Intelligent dynamic flood response and recovery strategy
Lead Research Organisation:
CRANFIELD UNIVERSITY
Department Name: School of Water, Energy and Environment
Abstract
iD-FRe2S explores the combined use of Unmanned Aerial Vehicles (UAV) and Artificial Intelligence (AI) to assess the effectiveness of drains to support surface water management. Efficient surface water management depends on a number of factors; one of the primary factors is the microtopography that influences the surface water flow path, direction, and velocity. Proper placement of drains at optimal locations is essential to ensure effective drainage of surface water into the subsurface sewer system to be transported to the treatment plant or discharged into a river. With advancements in technology, UAV now enable the capture of high-resolution data, including imagery in the visual spectrum and other products, which together provide detailed topographic information and realistic visual context. By combining high-resolution topographic data with imagery and AI techniques, we can generate valuable insights into managing surface water flooding.
In this project, we assessed the accuracy of UAV-derived Digital Elevation Models (DEMs) based on imagery and compared them to widely-used manned aerial LiDAR data to characterise microtopographic flood features such as drain points, floodgates, brick wall kerbs, road surfaces, and vegetated embankments. Given the critical role of storm drains in removing surface water, their optimal placement at low-lying areas and along natural flow paths is essential. Accurate identification of storm drain locations is therefore a priority. Leveraging the high-resolution RGB imagery from UAVs, we applied machine learning algorithms to detect and map storm drains. Once they were mapped, we explored the application of topographic indices maps generated from UAV DEMs to assess the spatial distribution of storm drains and assess their effectiveness. Further, we aim to investigate the effect of inefficient storm drains on a street-level flood risk. The specific objectives of the project are:
Objective 1- To critically evaluate the accuracy of UAS RGB-based DEM(Digital Elevation Model) and manned aircraft LIDAR DEM in identifying microtopographic features that influence surface water flooding
Objective 2- To evaluate the application of random forest and OBIA(Object based Image analysis) techniques on RGB imagery to map storm drain locations in urban areas.
Objective 3- To test the application of topographic indices from UAS-DEM to appraise the spatial distribution of storm drains and to assess their effectiveness for surface water flood management.
Objective 4: To investigate the impact of ineffective storm drains on street-level flood risk mapping.
The findings from this study can support decision-making for the maintenance and operation of drainage infrastructure. Additionally, the insights gained can inform the design and implementation of sustainable urban drainage systems (SuDS) based on Low Impact Development (LID) principles.
In this project, we assessed the accuracy of UAV-derived Digital Elevation Models (DEMs) based on imagery and compared them to widely-used manned aerial LiDAR data to characterise microtopographic flood features such as drain points, floodgates, brick wall kerbs, road surfaces, and vegetated embankments. Given the critical role of storm drains in removing surface water, their optimal placement at low-lying areas and along natural flow paths is essential. Accurate identification of storm drain locations is therefore a priority. Leveraging the high-resolution RGB imagery from UAVs, we applied machine learning algorithms to detect and map storm drains. Once they were mapped, we explored the application of topographic indices maps generated from UAV DEMs to assess the spatial distribution of storm drains and assess their effectiveness. Further, we aim to investigate the effect of inefficient storm drains on a street-level flood risk. The specific objectives of the project are:
Objective 1- To critically evaluate the accuracy of UAS RGB-based DEM(Digital Elevation Model) and manned aircraft LIDAR DEM in identifying microtopographic features that influence surface water flooding
Objective 2- To evaluate the application of random forest and OBIA(Object based Image analysis) techniques on RGB imagery to map storm drain locations in urban areas.
Objective 3- To test the application of topographic indices from UAS-DEM to appraise the spatial distribution of storm drains and to assess their effectiveness for surface water flood management.
Objective 4: To investigate the impact of ineffective storm drains on street-level flood risk mapping.
The findings from this study can support decision-making for the maintenance and operation of drainage infrastructure. Additionally, the insights gained can inform the design and implementation of sustainable urban drainage systems (SuDS) based on Low Impact Development (LID) principles.
People |
ORCID iD |
| Rakhee Ramachandran (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/V519509/1 | 30/09/2020 | 29/09/2027 | |||
| 2585153 | Studentship | EP/V519509/1 | 18/01/2021 | 07/08/2025 | Rakhee Ramachandran |
| Description | In objective 1, two commonly used survey strategies used in flood risk management, namely S1 ( i.e. UAS based RGB data) and S2 ( i.e. manned aircraft with LIDAR), are compared to understand how they can be integrated to better characterize microtopographic features. The key findings untilnow are : The key findings from the data analysis are: 1. When the difference in elevation between S1 and S2 was analyzed, the median difference of overall DSM had an elevation difference of 15cm. Most of the features showed an elevation difference larger than 10cm except for kerb. This indicates S1 and S2 cannot be substituted for each other on most features. 2. In the comparison of accuracy( i.e. difference in elevation between survey strategy and ground truth data) of DSM of S1 and S2, it was found that for the overall DSM, both S1 and S2 presented similar median accuracy <8cm confirming both S1 and S2 are suitable to capture microtopography at 10cm. In addition, this study also demonstrated that the accuracy of each flood feature varied for both S1 and S2. S1 presented better accuracy at the road ( 3cm), drainage( 4cm), and top of the wall (7cm), and S2 presented better accuracy at grassland (4cm), the bottom of the wall (30 cm), top of the vegetated crest (9cm) and bottom of the vegetated crest (24cm). It is to be noted although S1 presented better accuracy at bottom of wall and vegetated crest, the median accuracy value is greater than 10cm and so S1 and S2 are not suitable to capture elevation at the microtopographic level. 3. Based on the study on elevation difference and accuracy, now we know the error and elevation difference of individual microtopographic features can differ from the overall DSM. The choice of survey strategy is quite critical to characterise microtopograhic features accurately. A sampling decision framework was developed on how to collect and combine both surveying strategies to better characterise microtopography. The first step of the framework is to identify the features of interest. As this study is focussing on characterising microtopographic features, the tolerable error or difference is targeted at 10 cm. Depending on the elevation difference of the feature, we could check if the survey strategy could be substituted for one another and based on the accuracy values, the best survey strategy can be chosen. 4. S2 systematically underestimated the elevation while S1 overestimated the elevation. That means the choice of survey strategy would affect the flood predictions as well. The implication of overestimated flood prediction would be different in different cases. The choice of survey strategy needs to be made considering the purpose of the study. |
| Exploitation Route | The framework developed in this study will help to choose the appropriate survey strategy ( S1 or S2) to capture microtopographic features. Previous studies have proved that different datasets could be merged to better represent microtopography in flood models. This framework will help to identify how the 2 different datasets could be combined or merged to improve the representation of microtopography. Improved representation of microtopography will help to improve the shallow water surface water flood models where microtopography is critical. Better knowledge of microtopography would also support spatial planning and implementation of Sustainable urban drainage systems and maintenance of existing drainage assets. It could also support the management of property-level flood resilience measures and flood insurance. Thus, this study will help in improved flood management by authorities. |
| Sectors | Communities and Social Services/Policy Environment Government Democracy and Justice |