Fig 1 - uploaded by Marina Campolo
Content may be subject to copyright.
Map of the River Arno basin. (Nave di Rosano closes the upper part of the basin; two dams contribute to flow regulation.)
Source publication
The basin of the River Arno is a flood-prone area where flooding events have caused damage valued at more than 100 billion euro in the last 40 years. At present, the occurrence of an event similar to the 1966 flood of Firenze (Florence) would result in damage costing over 15.5 billion euro. Therefore, the use of flood forecasting and early warning...
Contexts in source publication
Context 1
... availability of new data, mainly radar data, should allow the time of prediction of the model to be extended to operationally useful values. Figure 1 shows the River Arno basin, Italy. In this study, only the upper part of the basin, closed by the section of Nave di Rosano, a few kilometres upstream of Firenze, is considered. ...
Context 2
... is larger for each class of water level for both the calibration and the validation sets. A reduced value is found only for h ≥ 6 m in the validation set. In order to understand the effect of the output selection (single value or segment of water-level time series) on model calibration, the map of the weights of the trained nets was studied. In Fig. 10 the values of the weights for each link between input and hidden nodes are plotted. The distribution of weights is extremely irregular for the multiple- output model, while the surface is only slightly deformed for the single-output model. Furthermore, a large number of weights is near to zero in the single-output case (Fig. 10(b)), ...
Context 3
... was studied. In Fig. 10 the values of the weights for each link between input and hidden nodes are plotted. The distribution of weights is extremely irregular for the multiple- output model, while the surface is only slightly deformed for the single-output model. Furthermore, a large number of weights is near to zero in the single-output case (Fig. 10(b)), indicating that input information from the corresponding input is neglected. Obviously, the transfer function to reproduce the water-level evolution for the next six hours is more complex than the one reproducing a single future value of water level and a less expensive model may be used to predict the water level 6 h in advance. ...
Similar publications
Glacier lake outburst floods (GLOFs) that originate in the Shaksgam Valley on the northern slope of the Karakoram Mountains (China) occur frequently and they cause many local and regional disasters in the Yarkant River basin. In order to monitor and assess their hazard potential, this study used both HJ-1A/B and Landsat TM/ETM+/OLI satellite images...
Reservoir inflow forecasting is crucial for appropriate reservoir management, especially in the flood season. Forecasting for this season must be sufficiently accurate and timely to allow dam managers to release water gradually for flood control in downstream areas. Recently, several models and methodologies have been developed and applied for infl...
Citations
... A water level forecasting system in a watershed will serve as an environmental protection early warning system for flood prone areas. With enough time, predicting the watershed's level could help decision makers in alerting individuals on specific preparation measures that can potentially mitigate flood effects [4], [13]- [16]. Particularly, days-ahead water level forecasting in a watershed is needed to provide time for the authorities to take appropriate flood protection measures such as evacuation. ...
... With the implicit importance of using past data to develop predictive models of hydrological bodies, various water level forecasting techniques using Artificial Neural Network (ANN), Adaptive Neuro Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM) have been emerging as a solution for specific decision makers in engineering and disaster risk reduction management [2], [11], [14], [16]- [19]. SVM is a machine learning methodology based on statistical learning theory that represents a data-driven method for solving classification and regression tasks [3], [15]. ...
... To match the date range of these datasets from three different data sources, the study used 5 years of historical data from February 2013 to October 2018 since these are the least common available data for the three data sources. Research involved in water level forecasting had 2-4 years of historical data and was enough to obtain accurate predictions, thus the 5-year historical data was considered sufficient to be used in calibrating the prediction model [16], [39], [44] Shown in Table 1 is the acquired data from each station before and after trimming with trimmed data from the Mandulog WLMS having 264, 246 rows, Rogongon ARG with 156, 246 rows and Digkilaan ARG with 156, 279 rows. From this point onwards, the term dataset will refer to the trimmed data. ...
Accurate water level forecasting in a watershed as an early warning system is of great importance especially in flood prone areas. This could help the disaster management agencies in alerting the people with enough time as well as have a real time control of hydraulic structures to mitigate flood effects. This paper presents a strategy for days-ahead water level forecasting of a watershed that utilizes Support Vector Regression Machine (SVRM). Data preparation was conducted in order to solve anomalies in terms of the 11.52% missing data and 0.023% time inconsistences attributed primarily from the Water Level Monitoring Stations (WLMS) that serves as the sensor that captures the watershed data. Min-Max scaling method was then used in data transformation so that the dataset can be implemented by a SVRM model that uses the t – 12, t − 24, t − 72, t − 168 Architecture with a Radial Basis Function (RBF) kernel having C = 115, ε = 0.01 and γ = 0.001 as SVRM parameters. Through proper data preparation and SVRM implementation, the results which compared the actual and the forecasted water level shows a Mean Absolute Percentage Error (MAPE) of 2.186 and Root Mean Square Error (RMSE) of 0.00601. This study clearly suggests that SVRM has the potential to be a viable days-ahead water level forecasting model for a watershed.
... The most commonly applied approach for flood prediction involves the use of two models: multiple linear regression (MLR) and autoregressive moving average (ARMA) (Bougadis et al. 2005). Presently, the utilization of artificial neural networks (ANN) for predicting floods has been suggested (Campolo et al. 2003;Lekkas et al. 2004). Limitations apply to both linear and non-linear approaches when working with non-stationary data (Adamowski 2008). ...
Climate change has significantly influenced the occurrence of extreme events and their outcomes in developing countries, like Pakistan. This research investigates the impact of climate variability on the development of Intensity Duration Frequency (IDF) curves using wavelet analysis across two Pakistani cities i.e. Abbottabad and Islamabad. IDF curves are produced utilizing the Statistical Software Package (HEC-SSP) and Hydrological Engineering Center and Watershed Modeling System (WMS), where daily meteorological (i.e., rainfall, and the approx. temperature) data from 1960 to 2020 (60 years) at Abbottabad and Islamabad was gathered from Pakistan Meteorological Department (PMD). Initially, the relation between the observed data was extracted by applying the slope approaches of Mann-Kendall and Sen. Following the removal of serial correlation, generalized IDF curves are developed utilizing the time and turnaround for both stations. Finally, climate variability's impact on IDF curves was studied using wavelet analysis applied to three different pairs of input data, i.e., maximum, minimum, and mean temperatures against the developed IDF curves. Results showed that wavelet analysis are extremely useful to monitor the climate variability's influence/role on frequency and return periods of flood events (i.e., IDF curves). The developed IDF curves showed higher intensities during the monsoon period, whereas lower intensities of IDF curves are observed in other months against the 24 hours' duration of different return periods. Results also depicted that Abbottabad has experienced higher intensity of rainfall as compared with Islamabad city, which might be linked due to the changing climate and the use of land in both cities and verified by the results obtained from wavelet analysis. Increased cloud cover and precipitation resulted from the orographic effect, coupled with the influence of lower temperatures at elevated altitudes. Wavelet analysis showed a strong impact of climate (i.e., temperature) on IDF curves, where significant changes in the period and frequency are observed between the two cities. Overall, this study will be useful to understand how IDF curves are affected by climate variability while predicting the future flood events and sustainable design of urban drainage system.
... The most commonly applied approach for flood prediction involves the use of two models: multiple linear regression (MLR) and autoregressive moving average (ARMA) (Bougadis et al. 2005). Presently, the utilization of artificial neural networks (ANN) for predicting floods has been suggested (Campolo et al. 2003;Lekkas et al. 2004). Limitations apply to both linear and non-linear approaches when working with non-stationary data (Adamowski 2008). ...
Climate change has significantly influenced the occurrence of extreme events and their outcomes in developing countries, like Pakistan. This research investigates the impact of climate variability on the development of Intensity Duration Frequency (IDF) curves using wavelet analysis across two Pakistani cities i.e. Abbottabad and Islamabad. IDF curves are produced utilizing the Statistical Software Package (HEC-SSP) and Hydrological Engineering Center and Watershed Modeling System (WMS), where daily meteorological (i.e., rainfall, and the approx. temperature) data from 1960 to 2020 (60 years) at Abbottabad and Islamabad was gathered from Pakistan Meteorological Department (PMD). Initially, the relation between the observed data was extracted by applying the slope approaches of Mann-Kendall and Sen. Following the removal of serial correlation, generalized IDF curves are developed utilizing the time and turnaround for both stations. Finally, climate variability’s impact on IDF curves was studied using wavelet analysis applied to three different pairs of input data, i.e., maximum, minimum, and mean temperatures against the developed IDF curves. Results showed that wavelet analysis are extremely useful to monitor the climate variability’s influence/role on frequency and return periods of flood events (i.e., IDF curves). The developed IDF curves showed higher intensities during the monsoon period, whereas lower intensities of IDF curves are observed in other months against the 24 hours’ duration of different return periods. Results also depicted that Abbottabad has experienced higher intensity of rainfall as compared with Islamabad city, which might be linked due to the changing climate and the use of land in both cities and verified by the results obtained from wavelet analysis. Increased cloud cover and precipitation resulted from the orographic effect, coupled with the influence of lower temperatures at elevated altitudes. Wavelet analysis showed a strong impact of climate (i.e., temperature) on IDF curves, where significant changes in the period and frequency are observed between the two cities. Overall, this study will be useful to understand how IDF curves are affected by climate variability while predicting the future flood events and sustainable design of urban drainage system.
... Different researchers have compared the results obtained using these variables on different datasets in different parts of the world or studies area. Prediction of flood was done using decision tree [46][47][48]. The parameters used were temperature, water level and rainfall [46,47]. ...
... The major drawback of these works is the inadequate data set. Artificial neural network model was adopted by [4, 49,50] which accepts input features or parameters such as rainfall data, water level, hygrometric data, temperature and information on dam operation [48] The evaluation was done with accuracy and mean absolute percentage error of relatively good performance. The main drawback of these studies was high computational cost due to use of artificial neural network. ...
Flood disaster is a natural disaster that leads to loss of lives, properties damage, devastating effects on the economy and environment; therefore, there should be effective predictive measures to curb this problem. Between the years 2002- 2023, flood has caused death of over 200,000 people globally and occurred majorly in resource poor countries and communities. Different machine learning approaches have been developed for the prediction of floods. This study develops a novel model using convolutional neural networks (CNN) for the prediction of floods. Important parameters such as standard deviation and variance were incorporated in the parameters tuned CNN model that performed flood images feature extraction and classification for better predictive performance. The enhanced model was assessed with accuracy and loss measurement and compared with the existing model. The model leverage on the unique features of region of Interest aligns to resolve the issues of misalignments caused by the use of region of Interest pooling engaged in the traditional Faster-RCNN. The techniques and the developed system were implemented using a Python-based integrated development environment called “Anaconda Navigator” on Intel Core i5 with 8G Ram hardware of Window 10 operating system. The developed model achieved optimal accuracy at 200 epochs with 99.80% and corresponding loss of 0.0890. The results confirmed that predictive performance of a model can be improved by incorporating standard deviation and variance on model, coupled with its parameters tunning approach before classification.
... Different researchers have compared the results obtained using these variables on different datasets in different parts of the world or studies area. Prediction of flood was done using decision tree [46][47][48]. The parameters used were temperature, water level and rainfall [46,47]. ...
... The major drawback of these works is the inadequate data set. Artificial neural network model was adopted by [4, 49,50] which accepts input features or parameters such as rainfall data, water level, hygrometric data, temperature and information on dam operation [48] The evaluation was done with accuracy and mean absolute percentage error of relatively good performance. The main drawback of these studies was high computational cost due to use of artificial neural network. ...
Flood disaster is a natural disaster that leads to loss of lives, properties damage, devastating effects on the economy and
environment; therefore, there should be effective predictive measures to curb this problem. Between the years 2002- 2023, flood has
caused death of over 200,000 people globally and occurred majorly in resource poor countries and communities. Different machine
learning approaches have been developed for the prediction of floods. This study develops a novel model using convolutional neural
networks (CNN) for the prediction of floods. Important parameters such as standard deviation and variance were incorporated in the
parameters tuned CNN model that performed flood images feature extraction and classification for better predictive performance. The
enhanced model was assessed with accuracy and loss measurement and compared with the existing model. The model leverage on the
unique features of region of Interest aligns to resolve the issues of misalignments caused by the use of region of Interest pooling engaged
in the traditional Faster-RCNN. The techniques and the developed system were implemented using a Python-based integrated
development environment called “Anaconda Navigator” on Intel Core i5 with 8G Ram hardware of Window 10 operating system. The
developed model achieved optimal accuracy at 200 epochs with 99.80% and corresponding loss of 0.0890. The results confirmed that
predictive performance of a model can be improved by incorporating standard deviation and variance on model, coupled with its
parameters tunning approach before classification
... In water resources studies, hydrologists frequently use ANNs. Indeed, ANNs have been used in forecasting rainfall (Zhang et al. 1997;Chiang et al. 2007), sediment (Partal & Cigizoglu 2008;Jothiprakash & Garg 2009), floods (Campolo et al. 2003;Chang et al. 2007;Aziz et al. 2015), evaporation (Moghaddamnia et al. 2009) and flow (Panagoulia 2006;Pramanik & Panda 2009;Kostićet al. 2016;Veintimilla-Reyesa et al. 2016;Zemzami & Benaabidate 2016). Uysal et al. (2016) developed snowmelt models with ANNs (multi-layer perceptron (MLP) and radial basis function) for mountainous region in Turkey. ...
Estimation accuracy of streamflow values is of great importance in terms of long-term planning of water resources and taking measures against disasters such as drought and flood. The flow formed in a river basin has a complex physical structure that changes depending on the characteristics of the basin (such as topography and vegetation), meteorological factors (such as precipitation, evaporation and infiltration) and human activities. In recent years, deep and machine learning techniques have attracted attention thanks to their powerful learning capabilities and accurate and reliable modeling of these complex and nonlinear processes. In this paper, long short-term memory (LSTM), random forest regression (RFR) and extreme gradient boosting (XGBoost) approaches were applied to estimate daily streamflow values of Göksu River, Turkey. Hyperparameter optimization was realized for deep and machine learning algorithms. The daily flow values between the years 1990–2010 were used and various input parameters were tried in the modeling. Examining the performance (R2, RMSE and MAE) of the models, the XGBoost model having five input parameters provided more appropriate results than other models. The R2 value of the XGBoost model was obtained as 0.871 for the testing set. Also, it is shown that deep and machine learning algorithms are used successfully for streamflow estimation.
... In addition, there has been a rise in the application of artificial neural networks [16][17][18], weights of evidence [19], and frequency ratio (FR) [20] in recent years. FR technique resembles one of the most prominent bivariate methods. ...
... If the null hypothesis is rejected, it means the data is statistically significant. However, if it fails to be rejected, there isn't enough evidence to do so (Campolo et al., 2003). ...
Purpose: The aim of the research was to analyze the drought hazard occurrence, perceptional differences among residents to drought hazard occurrence, and explore the resilience mechanisms at household, community level, and institutional level to recover from drought hazards Theoretical framework: The research dealt with the theoretical aspects of how hazards occurrence can be observed through the perceptional differences of the residents in a particular area. Adoption of short- and long-term drought resilience mechanisms is inevitable to recover from drought hazards. Design/methodology/approach: This study is based on exploratory research design. Four ‘kebeles’ were selected considering drought classifications and 200 respondents were chosen. Descriptive statistics, Timeline Approach and ANOVA were used to analyze the data collected. Findings: Majority of respondents know about drought hazard occurrence, causes at their level, and experienced the effects of drought. There are significant perceptional differences among respondents across educational and marital status, and monthly income. Research, Practical & Social implications: Household and community participation is essential for adopting drought resilience mechanisms with the support of governmental and non-governmental agencies. Originality/value: This Study focused on drought hazard occurrence and resilience mechanisms at household and community level to recover drought hazards. The originality / value of the study lies with perceptional differences on drought hazards occurrence using the Timeline approach and ANOVA. The findings of this study would contribute to design and implement effective drought resilience mechanisms.
... Using ten FCFs, Mahmoud and Gan [39] presented a flood susceptibility index with flow accumulation, runoff and soil type as the most influential factors. These results differed from other studies mentioned and, after a sensitivity analysis, the authors concluded that flood susceptibility maps should include more than six FCFs, while other studies suggest that a reduced number of independent FCFs can achieve accurate results [40]. ...
The identification and classification of flood-prone areas comprise a fundamental step in the Flood Risk Management approach, providing subsidies for land use planning, floodproofing policies, the design of mitigation measures and early warning systems. To address this issue, a frequently used preliminary tool is the flood susceptibility mapping of a region using a range of widely available data. Therefore, the present study introduces an index-based approach able to qualitatively assess flood-prone areas, named Physical Susceptibility to Floods Index (PhySFI), based on a multi-criteria decision-making method and developed in a GIS environment. The methodology presupposes a critical discussion of variables commonly used in other flood indexes, intending to simplify the proposed representation, and emphasizes the role of the user/modeler. PhySFI is composed of just four indicators, based on physical parameters of the assessed environment. This index was developed and first applied in the city of Rio de Janeiro, as part of the Rio de Janeiro Climate Change Adaptation Plan. The validation process was based on a comparative analysis with flood extent and height simulated by the hydrodynamic modeling of four watersheds within the study area, with different urbanization processes for each one. The results indicate that the index is a powerful preliminary tool to assess flood-prone areas in coastal cities.
... Some parameters are extensively used in flood mapping based on their importance. Some studies show that accurate results can be obtained by using specific least number of variables [19]. The existing research is based on the most influential parameters namely slope, rainfall intensity, distance from rivers, soil, LULC and altitude. ...
Flooding is Pakistan's most common natural hazard, and it is exacerbated by increased rainfall and urbanization. Khyber Pakhtunkhwa (KPK), Pakistan flood-prone zones were determined by superimposing six flood parameters in an ArcGIS environment: elevation, slope, rainfall accumulation, land cover, soil geometry, and gap/buffer from water channel. Cellular automata based on artificial neural network (CA-ANN) along QGIS plugin module of Land Use Change Simulations (MOLUSCE) was used for predicting year 2050 land use, with a kappa value of 0.83. The results indicated that of the 75775 km2 land area covered by this research region, 3.37% (2553.62 km2) falls in extremely high risk, 18.44% (13972.91 km2) falls in high risk, 11.26% (8532.27 km2) falls in moderate risk, 0.51% (386.45 km2) falls in low risk, and just 66.42% (50329.76 km2) falls in very low risk areas. In KPK, like in any other place, a multi-criteria flood risk-vulnerability assessment is consequently necessary for preparation and post-hazard planning. Without a doubt, the outcomes reported here are crucial for flood risk assessments and hazard management decision-making. Key words: natural disasters; floods; remote sensing; geographic information system, multi-criteria evaluation; weighted overlay.
