Sayısal Yükseklik Modellerindeki Mekânsal Çözünürlük Değişkenliğinin Taşkın Tehlike Analizine Etkileri
Taşkın tehlike ve risk çalışmalarında temel altlık veri olarak yüzey topografyasını temsil etmesi bakımından raster tabanlı Sayısal Yükseklik Modelleri (SYM) sıklıkla kullanılmaktadır. Bu çalışmanın amacı; küresel ve lokal ölçekte kullanılan ve birçok çalışmalara altlık oluşturan farklı kaynaklı ve farklı çözünürlükteki SYM’lerle taşkın tehlike analizleri gerçekleştirerek, Ulus yerleşmesi (Bartın) temelindeki tehlikenin değişkenliğini ortaya koymaktır. Bu amaç doğrultusunda atlık verileri, Ulus Çayı havzası ve Ulus yerleşmesi için elde edilen MERIT 90m, FABDEM 30m, TopoSYM 10m, SYM5m, LiDAR 1m ve İHA 0,1m çözünürlüklü SYM verileri, Ulus yerleşmesine akış gösteren Ulus üst kolu, Süleyman, Alpı ve Eldeş akarsularının SWAT yağış-akış modeliyle üretilmiş 500 yıllık akımları oluşturmaktadır. Bu veriler ile mekânsal çözünürlük değişkenliğini iyi ortaya koyabilmek için sabit Manning n değeri kullanılarak (n=0.035), 2 boyutlu LISFLOOD-FP hidrodinamik model temelinde taşkın tehlike analizleri gerçekleştirilmiştir. Sonuç olarak düşük çözünürlükten yüksek çözünürlüğe model zamanı ve ortalama hesaplama hataları artarken, suyun yayılışı, insan ve bina için üretilen tehlike sınıflarının alansal dağılışı azalış göstermiştir. Bölgesel yapılacak çalışmalarda FABDEM verisi daha avantajlı iken havza bazlı yapılacak çalışmalarda LiDAR verisi veya üzerindeki bina ve bitki örtüsü topluluklarına ait yüksekliklerin temizlenmesi koşuluyla SYM5 verisi kullanılabilir verilerdir.
The Effects of Spatial Resolution Variability of Digital Elevation Models on Flood Hazard Analysis
Raster-based Digital Elevation Models (DEMs) represent the surface topography as the primary input in flood hazard and risk studies. The study aims to reveal the variability of the hazard at the base of the Ulus settlement by performing flood hazard analyses with different source and resolution DEMs, which are used on a global and local scale and form a primary input for many studies. For this purpose, DEMs data, such as MERIT 90m, FABDEM 30m, TopoDEM 10m, DEM5m, LiDAR 1m, and UAV 0.1m, for the Ulus River basin and settlement and the 500-year flood produced for the river tributaries using the SWAT rainfall-runoff model were used. To examine spatial resolution variability, flood hazard analyses were performed based on the two-dimensional LISFLOODFP hydrodynamic model, using a fixed Manning n value (n=0.035). As a result, although there is an increase in cost, time, and model instabilities from low resolution to high resolution, it is essential to choose the most appropriate DEM according to the required detail and scale of the hazard analysis to be able to obtain more accurate results. While the model time and average computational errors from low resolution to high resolution increased, the water extent and the spatial distribution of the hazard classes produced for people and buildings decreased. The FABDEM data is more advantageous in regional studies than others, whereas the LiDAR data can be used in basin-scaled studies. In addition, the DEM5 data also can be used in basin-scaled studies after clearing the heights of the buildings and vegetation groups.
Flood hazard analysis is obtained by evaluating flood depths and velocity because combining these two features of floods increases the hazard size for people and structures. To obtain the depth and velocity of the flood water most accurately, the surface topography and the structures should be represented in the base data. Digital Elevation Models (DEMs), which contain this information, are the primary data used in the hydraulic modelling of floods and checking their accuracy. The present study aims to reveal the variability of the hazard at the base of the Ulus settlement (Bartın) by performing flood hazard analyses with different source and resolution DEMs, which are used in global and local scale studies as primary input data. For this purpose, DEM data, discharge data, and hydrodynamic model selection constitute the main components of the study. Six different data sets with different resolutions were used in this study. These are: MERIT DEM (90m), FABDEM (30m), TopoDEM (10m), DEM5 (5m), LiDAR DEM (1m), and UAV DEM (0.1m). The first two of these datasets (MERIT and FABDEM) are global datasets produced by Yamazaki et al. (2017) and Hawker et al. (2022), respectively. TopoDEM and DEM5 are national datasets, while LiDAR and UAV data are produced by project-based high-resolution data. The TopoDEM was generated from topographic contours scaled at 1:25000 using the TopotoRaster function in ArcGIS Pro. The DEM5 data was produced by the General Directorate of Mapping from the stereo photos taken in 2013, with ±3m vertical accuracy. The LiDAR and UAV data generation steps are provided in Figures 4 and 5. Since there is no gauge data in each tributary converging in the Ulus settlement, the SWAT (The Soil & Water Assessment Tool) model was used to obtain the data to simulate the hydraulic model. The SWAT rainfall-runoff model is a physical-based and semidistributed model and overlays the digital elevation model (DEM), land-use, soil, and slope to obtain the Hydrological Response Unit (HRU) level and force the model via weather data based on the water balance equation at HRU. The precipitation (mm), minimum/ maximum temperature (o C), and relative humidity (%) of the Ulus weather station (17615) between 1970 and 2020 and the gauge data of Emirce-Ulus station (D13A088) between 2016 and 2020 were used in the model (Figure 1). The sensitive parameters model was manually calibrated based on the suggestion of Abbaspour et al. (2015), with the NSE value being found to be 0.3 on a daily scale, which is sufficient to use in the model. For steady flow simulation, the Generalised Extreme Value (GEV) distribution includes three distributions. Namely, Gumbel, Frèchet, and Weibull were fitted to simulate each probability (return period) based on the annual maximum series (AMAX). Finally, the 500-year flow data was produced for the steady-state flow simulations (Fig. 6). We ran simulations for the 500-year flow data for four main tributaries using the two-dimensional hydrodynamic model LISFLOODFP to produce flood depth and velocity data. While modeling is done by distributing the 500-year flow of river tributaries to the back of the basin with Shreve order (1966) in MERIT, FABDEM, TopoDEM, and DEM5 data. The LiDAR and UAV DEMs-based modeling was performed with point-source flow data about 1 km behind the Ulus settlement. The results indicate that flood modeling studies, especially in areas where settlements spread up to the upper basins, are best to make modeling studies and hazard analyses involving the river networks in the entire basin instead of settlement based. However, the cost of acquiring data and difficulties and increases in computational power from low resolution to high resolution in flood modeling (Table 1) make it difficult to carry out these studies with high resolution across the entire basin. The flood extent of the 500-year flow has decreased 50% from the MERIT to the UAV data (Figure 9). Although the maximum depth values vary between 4.4 and 7.7 m, the increase in maximum velocity is more due to the increase in resolution. The high flood hazard class is distributed over a large area in flood hazard analyses, based on MERIT, FABDEM, and TopoDEM data with low resolution and no riverbed form. In the DEM5, LiDAR, and UAV data, where the bed form started to occur, there was a decrease of up to 50% in the areal distributions of this class. This situation has made the number of buildings exposed to floods and intervention more acceptable in the building-based flood hazard analysis, especially in the very high hazard class, from 152 buildings in the MERIT data to 2 and 0 buildings in the LiDAR and UAV data, respectively (Figure 10). As a result, flood hazard analyses vary depending on the area to be studied, the data detail, and computational power. It should not be forgotten that the results will vary depending on the characteristics of the projects or academic research and that evaluations and risk studies should be carried out within the resolution framework.
