High-Precision AI-Enabled Flood Prediction Integrating Local Sensor Data and 3rd Party Weather Forecast

Abstract

Flooding risk is a threat to many sea-level cities and residential areas in the world. In the city Kristianstad in southern Sweden, a large number of sensors of different types have been deployed to monitor rain and other weather conditions, water levels at sea and lakes, ground water levels, and water flows in the city’s storm-water and sewage systems. All the sensors are enabled by battery and wireless communication, and allow real-time data to be transferred and visualized on a cloud-based Internet of Things (IoT) portal. To better enable the system with capacity of foreseeing upcoming flooding threats and to allow early response from decision-makers, it is desired to build a real-time flood forecast system by utilizing the massive sensor data collected at the IoT portal and data from 3rd party weather forecast service. In this article, we have developed a smart flood forecast system using machine learning and artificial neural networks. The developed forecast system has successfully integrated data from multiple sources and can make accurate flood forecast at distributed locations for the coming days. After being successfully implemented as software product and integrated with the city’s IoT portal, our developed flood forecast system has significantly extended the basic monitoring functions of the city’s IoT infrastructure. This article presents the context of this work, the challenges that have been encountered during our development, our solutions and performance evaluation results. To the best of our knowledge, this is the first large-scale IoT-based real-time flood forecast system that has been enabled by artificial intelligence (AI) and deployed in real world.

Sensors on the Internet of Things Systems for Urban Disaster Management: A Systematic Literature Review

“…People develop flood detection systems based on more considerations such as rainfall [40][41][42][61][62][63][64], water level [40,43,44,46,47,62,63,[65][66][67][68], water pressure [41,44,45,47,50,69], soil moisture [42], solar radiation [42], vapor pressure [42], relative humidity [42], humidity [42], temperature [42,61,62,70], air pressure [42,62], wind speed [42,61], wind gust [42], wind direction [42], tilt [42], lighting [42], lighting average distance [42], the flow velocity [42,62]. Moreover, Ragnoli et al [50] also used GSM to detect locations in their flood monitoring systems.…”

“…Cellular communication technologies include GSM [57], GPRS [57,71], and 3G [42]. The communication protocols include Choco protocol [39], UDP [41], IPv6 with LoWPAN [41], IEEE 802.15.4 [16,41,52], Message Queuing Telemetry Transport (MQTT) [42,43,47,59,67,81], concurrent multipath transfer protocol [49,82], LoRaWAN [50,62,80], TCP/IP Internet protocol [50,83], Hyper Text Transfer Protocol (HTTP) [61], Zigbee [16,45,71], LRWiFi [59], Cat-M1 [59], CoAP [59], XBee [70]. The data is usually transmitted in JSON format [42,50,61].…”

“…Regarding the connectivity solutions for landslides, we can use Wi-Fi [55,56], cellular communication technology [57,59], MQTT [59], LRWiFi [59], Cat-M1 [59], and CoAP [59]. The connection solutions for floods include UDP [41], IPv6 with LoW-PAN [41], IEEE 802.15.4 [41], MQTT [42,43], Ethernet [42,43], cellular communication technology [42,71,80], Wi-Fi [43,68,72], concurrent multi-path transfer protocol [49], LoRaWAN [50,62,80], TCP/IP [50,83], radio [44], HTTP [61], ZigBee [45,71], 6LoW-PAN [74], and XBee [42,70]. Concurrent transfer can achieve higher throughput [49,82], accelerate transmission [49,82], reduce packet loss [49,82], and save energy [39].…”

“…To solve this problem, Park and Baek [75] suggested some quality management for sensor networks, such as data pre-processing (time allocation and filling short gaps in missing data), physical limit check, climate range check, internal consistency check, persistence check, step check, spatial consistency check, spatial outlier check, and data reconstruction using spatial and temporal gap-filling. In addition, Wang and Abdelrahman [62] suggested using a divide-and-conquer approach to process the high-dimensional data inputs from sensors. For example, we can group the sensors by their physical locations and customize the model to process each sensor.…”

“…Regarding data analysis methods, we can use some advanced techniques, such as machine learning [38,45,62,70,[85][86][87], deep learning [46,73], and time-series data analysis [58]. Some studies may use traditional methods, such as mathematical modeling (e.g., Markov Process, Laplace Transformation) [68], observing the signal (e.g., flare marker buoys) [76], and comparing the current situation with past cases [84].…”