Combining low-cost drone surveys and citizen science to identify flood-prone areas

Introduction: Natural disasters are becoming increasingly frequent, highlighting the need for accessible and reliable tools to support flood risk assessment, particularly in data-scarce regions.

Methods: This study evaluated the accuracy of Digital Elevation Models (DEMs) generated from UAV-based photogrammetry at two sites in the Upper Paranapanema River Basin (São Paulo State, Brazil). Surveys were conducted using Phantom 4 Pro and Mavic 3E RTK drones and validated against GNSS RTK geodetic measurements. UAV-derived DEMs were compared with freely available satellite datasets (SRTM and ANADEM) using RMSE, MAE, and PBIAS metrics. To assess practical implications, historical flood extents were reconstructed by combining DEMs with flood watermarks observed on utility poles, while resident interviews were used as an independent validation dataset.

Results: UAV-derived DEMs exhibited higher accuracy than satellite-based DEMs and showed improved agreement with citizen-reported flood limits. High-resolution UAV data better captured flood-relevant microtopography, particularly in urban areas, leading to more realistic flood inundation reconstructions.

Discussion: The proposed framework advances beyond conventional DEM-topographic survey comparisons by integrating low-cost UAV data with local flood observations, enabling operationally robust flood extent mapping. The results demonstrate that UAV-based DEMs represent an effective and affordable alternative to costly LiDAR or extensive GNSS surveys, reinforcing their potential for flood risk mapping in isolated and data-limited regions.

1 Introduction

Floods are a natural phenomenon that periodically affects rural and urban areas worldwide. Depending on their magnitude and duration, floods can be considered a disaster responsible for massive destruction in infrastructure and the environment, significant economic losses, social disturbances, and, in many cases, the loss of lives (Amirkhani et al., 2022; EM-DAT, 2015).

In recent years, the incidence of floods has increased in magnitude and frequency due to the combination of human-caused climate change and inadequate land management (Das and Gupta, 2021; Hagos et al., 2022). Because of the rapid urbanization, urban flooding has received more attention from the scientific community. The increase in frequency and magnitude of urban flooding has been linked to the increased impervious areas, changes in natural river channel morphology due to anthropogenic interventions, excessive urban sprawl, irrational urban planning, and inadequate flood management (Abebe et al., 2018; Miller and Hutchins, 2017; Choubin et al., 2019).

Sustainable management of flood risk is dependent upon the development of knowledge of the risk and probability of flood events (Binns, 2022). In recent years, potential flood risk area mapping has been considered as the major strategic component to effectively manage, reduce, and mitigate the potential impacts of flood hazards since these products provide residents and stakeholders with information on potential flood-prone areas (Abdelkarim et al., 2020; Rahmati et al., 2015; Mohanty and Simonovic, 2022).

Rapid urbanization has also been a major feature in Brazil. According to the 2022 Demographic Census (IBGE, 2022), of the total of more than 200 million people living in Brazil, 87.4% live in urban areas, while 12.6% reside in rural areas. Compared to 2010, when the degree of urbanization was 84.4%, there was an increase of 16.6 million people living in urban areas and a decrease of 4.3 million living in rural areas. The MapBiomas Project (MapBiomas, 2022) estimated that urbanized areas totalized 37 thousand km² (0.4 % of the whole area of Brazil), and approximately 4.2 thousand km² of urbanized areas were located within 3 m height above the nearest drainage. Moreover, 68% of the occupation of areas within this threshold occurred in the last 38 years, pointing to an intensification of the occupation of prone-flooded areas.

The increased frequency of flood events has led the Brazilian government to approve new legislation in 2012 (law 12,608) (Brasil, 2012) which established standards and guidelines to reduce the risks of disasters. Under this law, municipalities must identify and map areas susceptible to landslides, flash floods, and gradual floods, among other geological and hydrological relevant events. Despite the well-intention expressed in the law, the practical application of its principles after more than 10 years has been quite limited, due to the lack of high-resolution digital terrain models. The available DEM covering the whole country is often based on satellite-based low-resolution surface models with errors in altitude of the order of 2–5 m, which is clearly insufficient for the identification of prone-flooded areas and for flood risk mapping.

Because surveying flood-prone areas requires expensive topographic surveys, flood vulnerability studies, or even mapping is virtually non-existent in most of the 5,000 country's municipalities. Remote sensing and traditional photogrammetry are generally unable to achieve the level of detail required for applications in flood scenarios. On the other hand, conventional topographic surveying, using total stations or theodolites, or the use of GNSS receivers, usually provides greater accuracy, but in hard-to-reach isolated places, it can significantly increase costs or even be physically unfeasible.

Although urbanized areas tripled in magnitude over the period 1985–2022, only 97 of the 5,570 municipalities of Brazil have an urbanized area greater than 50 km², which indicates that despite the increase in urbanization of the country, a large proportion of the population still lives in smaller urban areas (less than 100 thousand inhabitants) which presented the higher rate of area expansion (MapBiomas, 2022). In addition, most of these municipalities are located in remote areas, with limited access to financial, technological, and human resources, where cost and availability of detailed topographic surveys became the main constraint for identifying flood-prone areas.

In this context, the use of unmanned aerial vehicles (UAV) platforms for environmental monitoring in recent years make possible the acquisition of spatial information on the river environment with high precision and high sampling frequency (Anderson and Gaston, 2013). The development of UAV technologies creates opportunities to generate sufficiently faster and more accurate data at much lower costs for the assessment of flood-prone areas (Feng et al., 2015). Moreover, UAVs can be used for acquiring high resolution data under complex urban landscapes as well as inaccessible areas due to hazardous environments (Boccardo et al., 2015; Feng et al., 2015).

Another way of producing maps of flooded areas is through the use of citizen science. This approach involves different types of data acquisition (Assumpção et al., 2018), which has been successfully used not only in data-scarce areas of least developed countries (Fohringer et al., 2015; Walker et al., 2016; Garrote, 2022) as well as in developed regions (Smith et al., 2015). As demonstrated by Sy et al. (2020), the use of citizen science could help reconstruct past flood events and help obtain useful information in areas of scarce or unavailable flow data records.

Several studies have already explored UAV applications for flood assessment, including delineation of floodwater extent (Mazzoleni et al., 2020), rapid post-event mapping (Boccardo et al., 2015), and hydraulic model support (Feng et al., 2015). However, fewer studies have evaluated how DEM accuracy, affected by flight altitude, GCP configuration, and platform type, propagates into errors in flood extent reconstruction. Recent research has emphasized that the suitability of Digital Elevation Models for hydrological applications must be carefully assessed, as DEM resolution and processing choices can significantly influence model outcomes (Vujović et al., 2024). In addition, Valjarević (2024) emphasized that the suitability of DEMs for hydrological applications should be evaluated not only in terms of geometric accuracy, but also in their ability to reproduce realistic drainage networks and basin characteristics. Moreover, almost no studies combine UAV-derived terrain models with citizen science flood recollections as an independent validation source. This gap motivated the integrated approach developed in the present study.

Therefore, the objective of this study is to evaluate the accuracy of DEMs obtained under different UAV acquisition configurations and compare them with free satellite-based DEMs using a geodetic reference; and to examine how DEM accuracy influences topography-based reconstruction of historical flood extents when combined with citizen-contributed recollections.

It is important to mention that in this study UAVs were not used to map water surfaces directly. Instead, UAV-derived DEMs were evaluated as the static topographic input controlling the delineation of historical flood extents, based on the premise that urban areas located in isolated areas are generally not accessible during extreme flood events, and carrying out surveys can be dangerous and unfeasible due to flood damage to infrastructure. Therefore, maximum water levels recorded on utility poles after major floods provided the reference elevation of past events. Citizen science interviews were then used as an independent validation dataset, allowing us to compare the agreement between community recollections and DEM-based flood reconstructions.

By linking UAV-derived terrain data, open-access DEMs, and participatory information, this study advances the understanding of how topographic accuracy drives flood delineation and offers a practical methodology for data-scarce regions seeking to improve flood risk assessment.

In this study, we present an innovative flood inundation mapping framework that moves beyond conventional DEM–topographic survey comparisons by enabling realistic and operationally robust flood extent mapping using low-cost optical UAV data, validated against citizen-science flood observations. The methodological approach offers an effective alternative to costly LiDAR/GNSS surveys and outperforms geodetic surveys in urban areas by imaging and documenting information of hard-to-access flood-relevant constructed areas.

2 Materials and methods

2.1 Study area

The drone surveys of this study were conducted in the Alto Paranapanema basin, in a region with a history of flooding, close to the confluence of the Itapetininga and Paranapanema rivers. The study area is dominated by a large floodplain, which extends to the south bank of the Itapetininga and at both margins of the Paranapanema, where the river morphology is characterized by meanders with several abandoned loops. Within the study site, two survey areas of the floodplain were delimited: the first site has a total area of 17.79 hectares and is located in the rural area dominated by pasture (top panel Figure 1), hereafter referred as rural site; the second is an area of 54.75 ha of the urban district where the landscape is a mixture of housing and riparian vegetation (bottom panel Figure 1), referred in the text as urban site.

Figure 1

Figure 1. Location of the survey areas within the Upper Paranapanema Basin. The top panel corresponds to the rural site, while the bottom indicates the urban site. The pink triangles indicate the locations of the interviews with local residents regarding past flood events, while the red triangles indicate the locations of utility poles where water marks of past floods were recorded (shown in detail at the left). Basemap: Google Satellite. Imagery © Google.

The landscape of the rural site is dominated by pastures for grazing beef cattle. The area presents gentle relief, dominated by pasture with low surface roughness, with few buildings and trees, favoring the drone survey and aerial photogrammetric reconstruction. These features facilitate the geodetic and aerial survey and the construction of the digital model.

In contrast, the urban sector is much more challenging for both surveys. In this site, geodetic measurements were only possible in the fronts of properties, streets, and the back of some properties located on riverbanks. In addition, the presence of buildings, walls, riparian vegetation, power lines, and other man-made objects obstructs sightlines and jams GNSS signals, posing additional constraints to surveying. Because of this, levels within the properties were obtained by interpolation between the front and back.

2.2 Citizen science data

In this study, historical flood data consist of post-event, community-generated evidence of past floods, including flood marks on utility poles and resident-reported maximum water levels, which were used to reconstruct flood extents in the absence of instrumental hydrometric records.

In order to evaluate drone survey products regarding flooded area extension, a citizen science–based approach was employed to reconstruct historical flood events in the urban site. Semi-structured interviews, indicated by pink triangles of Figure 1, were conducted. These interviews were guided by a script, which sought citizens' memories regarding floods that had hit the property or that had occurred in the region. The questions included: name; address; contact; relationship with the house; length of time living in the region; length of time living in the property; date and height of the flood water that affected housing; date and height of the flood water that had occurred in the region; and other reports or experiences. A total of 28 interviews were conducted to recover the flooded area of the historical floods of 1983, 1990, 1991, 1997, 1998, 2004, 2010, and 2016. The study was approved by the Research Ethics Committee (CEP) in accordance with Brazilian National Health Council Resolution No. 510/2016 (CAAE 61253522.3.0000.5504).

A more detailed discussion of the citizen science data collection used in this study and its inherent uncertainties is provided in Vasconcelos et al. (2025), where the authors explicitly acknowledge the qualitative and memory-dependent nature of interview-based flood reconstruction. The study emphasizes that such data are intended to validate recurring spatial patterns of inundation rather than precise water levels, and that recall uncertainty is mitigated through spatial redundancy of interviews and consistency with physical evidence and topographic controls.

In addition, this information was augmented by flood marks on utility poles (indicated by the red triangles and shown at the left of Figure 1) painted by local residents, which include dates and maximum water levels during past major floods beginning in 1983. These watermarks were geolocated using the RTK-GPS system to determine the altitude of each flood mark, and were used as a reference to determine the flooded area of past events for the derived DEMs.

Although citizen science typically involves public engagement throughout the scientific process, the concept is recognized as broad and flexible (Bonney et al., 2016; Njue et al., 2019). Participation may range from fully collaborative frameworks, where citizens contribute to study design, data collection, and interpretation, to simpler, yet still meaningful, forms of involvement, such as providing local observations or interpreting environmental information (Haklay, 2013; Nardi et al., 2022). In this study, we adopted this broader view and used the term citizen science to refer specifically to the contribution of local knowledge and community-generated flood evidence. These qualitative data, provided directly by residents, function as an independent validation layer for the flood extents reconstructed from DEMs. We therefore adopt a narrow definition of citizen science focused on participatory data contribution, consistent with approaches used in flood memory studies (Sy et al., 2020; Walker et al., 2016).

2.3 Geodesic and drone surveys

Aerial photogrammetric used imagery from unmanned aerial vehicles (UAVs) of two commercial low-cost equipment: A DJI manufactured drone (model Phantom 4 Pro), equipped with a 4K resolution camera and a 20-megapixel sensor (DJI, 2021); and second drone of the same manufacturer (model Mavic 3E RTK), also equipped with a 20-megapixel sensor and an integrated real-time kinematic positional correction (RTK) system.

The Phantom 4 Pro and the Mavic 3E RTK were selected because they represent two widely adopted commercial UAV platforms in Brazil, encompassing both conventional photogrammetric and RTK-enabled survey workflows commonly used in operational terrain mapping. In the literature, DJI UAV platforms such as the Phantom and Mavic series are frequently used for photogrammetric mapping applications, demonstrating their broad adoption in scientific and applied UAV-based terrain surveys (Kovanič et al., 2023).

For the rural site, three flight configurations were executed with the Phantom 4 Pro to evaluate the influence of flight altitude and the use of ground control points (GCPs) on DEM accuracy. The configurations were: P4PRO250—flight at 250 m altitude, without GCPs; P4PRO120wGCP—flight at 120 m altitude, with GCPs; P4PRO120woGCP – flight at 120 m altitude, without GCPs. In addition, the Mavic 3E RTK also performed a full photogrammetric survey at the rural site, and a detailed geodetic survey was conducted in the area.

These combinations enabled assessing how flight height affects image geometry and how the presence or absence of GCPs impacts the accuracy of the resulting DEMs.

For the urban site, photogrammetric data were obtained exclusively using the Mavic 3E RTK drone, complemented by a detailed geodetic survey.

For the geodetic survey in both areas, a pair of Geomax GNSS RTK receivers, model Zenith35PRO, was used. One of the receivers acted as a fixed base station, while the other operated as a mobile (rover), responsible for collecting the elevation points and ground control points (GCPs). For the rural site, which has a total area of 17.79 hectares, a total of 32 GCPs (indicated as white crosses painted on the ground) and 291 elevation points were implemented. GCPs were deployed in such a way as to homogeneously cover most of the area, in order to create a mesh allowing correction of the flight. With regard to the urban site, 644 elevations points were surveyed in an area of 54.75 ha.

The use of ground control points (GCPs) is a critical factor in increasing the accuracy of three-dimensional information obtained by aerial surveying (Oniga et al., 2018). Although three GCPs are, in principle, sufficient for basic georeferencing of images, increasing the number of points directly contributes to improving the final accuracy of the products generated, such as the point cloud, three-dimensional mesh, orthomosaic and digital surface model.

Oniga et al. (2018) conducted a controlled test in an area of approximately 1 hectare, where 50 control points were established, resulting in a density of 50 GCPs per ha. However, previous studies suggested uniformly distributed GCPs with a density ranging from 1 point per 3 to 5 ha, which ensures planimetric accuracy and meets engineering and mapping requirements (James and Robson, 2012; Westoby et al., 2012; Colomina and Molina, 2014).

2.4 Remote sensing products

We compared the DEM from the drone surveys to two products which are free available: The first is the version 3 of the Shuttle Radar Topography Mission (SRTM), which is a 30 m resolution digital topographic database (Farr et al., 2007); and the second is the ANADEM, which is a digital terrain model based on the Copernicus GLO-30 digital elevation model but with vegetation bias removal. With a spatial resolution of 30 meters and available for all of South America, the product is made freely available by the Brazilian National Water and Sanitation Agency (Laipelt et al., 2024).

2.5 Image and data processing

To ensure comparability, spatial consistency, and reproducibility of results, Remote Sensing for data acquisition and Geographic Information System (GIS) for analysis and visualization were fully integrated in the same software environment. In this study, the term DEM is used as a generic designation for gridded elevation datasets; however, all UAV-derived DEMs correspond to bare-earth Digital Terrain Models (DTMs), as vegetation and built structures were removed during the photogrammetric processing.

For the Mavic 3E RTK, raw data were corrected in Emlid Studio 1.9 (EMLID, 2024) using RTK with post-processing and integrated GNSS database files. After coordinate correction, the Mavic 3E RTK data followed the same initial processing workflow applied to the Phantom 4 Pro, based on the principles established by Buffon et al. (2018). The images were processed using Agisoft PhotoScan (AGISOFT L.L.C., 2014). The workflow began with image alignment, during which overlapping features were identified and matched between consecutive photographs, resulting in the generation of an initial sparse point cloud. This point cloud was subsequently densified to produce a dense point cloud.

Based on the dense point cloud, the DEMs from drone surveys were generated by classifying the dense point cloud in Agisoft Metashape with a spatial resolution of 1 m. The automatic classification was then visually inspected to detect inconsistencies, and manual editing was performed to refine the separation between soil, vegetation, and built structures. Only points classified as “ground” were retained for model generation, ensuring the effective removal of trees, buildings, and other aboveground objects from the elevation surface. Using this filtered dataset, a set of DEMs representing the bare-earth surface was generated.

In the geodetic survey, GNSS observations were post-processed using the PPP-IBGE service, which applies Precise Point Positioning (PPP) corrections based on precise satellite orbits and clock products. Elevation points were corrected by precise point positioning and imported into AutoCAD Civil 3D software, and a DEM with spatial resolution of 1 m was created using the TIN (Triangulated Irregular Network) method.

2.6 Deriving historical flood extent

Before reconstructing historical flood extents in the urban area, we first evaluated the relative performance of the available DEM products in a controlled environment, as described in Section 2.3. The rural site was selected for this initial assessment because its flat and geomorphologically simple terrain reduces topographic complexity and minimizes potential sources of error. This setting allowed a robust comparison among the DEM products considered in this study.

From this evaluation, we identified two products that showed the strongest concordance with the geodetic survey: (a) the free DEM with the highest overall agreement, and (b) the drone-derived DEM with the lowest elevation error. These two DEM configurations were then also produced for the urban site, where a geodetic survey was likewise conducted (as described in Section 2.3). This allowed their performance to be assessed in a more complex terrain with recurrent flooding. In doing so, we established an evaluation framework that moves beyond a conventional DEM-topographic survey comparison by testing the suitability of these products in a real flood-mapping application.

Using each selected DEM as a static topographic input, we reconstructed the historical flood extents for major events. Water marks painted on utility poles provided the maximum water surface elevation (in meters). Assuming a horizontal water surface at this elevation, an approach commonly adopted in studies with limited hydrodynamic data, we delineated flooded areas by identifying all DEM cells with elevation lower than or equal to the observed water level. This produced topography-based flood extent maps using consistent hydraulic assumptions across DEM sources.

Since the extensions of the flooded area of each event were determined assuming a horizontal water elevation projected from the maximum water level recorded in the utility poles, this approach neglects the effects of hydraulic gradients and the blockage effects of vegetation and urban construction on the water surface. However, this simplification is justified by the fact that the utility poles are located close to the flooded area, minimizing the effects of the surface gradients and providing an equal base of comparisons among DEM products with different native resolutions. Although the local effect of urban obstruction can be assessed through hydrodynamic modeling, the purpose of this study was to test simple approaches for data-scarce conditions.

Finally, citizen-science interviews served as an independent validation dataset. For each interview location, we classified the reported point as “flooded” or “not flooded” based on the flood maps generated from each DEM. This enabled a categorical accuracy assessment comparing modeled flood occurrence with residents' accounts.

2.7 Performance evaluation

To assess the accuracy of the drone-based DEMs, a series of surveys was conducted at the rural site using the Phantom 4 PRO and Mavic 3E drones at different flight altitudes, since altitude is a determining factor in imagery post-processing time. For the Phantom 4 PRO, post-processing was performed both with and without ground control points, enabling evaluation of the influence of field control on the results. The complete set of DEMs tested is presented in Table 1.

Table 1

Table 1. Details about the drone surveys and the DEMs evaluated.

The comparison of the elevation models was performed using, as reference DEM, the high-resolution geodetic surface (1 m spatial resolution). All DEMs, including those derived from satellite data (SRTM and ANADEM), drone-based surveys with the Phantom 4 PRO (at both flight altitudes and with and without ground control points), and the RTK-based Mavic 3E, were transformed to the UTM (Universal Transverse Mercator) coordinate system (zone 22S, central meridian −51°W), and resampled and spatially aligned to match the extent and resolution of the reference DEM.

The analysis was conducted in the R programming environment, where all rasters were adjusted to a common 1-meter spatial resolution and compared on a pixel-by-pixel basis. For each cell, the elevation value from the tested DEMs was compared to the corresponding value from the reference surface, enabling the generation of error maps. Accuracy was evaluated through statistical metrics, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Percent Bias (PBIAS), calculated using the hydroGOF package. The workflow was fully scripted and reproducible, using the terra and raster packages for spatial data manipulation, hydroGOF for accuracy metrics, ggplot2 for data visualization, and reshape2 for restructuring the data for graphical analysis. The full R script used for processing, spatial alignment, and accuracy assessment is openly available at Benso (2025), allowing reproducibility and further exploration of the methodology. Performance statistics were calculated as follows:

$\begin{array}{ll}MAE=\frac{1}{n}{\displaystyle \sum _{i=1}^n}|y_i-\widehat{y_i}|& (1)\end{array}$

$\begin{array}{ll}RMSE=\sqrt{\frac{1}{n}{\displaystyle \sum _{i=1}^{\text{n}}}{\left(y_i-\widehat{y_i}\right)}^2}& (2)\end{array}$

$\begin{array}{ll}PBIAS=100\times \frac{{\displaystyle {\sum }_{i=1}^n}(y_i-\widehat{y_i})}{{\displaystyle {\sum }_{i=1}^n}{y}_i}& (3)\end{array}$

Where:

y_i = observed value (reference elevation)

$\widehat{y_i}$ = predicted value (elevation from the DEM)

n = total number of observations (pixels)

Finally, an additional evaluation was carried out in the urban site using citizens' recollection of past events gathered in the questionnaires described in item 2.2. Based on the water marks registered in the utility poles of Figure 1, the extension of the flooded area for major past flood events was determined for each DEM. Then, for the location of each interview, it was verified if the flooded area matched the information reported by the local population. Statistics in this case were categorically, indicating the overall accuracy and the omission percentages.

3 Results

The comparison of the altimetric errors of the different Digital Elevation Models (DEMs) for the rural site, represented in Figure 2, shows significant variations between the analyzed products. The errors were expressed in meters and the color scale varies from −5 to +15 m, where darker tones (green) indicate smaller errors, and lighter tones (yellow to pink) represent greater discrepancies in relation to the reference model.

Figure 2

Figure 2. Spatial distribution of altimetric errors of the evaluated DEMs for the rural site.

The spatial distribution of altimetric errors revealed significant differences among the models evaluated. The SRTM model presented the largest altimetric errors, with values exceeding 15 meters in several regions of the study area.

The ANADEM model, despite having the same spatial resolution, showed better performance to the SRTM, with errors concentrated between 0 and 5 meters and a more homogeneous spatial distribution, although still limited in the representation of small-scale surface variations.

Except for the Phantom 4PRO without ground control points, the DEMs derived from drone surveys outperformed satellite products. Visual comparisons of Figure 2 between the DEMs produced by the Phantom 4 PRO flying at different altitudes (P4PRO250 against P4PRO120woGCP) indicates that the accuracy of the DEM is quite sensitive to the flight height. The DEM generated with images at 120 meters altitude without GCPs showed greater spatial variation of errors, with distortions at the edges. The flight at 250 meters, although it presented reasonable performance, resulted in more significant errors compared to the flight at 120 meters.

The use of ground control points produces major improvements in the accuracy of the DEM, as revealed by the low errors of the P4PRO120wGCP in comparison with the P4PRO120woGCP, highlighting the relevance of calibration by control points.

The DEM generated with the MAVIC 3E drone presented consistent performance, with low errors and good representation of topographic variations, being one of the positive highlights. The MAVIC3E DEM showed the lowest errors, with values close to zero and uniform distribution, standing out as the most accurate among those tested.

These qualitative patterns are consistent with the quantitative error metrics summarized in Table 2, which presents the RMSE, MAE, and PBIAS values for all DEMs. The numerical results reinforce the superior performance of drone-derived DEMs, especially the MAVIC 3E and the Phantom 4 PRO with GCPs, compared to satellite products.

Table 2

Table 2. Performance statistics for each of the DEMs evaluated compared to the geodetic reference DEM in the rural site.

Additionally, Figure 3 presents the statistical distribution of altimetric errors using boxplots for the different DEMs evaluated. The graphical analysis reinforces the conclusions obtained in the spatial evaluation, highlighting the inferior performance of the SRTM model, which presents the greatest dispersion of errors, with maximum values exceeding 15 meters and a positive median, indicating a tendency to overestimate altitude. The ANADEM DEM presents much less dispersion of errors, with a median close to zero and few outliers, which corroborates its superior performance.

Figure 3

Figure 3. Boxplots of the altimetric errors of the different elevation models.

The Phantom 4 Pro models also showed consistent behavior in terms of the flight height and use of GCPs. Flight at 120 meters with GCPs presents the lowest variability and practically zero median, indicating excellent accuracy. Without GCPs, there is an increase in error variability and occurrence of outliers, highlighting the importance of ground points in post-processing of the survey. The model obtained at 250 meters, although presenting a median close to zero, has greater dispersion and presence of negative outliers, indicating a loss of altimetric precision associated with the higher flight height.

The drone-derived models (MAVIC3E and Phantom 4 Pro) demonstrate superiority in relation to orbital data. The MAVIC3E presents good accuracy with a symmetrical distribution concentrated around the zero error, reinforcing its good performance already observed in the spatial analysis. As also confirmed by the performance statistics in Table 2, the DEM generated with the MAVIC 3E showed the lowest RMSE and MAE values, with errors close to zero and uniform distribution, standing out as the most accurate among those tested.

Figure 4 shows the spatial distribution of errors for the urban site for both the ANADEM and the MAVIC3E, which were the surveys that presented the best performance in the rural site. Again, comparisons of both DEMs against the geodetic reference indicate a better performance of the drone survey (Table 3). However, errors are comparatively higher compared to the rural site, indicating the difficulties in dealing with the interferences created by vegetation and housing. In general, the MAVIC3E DEMs show lower altitude than the geodetic reference, while the ANADEM showed an overestimation. Discrepancies for both DEMs are more evident in the south of the urban site, which is the area more populated.

Figure 4

Figure 4. Spatial distribution of altimetric errors of the ANADEM and MAVIC3E evaluated in the urban site compared to the geodetic reference DEM.

Table 3

Table 3. Performance statistics for each of the ANADEM and MAVIC3E evaluated in the urban site compared to the geodetic reference DEM.

Finally, Table 4 presents a validation for the urban site based on citizen science data. The interviews with the local population were used to reconstruct the area flooded for the events from 1983 through 2016, totaling 28 flooded points. As mentioned before, using the water marks of the utility poles of Figure 1 for each historical flood, the extension of the flooded area was determined, and the accuracy of each DEM was checked against the geolocated interviews with the local population. In this specific case the percentage of omission errors were, respectively, 14 %, 4 %, and 57 % for the Geodetic Survey, MAVIC 3E and ANADEM, while in terms of agreement, the values were, respectively, 86%, 96 %, and 43 %. Contrary to the expectations, the MAVIC3E DEM showed better performance compared to the geodetic survey.

Table 4

Table 4. Comparison between the flooded area determined for the DEMs from the geodetic survey, MAVIC3E and ANADEM in terms of number of omissions and agreement of historical events registered by the local population in the urban site.

Figure 5 illustrates the result of this analysis, showing the location of the interviews and the flooded areas projected from the water mark of the 2004 flood, which was the largest recorded in the area. It is clear that both the Geodetic survey and the MAVIC3E DEMs produce continuous flood areas, while the flooded area from the ANADEM shows discontinuities (isolated flooded—non flooded pixels). The surface generated by the MAVIC3E captured the events reported in the interviews in the area of higher density of housing at the center of the urban site, while the Geodetic survey missed these occurrences.

Figure 5

Figure 5. Flooded area projected using different DEMs from the water mark of the largest event recorded in the urban site. The dots indicate the location of the interviews from past events, highlighting in green color the agreements between the flooded area and the interviews, while the omissions are shown in red. Basemap: Google Satellite. Imagery © Google.

4 Discussion

The assessment in the rural site indicated that the SRTM-derived DEM showed lower performance compared to the other products tested. This performance is attributed to its low spatial resolution (~30 meters) and the radar acquisition methodology, which tends to capture the top of the vegetation instead of the actual surface of the terrain, compromising altimetric accuracy in environments with vegetation cover.

The ANADEM DEM showed better performance than SRTM, and even in the case of drone-derived DEMs, which did not employ ground correction points. In agreement with the results of this study, Garrote (2022) has shown that results related to the Copernicus DEM (which is the basis for the ANADEM product), provide better results compared to other free global DEMs.

In this context, it is worth noting that the SRTM model has limitations for applications that demand high precision and is more suitable for analyses on a regional scale. The ANADEM model, although it does not reach the performance of products derived from platforms with differential correction, proved to be a viable alternative for projects that require intermediate precision, superior to that offered by SRTM.

Comparing the DEMs generated from the flights performed with the Phantom 4 Pro, without the use of differential correction (RTK) or ground control points (GCPs), a significant difference in altimetric accuracy was observed as a function of flight height. These results show that increasing altitude compromises point cloud density, reduces spatial resolution, and increases altimetric errors, negatively impacting model accuracy.

Although both flights were carried out without precise georeferencing support, altitude proved to be a determining factor in the degradation of the positional quality of the generated models.

In the specific case of the model generated with the Phantom 4 Pro at 250 meters, the lack of precise georeferencing, combined with the higher flight altitude, compromised the positional quality of the product, making it unsuitable for technical applications that require greater accuracy. On the other hand, the models created using differential correction—Phantom 4 Pro with GCPs and MAVIC3E RTK—showed significantly superior performance when compared to a reference topographic model.

Data obtained by Martins et al. (2020), in a survey at 120 meters with GCPs, indicated errors of 0.97 m for longitude, 0.67 m for latitude, and 0.96 m for altitude. In contrast, in the present study, a flight at the same height, but without GCPs, presented errors of 5.61 m for latitude, 1.68 m for longitude, and 1.65 m for altitude. The Mavic 3E RTK demonstrated excellent performance, with errors of only 0.08 m for latitude, 0.08 m for longitude, and 0.27 m for altitude. The Phantom 4 Pro with GCPs also obtained expressive results, with errors of 0.08 m (latitude), 0.14 m (longitude), and 0.29 m (altitude), reinforcing the importance of using positional correction techniques to ensure the quality of cartographic products generated by UAVs.

Comparisons of the historical events based on citizen science information revealed discontinuities (i.e. isolated non-floodable pixels or small pixel clusters in floodable areas and vice versa) in the flooded area identified with the ANADEM. These discontinuities result from the interference of urban buildings and riparian vegetation. Although the ANADEM map is corrected for vegetation, it is clear that it has limitations in complex areas due to its coarser spatial resolution and the combination of vegetation and building which makes particularly challenging high altitude corrections. These isolated pixels are responsible for the lower metrics obtained by the ANADEM in the validation of historical events based on citizen science data. In contrast, discontinuities are not present in the flooded area resulting from both the geodetic survey and the MAVIC3E products.

The better performance of the MAVIC3E is concentrated in the area with a higher density of buildings. The reason for this is related to the fact that the geodetic survey measurements are limited to the main street that provides access to this neighbor, and the Geodetic survey used for comparisons was derived from the interpolation of the altitudes between the access roads and the rear property lines. In the case of the MAVIC3E, however, the images captured information of the backyards of the housing, allowing for a more continuous representation of the flooded areas. Figure 6 shows an aerial image of the urban areas, indicating in red crosses the altitude measured by the geodetic survey. Altitude information revealed that the access road is elevated compared to the terrain at the housing, and that there is a sudden drop in altitude at the front of the housing. Because the interpolation of the geodetic survey includes only information of the road and at the back of the housing, these features cannot be captured with enough accuracy.

Figure 6

Figure 6. Aerial image of the urban site showing details of the housing. The red crosses indicate the altitude of the points of the geodetic survey.

The promising results obtained from the MAVIC3E DEM highlight the potential for rapid surveys in risk areas, especially when there are limitations for implementing ground control points due to vegetation or difficult access. In practical terms, the MAVIC3E can represent the best solution between quality and operational cost, especially in surveys in regions that are difficult to access. More importantly, the use of UAVs has important economic implications: Mazzoleni et al. (2020) compared data obtained by drone, LiDAR and SRTM, highlighting the respective spatial resolutions: 6.0 cm for drone, 1.0 m for LiDAR, and 30.0 m for SRTM. In the study conducted in the Limpopo River plain, in Botswana, a cost of approximately 1,000 dollars was estimated for drone mapping, while complete coverage of the area with LiDAR would reach a cost close to 1,000,000 dollars, evidencing the cost-benefit of aerial photogrammetry with UAVs.

This study has identified flooded area based entirely on the altitude the water marks of past events, and projected the inundated area assuming a horizontal water surface, without considering hydrodynamic effects on the flood waves, or local effects related to vegetation and buildings obstructions that can alter the extension of the flooded area. Therefore, future studies should consider the use of hydraulic models and geomorphic descriptors that can provide a more complete description of the topography and hydrology of the study area (Magnini et al., 2023).

Because the two pilot sites are located in low-relief alluvial plains, they have subtle vertical variations that exert strong control over the extent and storage of floods. In these environments, sub-meter vertical accuracy becomes crucial, which explains the superior performance of elevation models derived from UAVs with RTK or ground control point support for delineating flood-prone areas. Although geodetic surveying provides highly accurate point measurements, its sampling nature limits the representation of microtopography in hard-to-reach urban areas, such as those in backyards, where access is not possible. In contrast, UAV-derived products offer continuous spatial coverage, justifying the greater agreement observed between the DEM of the Mavic 3E and the flood extents reported by residents in the urban area.

Since the reconstruction of flooded areas was based on the assumption of a spatially uniform horizontal water surface, which implies simplifications by disregarding hydraulic gradients, runoff resistance, and urban obstructions, the flood extents obtained should be interpreted as first-order approximations, suitable for comparison between elevation models, and not as complete hydrodynamic simulations.

For simplicity, vertical errors of the DEMs were not propagated in the estimation of the flooding areas. Although this might have altered the performance statistics, considering the same processing was applied to all products we are confident the main conclusions of our study will not change. It is recommended, however, that future study should take into account this limitation.

In addition, statistics derived from citizen science data are qualitative and entirely depend on the local people's memory, which is sometimes flawed, especially for past events that took place decades ago. Nevertheless, the consistency between independent reports, flood marks, and high-resolution topographic surfaces indicates that this information can provide relevant contextual validation in data-scarce regions (See et al., 2016). Thus, the proposed methodology proves particularly suitable for low-slope floodplains, where topography plays a dominant role in flood propagation, and should be applied with caution in steeper or hydraulically complex environments, where hydrodynamic modeling becomes necessary (Neal et al., 2012).

Although the results were consistent across both study sites, it is important to recognize that the performance of DEM products may vary under different geomorphological contexts. The Upper Paranapanema sites present relatively gentle slopes and low vertical relief, conditions that tend to favor accurate surface reconstruction from UAV imagery. Future studies should therefore replicate these analyses in areas with contrasting morphologies, such as steep hillslopes, deeply incised valleys, dense riparian forests, or highly urbanized terrains, to evaluate the robustness and generalizability of the methodology under more heterogeneous topographic and land cover conditions.

5 Conclusion

This study showed several advantages of using commercial UAVs for mapping flood-prone areas, not only in terms of the cost-benefits when compared to the traditional geodetic survey, but also because of being less time-consuming and laborious. However, surveys showed that the accuracy strongly depends on the flight height, which can be an issue for surveying larger areas because of the large number of images to be processed.

Additionally, cross-validation with citizen science information showed a better performance of the UAV's survey because it provided information on inaccessible areas, such as the backyards of housing, which were not contemplated in the geodesic survey.

However, the study also showed the limitations of drone surveys without RTK system, which requires a significant density of ground control points to provide accurate estimations. In this case, altitude information of areas with difficult access due to vegetation or housing, which is the main strength of RTK drone surveys compared to traditional geodetic surveys, might not stand because of the limitations of setting ground control points on those areas.

Citizen science information helped to elucidate relief details that were not evident from the geodetic survey alone, but appeared in the drone survey. This highlights the relevance of local information in areas with scarce data. For the preliminary mapping of flood-prone areas in isolated rural areas, this study demonstrated that using the ANADEM DEM, which showed better performance among the satellite products tested, in combination with local interviews, can produce an initial assessment of hazard areas with minimal resources.

Finally, this study highlighted the potential of UAVs as an intermediate solution between low-resolution global models and traditional topographic surveys, filling an important gap in the preliminary assessment of flood risk in urban and peri-urban areas. The ability to generate elevation models at a local scale, combined with operational flexibility and lower cost, makes this approach particularly relevant in contexts where detailed hydraulic data and consistent historical series are nonexistent or limited. In this sense, the integration of drone-derived topography with citizen science information enhances the interpretative robustness of the results by incorporating the historical knowledge of the affected populations as a complementary source of spatial validation.

From a methodological point of view, the results reinforce that the use of UAVs does not replace hydrodynamic modeling but constitutes an effective exploratory and comparative step for prioritizing vulnerable areas and supporting initial planning and risk management decisions. The applicability of the proposed workflow has been applied to low-slope alluvial plains, where topography plays a dominant role in flood propagation, and should be applied with caution in geomorphologically or hydraulically more complex systems, in which physically based approaches are necessary. In such cases, the integration of the methodological approach proposed in this study with hydrodynamic modeling is advisable.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement

The studies involving humans were approved by Plataforma Brasil - CAAE: 61253522.3.0000.5504. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Author contributions

HB: Formal analysis, Writing – review & editing, Data curation, Conceptualization, Writing – original draft. MF: Writing – original draft, Formal analysis, Visualization, Project administration, Writing – review & editing. JT: Formal analysis, Writing – original draft, Conceptualization, Writing – review & editing, Supervision. JS: Writing – original draft, Formal analysis, Visualization, Conceptualization. AV: Conceptualization, Data curation, Writing – original draft. GC: Conceptualization, Writing – original draft, Data curation. PG: Writing – original draft, Data curation, Conceptualization. MB: Writing – original draft, Formal analysis, Visualization. CP: Writing – original draft, Conceptualization, Data curation.

Funding

The author(s) declared that financial support was received for this work and/or its publication. The authors would like to thank the National Council for Scientific and Technological Development (CNPq) for funding the project (Grants Nos. 409527/2021-1 and 440016/2024-0).

Acknowledgments

The authors would like to thank the Postgraduate Program in Natural Disasters at the São Paulo State University “Júlio de Mesquita Filho” (UNESP), as well as the Postgraduate Program in Earth System Science at the National Institute for Space Research (INPE).

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

Abdelkarim, A., Al-Alola, S. S., Alogayell, H. M., Mohamed, S. A., Alkadi, I. I., and Ismail, I. Y. (2020). Integration of GIS-based multicriteria decision analysis and analytic hierarchy process to assess flood hazard on the Al-shamal train pathway in Al-qurayyat region, Kingdom of Saudi Arabia. Water 12:1702. doi: 10.3390/w12061702

Crossref Full Text | Google Scholar

Abebe, Y., Kabir, G., and Tesfamariam, S. (2018). Assessing urban areas vulnerability to pluvial flooding using GIS applications and Bayesian Belief Network model. J. Clean. Prod. 174, 1629–1641. doi: 10.1016/j.jclepro.2017.11.066

Crossref Full Text | Google Scholar

AGISOFT L.L.C. (2014). Agisoft Photoscan Pro. St. Petersburg: AGISOFT L.L.C. Available online at: http://www.agisoft.com (accessed June 22, 2025).

Google Scholar

Amirkhani, M., Ghaemimood, S., von Schreeb, J., El-Khatib, Z., and Yaya, S. (2022). Extreme weather events and death based on temperature and CO₂ emission – a global retrospective study in 77 low-, middle- and high-income countries from 1999 to 2018. Prev. Med. Rep. 28:101846. doi: 10.1016/j.pmedr.2022.101846

PubMed Abstract | Crossref Full Text | Google Scholar

Anderson, K., and Gaston, K. J. (2013). Lightweight unmanned aerial vehicles will revolutionize spatial ecology. Front. Ecol. Environ. 11, 138–146. doi: 10.1890/120150

Crossref Full Text | Google Scholar

Assumpção, T. H., Popescu, I., Jonoski, A., and Solomatine, D. P. (2018). Citizen observations contributing to flood modelling: opportunities and challenges. Hydrol. Earth Syst. Sci. 22, 1473–1489. doi: 10.5194/hess-22-1473-2018

Crossref Full Text | Google Scholar

Benso, M. R. (2025). DEM Accuracy Analysis Toolkit. Computer software. Available online at: https://github.com/marcosrbenso/dem_analysis (accessed June 22, 2025).

Google Scholar

Binns, A. D. (2022). Sustainable development and flood risk management. J. Flood Risk Manag. 15:e12807. doi: 10.1111/jfr3.12807

Crossref Full Text | Google Scholar

Boccardo, P., Chiabrando, F., Dutto, F., Tonolo, F. G., and Lingua, A. (2015). UAV deployment exercise for mapping purposes: evaluation of emergency response applications. Sensors 15, 15717–15737. doi: 10.3390/s150715717

PubMed Abstract | Crossref Full Text | Google Scholar

Bonney, R., Phillips, T. B., Ballard, H. L., and Enck, J. W. (2016). Can citizen science enhance public understanding of science? Public Underst. Sci. 25, 2–16. doi: 10.1177/0963662515607406

PubMed Abstract | Crossref Full Text | Google Scholar

Brasil (2012). Law No. 12,608, April 10, 2012. Establishes the National Policy on Civil Protection and Defense. Available online at: http://www.planalto.gov.br/ccivil_03/_ato2011-2014/2012/lei/l12608.htm (accessed June 22, 2025).

Google Scholar

Buffon, E., Sampaio, T., and da Paz, O. (2018). Veículo aéreo não tripulado (VANT)-aplicação na análise de inundações em áreas urbanas. GOT Rev. Geogr. Ordenamento Territ. 13:85. Portuguese. doi: 10.17127/got/2018.13.004

Crossref Full Text | Google Scholar

Choubin, B., Moradi, E., Golshan, M., Adamowski, J., Sajedi-Hosseini, F., and Mosavi, A. (2019). An ensemble prediction of flood susceptibility using multivariate discriminant analysis, classification and regression trees, and support vector machines. Sci. Total Environ. 651, 2087–2096. doi: 10.1016/j.scitotenv.2018.10.064

PubMed Abstract | Crossref Full Text | Google Scholar

Colomina, I., and Molina, P. (2014). Unmanned aerial systems for photogrammetry and remote sensing: a review. ISPRS J. Photogramm. Remote Sens. 92, 79–97. doi: 10.1016/j.isprsjprs.2014.02.013

Crossref Full Text | Google Scholar

Das, S., and Gupta, A. (2021). Multi-criteria decision-based geospatial mapping of flood susceptibility and temporal hydro-geomorphic changes in the Subarnarekha basin, India. Geosci. Front. 12:101206. doi: 10.1016/j.gsf.2021.101206

Crossref Full Text | Google Scholar

DJI (2021). Phantom 4 Pro. Available online at: https://www.dji.com/phantom-4-pro (accessed May 15, 2021).

Google Scholar

EM-DAT (2015). The Human Cost of Weather-Related Disasters, 1995–2015. Brussels: Centre for Research on the Epidemiology of Disasters

Google Scholar

EMLID (2024). EMLID Studio (Version 1.9). Computer software. Available online at: https://emlid.com/emlid-studio/ (accessed June 22, 2025).

Google Scholar

Farr, T. G., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., et al. (2007). The Shuttle Radar Topography Mission. Rev. Geophys. 45:RG2004. doi: 10.1029/2005RG000183

Crossref Full Text | Google Scholar

Feng, Q., Liu, J., and Gong, J. (2015). UAV remote sensing for urban vegetation mapping using random forest and texture analysis. Remote Sens. 7, 1074–1094. doi: 10.3390/rs70101074

Crossref Full Text | Google Scholar

Fohringer, J., Dransch, D., Kreibich, H., and Schröter, K. (2015). Social media as an information source for rapid flood inundation mapping. Nat. Hazards Earth Syst. Sci. 15, 2725–2738. doi: 10.5194/nhess-15-2725-2015

Crossref Full Text | Google Scholar

Garrote, J. (2022). Free global DEMs and flood modelling—a comparison analysis for the January 2015 flooding event in Mocuba City (Mozambique). Water 14:176. doi: 10.3390/w14020176

Crossref Full Text | Google Scholar

Hagos, Y. G., Andualem, T. G., Yibeltal, M., and Mengie, M. A. (2022). Flood hazard assessment and mapping using GIS integrated with multi-criteria decision analysis in Upper Awash River Basin, Ethiopia. Appl. Water Sci. 12:148. doi: 10.1007/s13201-022-01674-8

Crossref Full Text | Google Scholar

Haklay, M. (2013). “Citizen science and volunteered geographic information – overview and typology of participation,” in Crowdsourcing Geographic Knowledge: Volunteered Geographic Information (VGI) in Theory and Practice, Eds. D. Z. Sui, S. Elwood, and M. F. Goodchild (Berlin: Springer), 105–122.

Google Scholar

IBGE (2022). Censo Demográfico. Available online at: https://www.ibge.gov.br/estatisticas/sociais/populacao/22827-censo-demografico-2022.html (accessed June 22, 2025).

Google Scholar

James, M. R., and Robson, S. (2012). Straightforward reconstruction of 3D surfaces and topography with a camera: accuracy and geoscience application. J. Geophys. Res. Earth Surf. 117:F03017. doi: 10.1029/2011JF002289

Crossref Full Text | Google Scholar

Kovanič, L., Topitzer, B., Petovský, P., Blištan, P., Gergelová, M. B., and Blištanová, M. (2023). Review of photogrammetric and lidar applications of UAV. Appl. Sci. 13:6732. doi: 10.3390/app13116732

Crossref Full Text | Google Scholar

Laipelt, L., Comini de Andrade, B., Collischonn, W., de Amorim Teixeira, A., Paiva, R. C. D., Ruhoff, A., et al. (2024). ANADEM: a digital terrain model for South America. Remote Sens. 16:2321. doi: 10.3390/rs16132321

Crossref Full Text | Google Scholar

Magnini, A., Lombardi, M., Bujari, A., Mattivi, P., Shustikova, I., Persiano, S., et al. (2023). Geomorphic flood hazard mapping: from floodplain delineation to flood hazard characterization. Hydrol. Sci. J. 68, 2388–2403. doi: 10.1080/02626667.2023.2269909

Crossref Full Text | Google Scholar

MapBiomas (2022). Áreas urbanizadas no Brasil 1985–2022. Available online at: https://brasil.mapbiomas.org/wp-content/uploads/sites/4/2023/10/FACT_Areas-Urbanas-no-Brasil_31.10_v2.pdf (accessed June 22, 2025).

Google Scholar

Martins, M., de Souza Lopes, C., Cantarelli Camargo, E., Ivo Pfeifer, M., Paes Marangon, G., and Luis Schünemann, A. (2020). Acurácia de levantamento aéreo realizado por VANT sem o auxílio de pontos de controle. Anais Salão Int. Ensino Pesqui. Extensão 11:2. Portuguese.

Google Scholar

Mazzoleni, M., Paron, P., Reali, A., Juizo, D., and Brandimarte, L. (2020). Testing UAV-derived topography for hydraulic modelling in a tropical environment. Nat. Hazards 103, 139–163. doi: 10.1007/s11069-020-03963-4

Crossref Full Text | Google Scholar

Miller, J. D., and Hutchins, M. (2017). The impacts of urbanisation and climate change on urban flooding and urban water quality: a review of the evidence concerning the United Kingdom. J. Hydrol. Reg. Stud. 12, 345–362. doi: 10.1016/j.ejrh.2017.06.006

Crossref Full Text | Google Scholar

Mohanty, M. P., and Simonovic, S. P. (2022). A comprehensive approach for floodplain mapping through identification of hazard using publicly available data sets over Canada. Water 14:2280. doi: 10.3390/w14142280

Crossref Full Text | Google Scholar

Nardi, F., Cudennec, C., Abrate, T., Allouch, C., Annis, A., Assumpção, T., et al. (2022). Citizens AND HYdrology (CANDHY): conceptualizing a transdisciplinary framework for citizen science addressing hydrological challenges. Hydrol. Sci. J. 67, 2534–2551. doi: 10.1080/02626667.2020.1849707

Crossref Full Text | Google Scholar

Neal, J., Villanueva, I., Wright, N., Willis, T., Fewtrell, T., and Bates, P. (2012). How much physical complexity is needed to model flood inundation?. Hydrol. Process. 26, 2264–2282. doi: 10.1002/hyp.8339

Crossref Full Text | Google Scholar

Njue, N., Kroese, J. S., Gräf, J., Jacobs, S. R., Weeser, B., Breuer, L., et al. (2019). Citizen science in hydrological monitoring and ecosystem services management: state of the art and future prospects. Sci. Total Environ. 693:133531. doi: 10.1016/j.scitotenv.2019.07.337

PubMed Abstract | Crossref Full Text | Google Scholar

Oniga, V.-E., Breaban, A.-I., and Statescu, F. (2018). Determining the optimum number of ground control points for obtaining high precision results based on UAS images. Proceedings 2:352. doi: 10.3390/ecrs-2-05165

Crossref Full Text | Google Scholar

Rahmati, O., Zeinivand, H., and Besharat, M. (2015). Flood hazard zoning in Yasooj region, Iran, using GIS and multi-criteria decision analysis. Geomatics Nat. Hazards Risk 7, 1000–1017. doi: 10.1080/19475705.2015.1045043

Crossref Full Text | Google Scholar

See, L., Mooney, M., Foody, G., Bastin, L., Comber, A., Estima, J., et al. (2016). Crowdsourcing, citizen science or volunteered geographic information? The current state of crowdsourced geographic information. ISPRS Int. J. Geo-Inf. 5:55. doi: 10.3390/ijgi5050055

Crossref Full Text | Google Scholar

Smith, L. S., Liang, Q., James, P., and Lin, W. (2015). Assessing the utility of social media as a data source for flood risk management using a real-time modelling framework. J. Flood Risk Manag. 10, 370–380. doi: 10.1111/jfr3.12154

Crossref Full Text | Google Scholar

Sy, B., Frischknecht, C., Dao, H., Consuegra, D., and Giuliani, G. (2020). Reconstituting past flood events: the contribution of citizen science. Hydrol. Earth Syst. Sci. 24, 61–74. doi: 10.5194/hess-24-61-2020

Crossref Full Text | Google Scholar

Valjarević, A. (2024). GIS-based methods for identifying river networks types and changing river basins. Water Resour. Manag. 38, 5323–5341. doi: 10.1007/s11269-024-03916-7

Crossref Full Text | Google Scholar

Vasconcelos, A. F., Santos, J. C., Cilto, G. G., Galvão, P. R., Fava, M. C., Menezes Filho, F. C. M., et al. (2025). “Contribuições da ciência cidadã para o gerenciamento de riscos de inundações: o caso de Campina do Monte Alegre – SP, Brasil,” in Estudos de Caso em Mapeamento Colaborativo, ed. R. D. Souto (Rio de Janeiro: IVIDES.org), 303–334. Portuguese.

Google Scholar

Vujović, F., Culafić, G., Valjarević, A., Brdanin, E., and Durlević, U. (2024). Comparative geomorphometric analysis of drainage basin using AW3D30 model in ArcGIS and QGIS environment: case study of the Ibar River drainage basin, Montenegro. Agric. For. 70, 217–230. doi: 10.17707/AgricultForest.70.1.15

Crossref Full Text | Google Scholar

Walker, D., Forsythe, N., Parkin, G., and Gowing, J. (2016). Filling the observational void: scientific value and quantitative validation of hydrometeorological data from a community-based monitoring programme. J. Hydrol. 538, 713–725. doi: 10.1016/j.jhydrol.2016.04.062

Crossref Full Text | Google Scholar

Westoby, M. J., Brasington, J., Glasser, N. F., Hambrey, M. J., and Reynolds, J. M. (2012). Structure-from-Motion photogrammetry: a low-cost, effective tool for geoscience applications. Geomorphology 179, 300–314. doi: 10.1016/j.geomorph.2012.08.021

Crossref Full Text | Google Scholar