Geo-foundation models and UAV data for post flooding damage assessment in Mozambique

Visual scene understanding is the core task in making any crucial decision in any computer vision system. Although popular computer vision datasets like Cityscapes, MS-COCO, PASCAL provide good benchmarks for several tasks (e.g. image classification, segmentation, object detection), these datasets are hardly suitable for post disaster damage assessments. On the other hand, existing natural disaster datasets include mainly satellite imagery which has low spatial resolution and a high revisit period. Therefore, they do not have a scope to provide quick and efficient damage assessment tasks. Unmanned Aerial Vehicle (UAV) can effortlessly access difficult places during any disaster and collect high resolution imagery that is required for aforementioned tasks of computer vision. To address these issues we present a high resolution UAV imagery, FloodNet, captured after the hurricane Harvey. This dataset demonstrates the post flooded damages of the affected areas. The images are labeled pixel-wise for semantic segmentation task and questions are produced for the task of visual question answering. FloodNet poses several challenges including detection of flooded roads and buildings and distinguishing between natural water and flooded water. With the advancement of deep learning algorithms, we can analyze the impact of any disaster which can make a precise understanding of the affected areas. In this paper, we compare and contrast the performances of baseline methods for image classification, semantic segmentation, and visual question answering on our dataset. FloodNet dataset can be downloaded from here: https://github.com/BinaLab/FloodNet-Supervised_v1.0.

Measuring and monitoring of river channel evolution or changes under laboratory conditions is an important scope in hydraulic assessment. Measurement such as changes in river channel profile for instance, provides an important indicator on erosion and accretion rates in hydraulic modelling. Under controlled conditions, the changes in channel profile are usually measured using a high precision point gauge. However, when large numbers of points of river profile need to be measured, the use of point gauge method becomes laborious and time consuming. This study proposed a digital close range photogrammetry technique to measure natural river channel bed profile changes in laboratory. The objective of this study was to investigate the changes of physical river model profile for pre and post flooding simulations using digital close range photogrammetry technique. Small scale physical model experimental works were conducted in the Hydraulics and Hydrology Laboratory to observe the river profile evolution during the events. The flood flume utilized in this study is 4.95 m long, 1.38 m wide and 1.26 m deep with carved V-shaped natural main channel with a bed slope of 1:500. Data measured from digital close range photogrammetry technique during pre-flooding at t = 0 second and post flooding events at t= 9660 second were compared to determine the channel profile evolution or changes after simulation of flood event. The results show the changes in invert level or bed level between non-flooding and flooding events due to erosion is varies from 1 mm (minimum) to 6 mm (maximum) along the channel between chainage 0mm and chainage 2000 mm with total volume of erosion is 1157 x 103 mm3 respectively. It can be concluded that the digital close range photogrammetry technique can be used as a complimentary method to measure and monitor the changes of river channel profile in the laboratory

Abstract. Reliable information on building stock and its vulnerability is important for understanding societal exposure to floods. Unfortunately, developing countries have less access to and availability of this information. Therefore, calculations for flood damage assessments have to use the scarce information available, often aggregated on a national or district level. This study aims to improve current assessments of flood damage by extracting individual building characteristics and estimate damage based on the buildings' vulnerability. We carry out an object-based image analysis (OBIA) of high-resolution (11 cm ground sample distance) unmanned aerial vehicle (UAV) imagery to outline building footprints. We then use a support vector machine learning algorithm to classify the delineated buildings. We combine this information with local depth–damage curves to estimate the economic damage for three villages affected by the 2019 January river floods in the southern Shire Basin in Malawi and compare this to a conventional, pixel-based approach using aggregated land use to denote exposure. The flood extent is obtained from satellite imagery (Sentinel-1) and corresponding water depths determined by combining this with elevation data. The results show that OBIA results in building footprints much closer to OpenStreetMap data, in which the pixel-based approach tends to overestimate. Correspondingly, the estimated total damage from the OBIA is lower (EUR 10 140) compared to the pixel-based approach (EUR 15 782). A sensitivity analysis illustrates that uncertainty in the derived damage curves is larger than in the hazard or exposure data. This research highlights the potential for detailed and local damage assessments using UAV imagery to determine exposure and vulnerability in flood damage and risk assessments in data-poor regions.

Due to climate change, the frequency and intensity of Hydro-Meteorological disasters, such as floods, are increasing. Therefore, the main purpose of this work is to assess tangible future flood damage in the urban watershed of the To Lich River in Hanoi, Vietnam. An approach based on spatial analysis, which requires the integration of several types of data related to flood characteristics that include depth, in particular, land-use classes, property values, and damage rates, is applied for the analysis. To simulate the future scenarios of flooding, the effects of climate change and land-use changes are estimated for 2030. Additionally, two scenarios based on the implementation of flood control measures are analyzed to demonstrate the effect of adaptation strategies. The findings show that climate change combined with the expansion of built-up areas increases the vulnerability of urban areas to flooding and economic damage. The results also reveal that the impacts of climate change will increase the total damage from floods by 26%. However, appropriate flood mitigation will be helpful in reducing the impacts of losses from floods by approximately 8% with the restoration of lakes and by approximately 29% with the implementation of water-sensitive urban design (WSUD). This study will be useful in helping to identify and map flood-prone areas at local and regional scales, which can lead to the detection and prioritization of exposed areas for appropriate countermeasures in a timely manner. In addition, the quantification of flood damage can be an important indicator to enhance the awareness of local decision-makers on improving the efficiency of regional flood risk reduction strategies.

Understanding historical changes in flood damage and the underlying mechanisms is critical for predicting future changes for better adaptations. In this study, a detailed assessment of flood damage for 1950–1999 is conducted at the state level in the conterminous United States (CONUS). Geospatial datasets on possible influencing factors are then developed by synthesizing natural hazards, population, wealth, cropland and urban area to explore the relations with flood damage. A considerable increase in flood damage in CONUS is recorded for the study period which is well correlated with hazards. Comparably, runoff indexed hazards simulated by the Variable Infiltration Capacity (VIC) model can explain a larger portion of flood damage variations than precipitation in 84% of the states. Cropland is identified as an important factor contributing to increased flood damage in central US while urbanland exhibits positive and negative relations with total flood damage and damage per unit wealth in 20 and 16 states, respectively. Overall, flood damage in 34 out of 48 investigated states can be predicted at the 90% confidence level. In extreme cases, ~76% of flood damage variations can be explained in some states, highlighting the potential of future flood damage prediction based on climate change and socioeconomic scenarios.

Damage assessment of post-disaster built environments is critical in emergency response and recovery processes. Artificial intelligence is increasingly used in this field thanks to its ability to quickly analyses large, diverse, complex data and automate decision processes. This study evaluates the research trends, methods, and data utilization strategies in artificial intelligence-based structural damage assessment by examining 464 academic publications published between 2000 and 2024 through bibliometric analysis. The Web of Science Core Collection database data were analyzed using VOSviewer 1.6.20 and Microsoft Excel 2023. The analyses were based on the number of citations normalized by publication year. The findings reveal that the studies are mainly focused on earthquake-related issues, and the most common artificial intelligence techniques are supervised learning and deep learning architectures. Satellite imagery, UAV data, and 3D point clouds are the most frequently used data sources. The study summarizes the field's current state and provides directions for future research.

Floods are among the most severe consequences of climate change, causing significant damage across multiple sectors, including agriculture. Nevertheless, the assessment of agricultural flood damage remains limited, particularly in agriculturally intensive regions where timely support is crucial. This work proposes a data-driven approach for assessing crop flood damage through a machine learning classification framework applied to features derived from Earth Observation (EO) data, trained and tested on field-level damage data collected by agronomists. Specifically, we applied a Random Forest model to classify fields into three damage classes by integrating Sentinel-2–derived indices, topographic information, and flood extent maps. The analysis focused on the flood event that struck the Emilia-Romagna region (Italy) in May 2023, one of the costliest floods globally that year. The model was trained and tested on 412 fields, achieving an overall accuracy of 0.74, with a precision, recall, and F1 score of 0.75, 0.74, and 0.74, respectively. The model accurately identified high-damage fields, which were characterized by greater flood exposure, lower elevations, and pronounced declines in vegetation indices. However, it struggled to distinguish between no-damage and medium-damage fields, particularly for permanent crops, where damage often occurs beneath the canopy and flooded areas may be partially occluded. The main novelty of this work lies in the use of in situ crop damage assessments, enabling a data-driven estimation of flood impacts. These results have direct implications for policymakers: the framework relies on free EO data, providing a tool that can support post-event compensation and decision-making in flood-prone regions. • Beyond flood extent: a new approach to assessing flood damage in agriculture. • Quantifying agricultural flood damage across three severity classes. • Classifying flood damage using a random forest model trained on in-situ damage. data • A potential tool to support fair compensation for flood-affected farmers.

<p>The European Floods Directive (2007/60/EC) requires that Member States develop flood hazard and risk maps, to be used as the information basis for the development of Flood Risk Management Plans (FRMPs), and to update them every 6 years. To support such a process, the Po River District Authority signed in May 2020 an agreement with 20 Italian Universities and the Italian National Research Council (CNR) with the aim of transferring the state of the art about hydrology (including climate change), hydraulics and damage modelling into the production of the new maps, to be delivered by December 2021. This contribution describes the results obtained by the damage modelling group, composed by 8 universities and the CNR, within the project MOVIDA (MOdello per la Valutazione Integrata del Danno Alluvionale – Model for integrated evaluation of flood damage).<br>The objective of the project was to provide an Information System able to perform an analytical evaluation and mapping of expected damage, overcoming the limitations of present maps where the evaluation of risk remains highly qualitative and subjective. First, proper exposure and damage assessment tools were identified, for all the categories of exposed elements included in the Directive: population, infrastructures, economic activities (classified as: residential buildings, commercial and industrial activities, and agricultural activities), environmental and cultural heritage, and na-tech sites. These tools address specific requirements: (i) being valid/applicable for the whole District, (ii) being based on standardised and institutional data, available at national level, (iii) being calibrated (and possibly validated) in the Italian context. Second, the dedicated open-source Geographic Information System ISYDE (open-source Information SYstem for Damage Estimation) was developed to support technicians in the implementation of the proposed tools and in the visualisation and processing of damage results. Finally, ISYDE was transferred to Regional Authorities for flood damage evaluation and mapping to all areas at significant risk in the Po District.<br>Results of damage assessment are collected in dedicated sheets (one for each area) and are presented by means of tables, graphs, and maps, in order to provide both the overall expected damage in a specific area as well as its spatial distribution, for all the five categories of exposed elements. The new developed maps represent a significant improvement with respect to the previous ones, supporting the Cost Benefit Analysis (CBA) of mitigation measures to be defined in FRMPs. Indeed, they enable the identification and localization of typological and spatial hotspots within a single area as well as the comparison among different scenarios, in terms of CBAs.</p>

<p><span><span>Reliable information on building stock and its vulnerability is important for understanding societal exposure to flooding and other natural hazards. Unfortunately, this often lacks in developing countries, resulting in flood damage assessments that use aggregated information collected on a national- or district level. In many instances, this information does not provide a representation of the built environment, nor its characteristics. </span>This study aims to improve current assessments of flood damage by extracting structural characteristics on an individual building level and estimating flood damage based on its related susceptibility. An Object-Based Image Analysis (OBIA) of high-resolution drone imagery is carried out, after which a machine learning algorithm is used to classify building types and outline building shapes. This is applied to local stage-dependent damage curves. To estimate damage, the flood impact is based on the flood extent of the 2019 mid-January floods in Malawi, derived from satellite remote sensing. Corresponding water depth is extracted from this inundation map and taken as the damaging hydrological parameter in the model. The approach is applied to three villages in a flood-prone area in the Southern Shire basin in Malawi. By comparing the estimated damage from the individual object approach with an aggregated land-use approach, we highlight the potential for very detailed and local damage assessments using drone imagery in low accessible and dynamic environments. The results show that the different approaches on exposed elements make a significant difference in damage estimation and we make recommendations for future assessments in similar areas and scales.</span></p>

Abstract. Natural disasters cause human sufferings and property loss, if not managed properly. It cannot be prevented but their adverse impacts can be reduced through proper planning and disaster mitigation measures. The floods triggered by heavy rains during July 2010 in Pakistan caused swallowing of rivers causing human, agriculture, livestock and property losses in almost all over the country. The heavy rains in upper part of country were attributed to El-Nina effect. Accumulated water in the rivers floodplain overtopped and breached flood protective infrastructure. Flood damage particularly in Sindh province was caused by breaches in the embankments and even after months of flood recession in rivers, flood water affected settled areas in the province. This study evaluates the role of satellite remote sensing particularly in assessment of breaches and consequential damages as well as measures leading to minimize the effects of floods caused by breaches in flood protective infrastructure. More than 50 SPOT-5 imageries had been used for this purpose and breached areas were delineated using pre and post flood imageries, later on rehabilitation work were also monitored. A total 136 breaches were delineated out of which 60 were in the Punjab and 76 in Sindh province. The study demonstrates the potentials of satellite remote sensing for mapping and monitoring natural disasters and devising mitigation strategies.

A Decision-Making Framework Integrating Fluid and Solid Systems to Assess Resilience of Coastal Communities Experiencing Extreme Storm Events

Flood damage assessment (FDA) provides an effective basis for the flood risk management process to evaluate and treat risks. Current FDA methods, however, are ill-equipped to account for the distinctiveness of buildings in the analysis. This is mainly due to their limited input data that on the one hand requires complete building information (well-represented in building information models [BIM]) and, on the other hand, the flood information (commonly managed by geographic information systems [GIS]). While such requirements cannot be satisfied via independent use of BIM or GIS, their integration can potentially be used for this purpose. The current BIM-GIS integration methods, however, are limited or application-specific. Therefore they fall short in satisfying the micro-level FDA information requirements. This paper presents a new data model as a profile of Geography Markup Language (GML) for BIM-GIS integration to support the detailed assessment and 3D visualisation of flood damage to buildings. The model was further evaluated using a case study and found effective for this purpose.

Abstract. Estimating flood damage, although crucial for assessing flood risk and for designing mitigation policies, continues to face numerous challenges, notably the assessment of indirect damage. It is widely accepted that damage other than direct damage can account for a significant proportion of total damage. Yet due to scarcer data sources and lack of knowledge on links within and between economic activities, indirect impacts have received less attention than direct impacts. Furthermore, attempts to grasp indirect damage through economic models have not gone below regional levels. Even though local communities can be devastated by flood events without this being reflected in regional accounts, few studies have been conducted from a microeconomic perspective at local level. What is more, the standard practices applied at this level of analysis tackle entities but ignore how they may be linked. This paper addresses these two challenges by building a novel agent-based model of a local agricultural production chain (a French cooperative wine-making system), utilized as a virtual laboratory for the ex ante estimation of flood impacts. We show how overlooking existing interactions between economic entities in production chains can result in either overestimation (double counting) or underestimation (wrong estimation of the consequences for the activity) of flood damage. Our results also reveal that considering interactions requires thorough characterization of their spatial configuration. Based on both the application of our method and the results obtained, we propose balanced recommendations for flood damage estimation at local level.

Natural hazards, such as flooding, have been negatively impacting developed and emerging economies alike. The effects of floods are more prominent in countries of the Global South, where large parts of the population and infrastructure are insufficiently protected from natural hazards. From this scope, a lot of effort is required to mitigate these impacts by continuously providing new and more reliable tools to aid in mitigation and preparedness, during or after a flood event. Flood mapping followed by damage assessment plays an important role in all these stages. In this work we investigate a new dataset provided by DrivenData Labs based on Sentinel-1 (S1) imagery (VH, VV imagery and labels) to help map floods in the city of Beira in Mozambique. Exploiting Google Earth Engine (GEE), we deployed supervised and unsupervised machine learning (ML) methods on a dataset comprising imagery from 13 countries worldwide. We first mapped the floods country-by-country including Mozambique. This first part was helpful to understand the sensitivity of each method when applied to data from different regions and with different polarizations. We then trained the supervised model globally (in all 13 countries) and used it to predict floods in Beira. To assess the accuracy of the experiments we used the intersection over the union (IoU) metric, results of which we compared with the benchmark IoU achieved by the winner in the DrivenData competition for flood mapping in 2021. The implementation of unsupervised and supervised ML using VH and VV+VH produced satisfactory results, and showed to be better than using VV imagery; in Cambodia and Bolivia with VH polarization yielded IoUs values ranging from 0.819 to 0.856 which is above the benchmark (0.8094). The predictions in Beira using VH imagery yielded IoU of 0.568, which is a reasonable outcome. The proposed approach is a reliable alternative for flood mapping, especially in Mozambique due to its low cost and time effectiveness as even with unsupervised approaches, relatively high-quality results are yielded in near real-time. Finally, we used Sentinel-2 (S2) imagery for a land cover classification to perform damage assessment in Beira and integrated population data from Beira to enhance the quality the results. The results show that 20% of agricultural area and about 10% of built up area were flooded. Flooded built up area includes highly populated neighborhoods such as Chaimite and Ponta Gea that are located in the center of the city.

India is one of the world's regions that is most vulnerable to natural disasters, which results in yearly in Hyderabad also losses of property, infrastructure, and human lives. Monitoring and evaluating the floods and other disasters can be done from a unique vantage point by using air and space-based observations of the globe. Geospatial technology has significantly enhanced flood research, particularly in the first three phases of floods. Specifically, there are three phases: a) preparation for floods; b) monitoring for floods; and c) recovery from floods (damage assessment and mitigation phase). Agriculture, socioeconomic, communication, population, and infrastructure data are all included in the GIS data set that was developed during the preparation phase. In the event of a serious flood crisis in Hyderabad, this can be utilised in conjunction with the flooding data to adopt an evacuation strategy, rehabilitation planning, and damage assessment.