Flood Detection in Urban Areas Using Satellite Imagery and Machine Learning

Abstract

Urban flooding poses risks to the safety of drivers and pedestrians, and damages infrastructures and lifelines. It is important to accommodate cities and local agencies with enhanced rapid flood detection skills and tools to better understand how much flooding a region may experience at a certain period of time. This results in flood management orders being announced in a timely manner, allowing residents and drivers to preemptively avoid flooded areas. This research combines information received from ground observed data derived from road closure reports from the police department, with remotely sensed satellite imagery to develop and train machine-learning models for flood detection for the City of San Diego, CA, USA. For this purpose, flooding information are extracted from Sentinel 1 satellite imagery and fed into various supervised and unsupervised machine learning models, including Random Forest (RF), Support Vector Machine (SVM), and Maximum Likelihood Classifier (MLC), to detect flooded pixels in images and evaluate the performance of these ML models. Moreover, a new unsupervised machine learning framework is developed which works based on the change detection (CD) approach and combines the Otsu algorithm, fuzzy rules, and iso-clustering methods for urban flood detection. Results from the performance evaluation of RF, SVM, MLC and CD models show 0.53, 0.85, 0.75 and 0.81 precision measures, 0.9, 0.85, 0.85 and 0.9 for recall values, 0.67, 0.85, 0.79 and 0.85 for the F1-score, and 0.69, 0.87, 0.83 and 0.87 for the accuracy measure, respectively, for each model. In conclusion, the new unsupervised flood image classification and detection method offers better performance with the least required data and computational time for enhanced rapid flood mapping. This systematic approach will be potentially useful for other cities at risk of urban flooding, and hopefully for detecting nuisance floods, by using satellite images and reducing the flood risk of transportation design and urban infrastructure planning.

An Integrated Artificial Intelligence of Things Environment for River Flood Prevention

“…The approach of Khalaf et al [ 57 ] was third with 85.9% accuracy. The remaining approaches [ 26 , 41 , 44 ] all had an 85% accuracy.…”

“…The LSTM algorithm has provided the best flood level predictions. Tanim et al [ 44 ] put forward an unsupervised Machine Learning approach based on a combination of the Change Detection method, the Otsu algorithm, and fuzzy rules. This approach was tested on flood images captured by the ESA’s “Sentinel 1” satellite and has proved better classification performance when compared to the Random Forest (RF) algorithm, Support Vector Machines (SVM), and the Maximum Likelihood Classifier (MLC).…”

Flood extent delineation using Sentinel-1 data as information source: systematic review of processing methods

“…The most common ML classification methods are Artificial Neural Networks (ANN), k-Nearest Neighbors (kNN), Decision Trees (DT), Support Vector Machines (SVM), Random Forest (RF) (Hassan et al, 2020;Antzoulatos et al, 2022;Tanim et al, 2022). The most common ML methods found in this review are Support Vector Machine and Random Forest models.…”

“…The tools for thermal noise removal and border noise removal are optional steps, the last one used when dealing with subset images, cropped from bigger data. Thermal noise is background energy caused by the microscopic motion of electrons due to temperature and can also be removed in pre-processing (Phan et al, 2019;Tanim et al, 2022).…”

“…It is possible and has become very common when the test of several scenarios with different methods that originally were seen apart in literature are united for flood classification improvements or in extensive comparison procedures. In Tanim et al (2022), for example, a fuzzy classification made use of a set of rules based on thresholding values to delimitate flood class features, while a machine learning model was set up to be compared with the fuzzy results. In this case, the authors called the machine learning models supervised versions of classification, against the unsupervised method presented by a fuzzy model.…”