Methodology for surface water flood modelling based on analysis of digital elevation model and hydrological data

. A methodology for surface water flood modelling is proposed, based on the use of small-scale mapping tools together with hydrological observation data. To reproduce the flooding surface during extreme floods, information on historical maximum water levels was selected, and a digital elevation model (DEM) was used as the cartographic basis. The proposed methodology is universal and provides possibility to determine the boundaries of the potential flooding area in river sections during extreme water levels, to identify objects in the potential flooding area, as well as to make operational decisions to prevent disastrous situations arising from floods and minimize adverse consequences. The results of application of the proposed method for certain river sections of the North Caucasus, a region characterized by a high degree of flood hazard, are presented.


Introduction
Rivers floods are one of the dangerous hydrological phenomena posing a serious threat to the population and infrastructure facilities. In recent decades floods have resulted in greater economic losses worldwide [1,2]. Scientists are therefore faced with the task of not only thoroughly studying the hydrological regime of rivers, but also forecasting possible extreme events and negative consequences associated with them. This will make it possible to develop targeted measures to prevent extreme situations in rivers and reduce the consequences in case of their adverse scenarios [3,4,5]. The case study has been conducted for the North Caucasus characterized by a high degree of natural flood hazard [10]. Related studies of the river water regime are actively conducted in the region [6][7][8][9].
The aim of the study is to develop a methodology for the rapid delineation of potential flood zones that requires a minimum set of input data -a map base and hydrological observation data.
Modelling of extreme flood surfaces provides possibility to determine the boundaries of potentially hazardous river sections and assess the probable damage from their passage [11][12][13]. The flood surfaces can be successfully determined by means of hydrodynamic modelling [13][14][15] and geoinformation analysis tools [16,17] that require a high degree of knowledge of the study area.

Materials and methods
The input data for the proposed method of surface water flood modelling are divided into two components -cartographic and hydrological. The SRTM DEM is used as the cartographic basis [18]. The hydrological input includes extreme water levels, the zero points of hydrological gauges and their geographical coordinates. The proposed method of surface water flood modelling is universal. Any maximum water levels, whether for a certain period of time (year, season, month, water regime phase) or water levels of rare recurrence (P=0,1-5%) can be selected as extreme water levels. Thus, the methodology allows reproducing the situation when water in the rivers reaches extreme levels at one time.
The study considered maximum historical water levels (Hmax) recorded at 252 hydrological stations for the period of 1932-2015. The SRTM DEM with 1" resolution for the territory of the North Caucasus was used.
The ArcGIS software package was used as a medium for the implementation of the methodology. A vector point layer of the hydrological gauges with corresponding hydrological information in the attributive table was created.
When modelling the inundation surface, the slope of the water surface in the crosssectional profile should be minimized. For this propose additional doubling of points at the zero height of the base hydrological gauge H0 on the opposite banks of the river valley was performed (Fig. 1). The elevation of the points corresponding to the gauging station was specified using the DEM; as the same elevation value can correspond to several pixels of the DEM, the points were selected perpendicular to the river channel at the greatest distance from each other.
After doubling the points in the alignment of gauging stations, the absolute values of historical maximum water levels were determined by adding maximum water levels Hmax to the zero points of hydrological gauges H0). Thus, the elevations of two points in the alignment of each gauging station lying on the flood surface, were obtained. Based on these values, interpolation was performed to determine the position of the flood surface on the rivers (Fig.  1). The DEM was then "cut" by the obtained flood surface determining the flooded areas in river valleys including areas not covered by the hydrological observations.

Results
The methodology was used successfully on a sample of 252 hydrological stations in the North Caucasus (~1 station per 1000 km 2 ). The boundaries of the inundation surface were determined quite unambiguously during the passage of historical maximums. Fig. 2 shows the results of testing the proposed methodology for selected river reaches of the North Caucasusthe Terek River, the Sunzha River and the Ulluchay River. The hydrological stations considered in the study are grouped according to the date of the maximum historical water level. The prevalence of gauges with historical maximums in the rivers of the upper Terek River basin was shown to be from 2000 to 2010: 7 gauges out of 16, or 44% of the total number of gauges in the study area (Fig. 2). The historical maximum water level on the Ulluchay River according to the Madzhalis station was also observed in 2000-2010. In the Sunzha River basin, in contrast, maximum water levels were observed before 1980 on most of gauges (3 out of 7, or 43% of all gauges in the study area). An attempt to model flood surface for the North Caucasus area based on data only from 59 hydrological gauges in the study area was not successful due to lower density of the gauging network (~0,2 gauges per 1000 km 2 ). Increasing the accuracy of modelling of the potential river flooding surface based on the proposed methodology could be achieved by the use of DEMs with a higher spatial resolution.
The methodology for surface water flood modelling includes the following main steps (the ArcGIS tools used for its implementation are given): Step 1. Preparation of the hydrological database, namely the collection of information on the extreme water levels Hmax at gauging stations during the period of interest, their geographical coordinates (X, Y) and the zero points of the gauges Н0.
Step 2. Preparation of the DEM. Creation of a point vector layer of hydrological gauges with accumulated information in an attribute table (File -Add data -Add XY data).
Step 3. Bipointing of hydrological stations on either side of the channel at the zero points elevation using a DEM perpendicular to the channel in manual editing mode in a point vector layer (Edit Features -Start Editing).
Step 4. Create an additional field in the attribute table of the point vector layer (Open Attribute Table -Add field) and calculating the absolute height of the points (Spatial Analyst -Extraction -Extract Values to Points).
Step 5. Determination of the absolute height of the inundation surface at hydrological gauges HFSn=Н0+Нmax, where Н0 -the zero points of the gauges in the Baltic system heights, m, Нmax -maximum water level, n -number of hydrological gauges in the survey (Open Attribute Table -Field Calculator: 'Н0' + 'Нmax').
Step 6 Interpolation of the derived HFSn values in the whole study area and creation of a flood surface (ArcToolBox -Spatial Analyst -Interpolation -Kriging). Note: other interpolation methods can be used within the proposed methodology. The Kriging interpolation method in this case showed the best result with the flood surface boundaries following the outline of the river valleys which was not the case with other methods.
Step 7. Calculation of DEM and flood surface difference to determine flood boundary at river sections (ArcToolBox -Spatial Analyst -Map Algebra -Raster Calculator: = 'DEM' -'flood surface'). Note: this step makes it possible to determine the flood boundary not only at the gauging stations, but also for the areas not covered by hydrological observations. However, the results of the construction of flood surfaces for unstudied rivers are not always adequate and require verification with additional information.
Step 8. Selection of pixels by DEM and flood surface difference close to zero to determine the exact boundary of the flooding surface (ArcToolBox -Spatial Analyst -Conditional -Con).
Step 9. Convert flood surface to vector format (ArcToolBox -Conversion Tools -From Raster -Raster To Polygon). Note: the image must be an integer, otherwise the tool will not work. To make the image an integer, apply the Int tool (ArcToolBox -Spatial Analyst -Math -Int).

Conclusion
The study proposes a methodology for surface water flood modelling by means of small-scale mapping tools to enable rapid delineation of potentially flood-prone areas. The methodology requires a minimum set of input data: a digital elevation model of the study area as a cartographic basis, and information on extreme levels in the period under consideration at hydrological stations, their geographical location and the zero points of the gauges. The main advantage of the methodology is its versatility: any value of interest can be set as the water level. The methodology also allows flood levels to be determined at sites not covered by hydrological observations, which, however, requires further investigation.
The introduction of the proposed methodology for modelling flood surfaces into real practice could solve the problem of promptly identifying buffer zones during catastrophic floods in order to develop measures for timely prevention of adverse consequences within dangerous river sections, determine the proportion of population and infrastructure facilities in the potential flood zone and subsequent assessment of damage during their passage.

Acknowledgements
This work was supported by the Russian Foundation for Basic Research (project №20-35-90120).