfall Forecasting with ARIMA Exogenous Variables and Support Vector Regression

There is a seasonal element every year, with the dry season often lasting from May to October and the rainy season lasting from November to April. However, climate change causes the changing of the rainy and dry seasons to be erratic, so it is necessary to anticipate weather conditions. Prediction of rainfall is used to see natural conditions in the future with time series modeling. The rainfall modeling method at the six Surabaya observation posts used is the Autoregressive Integrated Moving Average with exogenous variables (ARIMAX) and Support Vector Regression. The exogenous variable used is the captured seasonal pattern of rainfall. The SVR model uses input lags from the ARIMAX model and parameter tuning uses the Kernel Radial Based Function. Selection of the best model uses the minimum RMSE value. The results showed that the average occurrence of rain at the six rainfall observation posts occurred in January, February, March, April, November and December. The ARIMAX method in this study is well used to predict rainfall in Gubeng and rainfall in Wonorejo. The SVR input lag ARIMAX method is good for predicting rainfall for Keputih, Kedung Cowek, Wonokromo and Gunung Sari. Nonparametric methods are better used to forecast rainfall data because they are able to capture data patterns with greater volatility than parametric methods, one of which is the SVR method.

An artificial neural network and SCS–CN-based model for runoff estimation: a case study of the Peddavagu watershed

ABSTRACT The accurate prediction of runoff from rainfall events is crucial for effective water resource management, especially in regions with diverse climatic patterns like India. This study proposes a novel approach by integrating the soil conservation service (SCS)–curve number (CN) method with artificial neural networks (ANNs) to model rainfall–runoff relationships. In this research, an SCS–CN method is utilized to estimate initial runoff volumes, accounting for local soil and land-cover characteristics. Subsequently, an ANN-based model is developed to capture the complex nonlinear relationships between rainfall inputs and runoff outputs. Taking the rainfall of the watershed as the inputs, an ANN model was developed in MATLAB for runoff simulation at the main outlet of the Peddavagu watershed. A feed-forward back-propagation method is employed in the ANN model. The architecture with a 1-10-1 configuration based on tan–sig transfer function performed well in terms of MSE 0.019. The model efficiency was satisfactory with the coefficient of correlation (R), 0.96 for training, 0.98 for validation, and 0.98 for testing period. The overall value of R (0.96) indicates the utility of this ANN–SCS-based coupled model for rainfall–runoff simulation.

A hybrid CNN–RNN model for rainfall–runoff modeling in the Potteruvagu watershed of India

AbstractAccurate rainfall‐runoff analysis is essential for water resource management, with artificial intelligence (AI) increasingly used in this and other hydrological areas. The need for precise modelling has driven substantial advancements in recent decades. This study employed six AI models. These were the support vector regression model (SVR), the multilinear regression model (MLR), the extreme gradient boosting model (XGBoost), the long‐short‐term memory (LSTM) model, the convolutional neural network (CNN) model, and the convolutional recurrent neural network (CNN‐RNN) hybrid model. It covered 1998–2006, with 1998–2004 for calibration/training and 2005–2006 for validation/testing. Five metrics were used to measure model performance: coefficient of determination (R2), Nash‐Sutcliffe efficiency (NSE), mean absolute error (MAE), root‐mean square error (RMSE), and RMSE‐observations standard deviation ratio (RSR). The hybrid CNN‐RNN model performed best in both training and testing periods (training: R2 is 0.92, NSE is 0.91, MAE is 10.37 m3s−1, RMSE is 13.13 m3s−1, and RSR is 0.30; testing: R2 is 0.95, NSE is 0.94, MAE is 12.18 m3s−1, RMSE is 15.86 m3s−1, and RSR is 0.25). These results suggest the hybrid CNN‐RNN model is highly effective for rainfall‐runoff analysis in the Potteruvagu watershed.

Assessment of soil erosion and sediment yield in the Peddavagu watershed, India, using a revised universal soil loss equation model (RUSLE) and GIS techniques

ABSTRACT The present investigation was carried out within the Peddavagu watershed, which is located in India. The necessary datasets, including soil, land use land cover, rainfall, and digital elevation model, were processed and analysed within a Geographic Information System framework. To evaluate soil loss within the watershed, the present investigation employed the revised universal soil loss equation (RUSLE) model. Subsequently, the sediment yield is estimated based on the sediment delivery ratio (SDR). The average annual soil loss was estimated at 17.91 tonnes/hectare/year, which is high soil erosion risk. The RUSLE model's accuracy is 82.1%. Moreover, the findings revealed that sub-watersheds (SW) 9 and SW 3 exhibited the maximum and minimum average annual soil loss. The Peddavagu watershed's SDR was 0.210. Annually, 3.76 tonnes/hectare/year of sediment were transported to the Peddavagu watershed outlet. The findings revealed that SW 9 and SW 5 exhibited the maximum and minimum average annual sediment yield. The model's performance was evaluated by comparing its predictions with gauge data for validation. The observed actual data indicated a yield of 3.66 tonnes/hectare/year, while the model predicted a yield of 3.76 tonnes/hectare/year. This resource offers significant insights for policymakers and decision-makers on sustainable watershed management techniques.

A stochastic rainfall model for reliability analysis of rainfall-induced landslides

ABSTRACT Rainfall is an important trigger for a large number of landslides. Physics-based quantitative risk assessment of rainfall-induced landslides requires modelling of rainfall uncertainty in addition to soil parameters. Explicit modelling of rainfall uncertainty should consider the uncertainties in rainfall intensity, duration, as well as the occurrence of rainfall events. Towards this aspect, this study proposes a stochastic rainfall model, where the joint distribution of rainfall duration and intensity is constructed using copula theory and the occurrence of rainfall events is described using Poisson process. Based on the proposed stochastic rainfall model, a computationally efficient reliability method with a machine learning-based surrogate model is developed for assessing the failure probability of a slope subjected to rainfall infiltration within a given time period. An illustrative example with real rainfall data from Singapore is utilised to demonstrate the proposed approach. The results suggest that the slope failure probability is less sensitive to the uncertainty in rainfall occurrence, but highly sensitive to the adopted inter-event time definition for characterisation of the rainfall event. Overall, this study provides a useful stochastic rainfall model and some practical guidelines for quantitative risk assessment of rainfall-induced landslides.

High-Resolution Downscaling with Interpretable Relevant Vector Machine: Rainfall Prediction for Case Study in Selangor

Due to the discrepancy in resolution between existing global climate model output and the resolution required by decision-makers, there is a persistent need for climate downscaling. We conducted a study to determine the effectiveness of Relevant Vector Machine (RVM), one of the machine learning approaches, in outperforming existing statistical methods in downscaling historical rainfall data in the complex terrain of Selangor, Malaysia. While machine learning eliminates the requirement for manual feature selection when extracting significant information from predictor fields, considering multiple pivotal factors is essential. These factors include identifying relevant atmospheric features contributing to rainfall, addressing missing data, and developing a significant model to predict daily rainfall intensity using appropriate machine-learning techniques. The Principal Component Analysis (PCA) technique was employed to choose relevant environmental variables as input for the machine learning model, and various imputation methods were utilized to manage missing data, such as mean imputation and the KNN algorithm. To assess the performance of the RVM-based rainfall model, we collected a dataset from the Department of Irrigation and Drainage Malaysia. We used Nash-Sutcliffe Efficiency (NSE) and Root Mean Square Error (RMSE) as evaluation metrics. This study concluded that Relevance Vector Machine (RVM) models are suitable for forecasting future rainfall since they can support large rainfall extremes and generate reliable daily rainfall estimates based on rainfall extremes. In this study, the RVM model was employed to determine a predictive association between predictand variables and predictors.

Constructing multivariate distribution of rainfall characteristics: A Bayesian vine algorithm

Constructing a multivariate distribution model for rainfall characteristics is an important task for risk-based hydrologic design. However, most stochastic rainfall models are limited to bivariate frequentist distribution and cannot be easily generalized to multivariate (≥3) Bayesian settings. Towards this problem, this study presents a Bayesian vine algorithm. The presented algorithm translates the high-dimensional Bayesian update task into multiple lower-dimensional update tasks using the vine approach. The algorithm also computes the evidence, allowing Bayesian model selection of marginals and copulas in the joint distribution. Two numerical examples, a synthetic rainfall data from Canada, and a real rainfall data from China are utilized to demonstrate the algorithm’s capability. The results suggest that the algorithm can satisfactorily identify and characterize the underlying marginal and dependence for the rainfall characteristics. Since the high-dimensional Bayesian update task is translated into lower-dimensional Bayesian updates, the algorithm is scalable, stable, and computationally efficient in higher dimensions. A 20-year continuous hourly rainfall record can be characterized for its depth, duration, and intensity in around 60 s without any convergence or tuning issues. Overall, this study provides a powerful Bayesian tool for constructing stochastic rainfall models, which in turn can be used in multiple civil engineering disciplines concerned with rainfall interaction. This study does not consider temporal/serial dependence but only the cross-dependence among the rainfall characteristics.

BAYESIAN NEURAL NETWORK RAINFALL MODELLING: A CASE STUDY IN EAST JAVA

Rainfall is an important parameter in meteorology and hydrology, and it measures the amount of rain that falls from the atmosphere to the ground surface in liquid form. However, in the process of measuring rainfall, changes in the rainfall cycle sometimes occur due to climate change, global warming, and other factors. Therefore, this research aims to model daily rainfall using the Bayesian Neural Network (BNN) approach, combining the Bayesian Method and Artificial Neural Network (ANN). ANN is suitable for rainfall models that have intermittent characteristics. Meanwhile, the Bayesian method provides advantages in producing model parameter inferences that provide uncertainty measurements in predictions. BNN is expected to deliver better daily rainfall predictions than ANN. This research used daily rainfall data in East Jawa, and the results show that the Bayesian Neural Network produces better rainfall predictions when describing rainfall in East Java. These predictions will be very useful for the government and the people of East Java province to prevent flooding. Also, with rainfall predictions, people will know more about what crops should be planted during the rains.