Offering flood prevention solutions using remote sensing and approaches integrating fuzzy logic and agent-based modeling

Population growth, urbanization and land use change in recent decades have made floods one of the most devastating natural disasters in the world. Therefore, understanding this phenomenon, its effects and methods used to deal with it is considered to be among the most important issues crisis management planners and policymakers in urban and rural areas should pay attention to. Iran faces many natural disasters among which flood is one of the most serious ones. Monitoring and controlling accidents, assessing damages and providing relief are among the main concerns of government and crisis management experts. Continuous monitoring before the occurrence, and accurate assessment during and after the event can decrease damages to human and natural resources. Preventing flood related hazards, organizing and managing flood water in channels and ultimately improving channels require identifying and determining flood zones.
 
Materials & Methods:

Agent-based modeling (ABM) provides simulation and abstract systems used to identify patterns of land forms in the study area. As a new approach, agent-based modeling is used to develop simulation tools for complex phenomena in various fields such as natural disasters, biological studies and relief provision in flood occurrences. In fact, agent-based modeling (ABM) has been increasingly used to confront the risk of flood and its challenges in recent years. The present study applies fuzzy inference approach (using parameters affecting the occurrence of flood and remote sensing data) and agent-based modeling to prepare a flood risk map and provide a deterrent solution for flood risk management and decision making before the occurrence. In the fuzzy inference system, various maps are prepared showing parameters affecting the occurrence of floods such as slope, soil type and rivers. Then, Fuzzy Overlay model is used to define the flood risk zones and overlay the fuzzy parameters. The present study applies fuzzy gamma operator with a coefficient of 0.8 in the final fuzzy overlay calculation.
 
Results & Discussion:

Comparing the results obtained from overlaid maps reveals that most flood plains are located in areas covered with Affisols (clay-rich soil) and low-lying arable lands and orchards. In agent-based modeling, GIS plugin of NetLogo was used to investigate the flood phenomenon based on the digital elevation model of the area. In this model, raindrop cycle was simulated in the DEM raster layer of Gilan. DEM layer can be used to calculate the slope (vertical angle) and slope direction (horizontal angle) of the ground surface. Simulated images shows the movement and accumulation of agents along the rivers and their surroundings and in low altitude areas. Analysis confirms the risk of floods in rivers and low-lying areas. Finally, georeferenced images of points in risk of possible flood (agents in the slopes of the study area), land use map and soil cover map can be overlaid to evaluate the obtained results. Results indicate that the highest number of agents (white markings on the map) are located in agricultural land use covered with Affisols while a relatively moderate number of agents are located in agricultural lands covered with Inceptisols. As previously mentioned, these agents simulate the amount of runoff accumulation due to atmospheric precipitation. Results indicate that precipitation models simulated using artificial intelligence lead to almost the same result Fuzzy analysis method shows (regarding the prediction of flood occurrence).

Finally, these two approaches are compared and their functions are examined. It should be noted that multi-criteria methods such as fuzzy inference approach has a higher level of complexity and accuracy, while methods based on artificial intelligence and agent-based modeling are faster. On the other hand, agent-based modelling method use relatively ready programs and thus has a lower level of complexity. The level of accuracy in this method is also lower than the fuzzy logic method.