Spatial Variations Analysis of Flood hazard Susceptibility based on a new ensemble model (Case Study: Aland Chai Basin, Khoy city)

Document Type : Applied Article


1 PhD Student in Geomorphology, Department of Geomorphology, University of Tabriz

2 Professor of Geomorphology, Department of Geomorphology, University of Tabriz

3 Associate Professor of Geomorphology, Department of Geomorphology, University of Tabriz

4 Associate Professor of RS and GIS, Department of GIS and RS, University of Tabriz


Flood is a disaster which causes a lot of economic damages to farmlands, forests, gas and power transmission lines, roads, engineering structures, and buildings. There are numerous floods in the northwest of country at the beginning of spring and the start of spring rains, which in most cases results in heavy damage [4]. The aim of the present study was to prepare a map of spatial variations in flood risk susceptibility in the Aland Chai basin located in West Azerbaijan province and Khoy city. To achieve this aim, 13 effective parameters in the occurrence of this phenomenon have been used. These parameters include lithology, soil hydrological groups, NDVI, land use, slope, aspect, elevation, distance to river, river density, precipitation, topographic wetness index, stream power index, and sediment transport index.
Study area
Aland Chai basin is located between 38°- 30¢-14² and 38°- 48¢-22² N and between 44°- 15¢- 13² and 45°- 01¢-02² E in the Northwest of Iran and the Western Azerbaijan province. This basin is one of the sub-basins of the Aras River basin to which surface water flows after joining the grand Qotour River. Basin elevation variations are from 1093m in the Aland Chai River bed to 3638m above sea level in the Avrin Mountain [4].
Materials and Methods
The following data, software, and methods were used to analysis flood risk susceptibility and prepare flood risk maps in the study area:
- A frame of Landsat 8 satellite image OLI scanner with path of 169 and row of 33, in 30m spatial resolution
- Geological maps in 1:100000 and 1:250000 scale from Khoy and Dizaj
- Topographic map in 1:50000 scale from Khoy city
- Digital Elevation Model (DEM) in 12.5m spatial resolution
- ArcGIS software to generate maps
- ENVI software for land use mapping
- WEKA software to data mining
The new ensemble model of FURIA-GA-AdaBoost have been used to investigate the role of parameters in the occurrence of floods. FURIA is a fuzzy rule-based classification method, an extension of the Repeated Incremental Pruning to Produce Error Reduction (RIPPER) rule learner (Cohen, 1995), introduced by Hühn and Hüllermeier (2009) [1, 3]. AdaBoost is a machine learning algorithm introduced by Freund and Schapire in 1997 [2].
Discussion and Results
To implementation the FURIA-GA algorithm, the following characteristics were obtained after trial and error. For the GA, crossover probability was set to 0.2, mutation probability was set to 0.035, population size at 250, and the number of generations set to 50. For the FURIA evaluator, a 10-fold cross-validation technique with T-Norm product as a fuzzy aggregation operator was trained to combine rule antecedents. The results showed that the FURIA-GA classification with 86.45% was very accurate. The following settings are used to run the AdaBoost algorithm: batchsize, 100; number of iterations, 12; seed, 1. Decision tree C4.5 was also selected as the base classifier. WEKA software was used to perform these algorithms.
The present study was an attempt to investigate the susceptibility of flood risk in the Aland chai basin. Therefore, 13 effective parameters in flood occurrence were used to prepare a flood risk susceptibility map. ArcGIS and ENVI software were used to prepare each of the information layers. In order to perform the relevant analyzes on each of the parameters, the new ensemble model FURIA-GA-AdaBoost in WEKA software was used. The results of these studies showed that slope, soil hydrological groups, altitude, vegetation, and land use have an important role in the occurrence of floods in the area. Flood risk susceptibility map was prepared in 5 classes of very low, low, medium, high, and very high susceptibility. The results showed that the areas that are highly susceptible in terms of flood risk are mainly concentrated downstream of the basin, which includes flat and low areas. Generally, 26% of the total area of the Aland Chai basin is located in high and very high risk for floods.


[1]. Cohen, W.W., (1995). “Fast effective rule induction”, In: Prieditis, A., Russell, S. (Eds.), Proceedings of the 12th International Conference on Machine Learning, 115–123. Morgan Kaufmann.
[2]. Freund, Y., Schapire, R., (1997). “A decision-theoretic generalization of on-line learning and an application to boosting”, Journal of Computer and System Sciences, 55(1): 119-139.
[3]. Hühn, J., Hüllermeier, E., (2009). “FURIA: an algorithm for unordered fuzzy rule induction”, Data Mining and Knowledge Discovery, 19(3): 293–319.
[4]. Rezaei Moghaddam, M.H., Hejazi, S.A., Valizadeh Kamran, K., Rahimpour., (2020). “Analysis of Hydrogeomorphic Properties of Aland Chai Basin to Prioritize Sub-Basins in terms of Flood Sensitivity”, Geography and Environmental Hazards. 33: 15-20.