{"id":29093,"date":"2024-04-16T08:52:13","date_gmt":"2024-04-16T07:52:13","guid":{"rendered":"https:\/\/www.vtei.cz\/2024\/04\/porovnani-kvality-schematizace-udolnice-extrahovane-z-dat-dmr-4g-dmr-5g-a-jeho-derivatu-2\/"},"modified":"2024-08-21T20:02:21","modified_gmt":"2024-08-21T19:02:21","slug":"comparison-of-the-quality-of-thalweg-lines-extracted-from-data-of-dmr-4g-dmr-5g-and-its-derivatives","status":"publish","type":"post","link":"https:\/\/www.vtei.cz\/en\/2024\/04\/comparison-of-the-quality-of-thalweg-lines-extracted-from-data-of-dmr-4g-dmr-5g-and-its-derivatives\/","title":{"rendered":"Comparison of the quality of thalweg lines extracted from data of DMR 4G, DMR 5G and its derivatives"},"content":{"rendered":"

ABSTRACT<\/h2>\n

Determining the\u00a0gradient of\u00a0watercourses in\u00a0the\u00a0case of\u00a0local applications is\u00a0a\u00a0common problem, which is\u00a0most often dealt with by\u00a0geodetic surveying. However, determining the\u00a0gradient of\u00a0all watercourses in\u00a0the\u00a0Czech Republic is\u00a0a\u00a0challenge. The use of\u00a0geodetic methods on\u00a0such a\u00a0scale is\u00a0usually unrealistic. Therefore, it\u00a0is\u00a0necessary to\u00a0choose a\u00a0different approach, such as\u00a0the\u00a0extraction of\u00a0the\u00a0gradient lines from other already existing elevation data. The\u00a0DMR\u00a04G and DMR 5G are elevation models currently available for the\u00a0Czech Republic. For\u00a0the\u00a0extraction of\u00a0gradient lines, it\u00a0is\u00a0necessary to\u00a0create a\u00a0digital terrain model (DTM) from the\u00a0available datasets. Various interpolation methods are used for this. But which of\u00a0the\u00a0available interpolation methods is\u00a0the\u00a0most suitable? What role does the\u00a0size of\u00a0spatial resolution play in\u00a0the\u00a0quality of\u00a0altimetry representation and subsequent sizes of\u00a0the\u00a0stored DTMs? To\u00a0find answers to\u00a0these questions, we\u00a0chose four study sites (fourth order catchments) in\u00a0the\u00a0Otava river basin. Eight different DTMs were then created at\u00a0each site, which were then compared. The results show that choice of\u00a0raster size has a\u00a0significantly greater influence on\u00a0the\u00a0resulting quality of\u00a0the\u00a0gradient lines than the\u00a0choice of\u00a0interpolation method in\u00a0the\u00a0case of\u00a0DTM creation from DMR 5G data. DTM from DMR 4G data gives worse results than from DMR 5G at\u00a0the\u00a0same raster resolution.<\/p>\n

INTRODUCTION<\/h2>\n

Determining the\u00a0longitudinal slope of\u00a0a\u00a0watercourse bed is\u00a0important from the\u00a0point of\u00a0view of\u00a0a\u00a0wide range of\u00a0engineering and scientific applications, such as\u00a0analysis of\u00a0bed stability, detection of\u00a0transverse obstacles, design of\u00a0bed modification, and assessment of\u00a0watercourse hydropower potential. In\u00a0most cases, local studies requiring the\u00a0slope of\u00a0a\u00a0watercourse bed use geodetic methods to\u00a0survey them (tachymetry, positioning of\u00a0GPS points). Geodetic approaches are very accurate; however, their use in\u00a0the\u00a0case of\u00a0regional, district, or\u00a0nationwide projects is\u00a0not realistic. To\u00a0survey an\u00a0area the\u00a0size of\u00a0the\u00a0Czech Republic, spatial data collection methods can be\u00a0used. Satellite measurements or\u00a0airborne laser scanning (ALS) methods are mainly used for this. Even today, satellites produce altimetry data with an\u00a0error of\u00a0several metres\u00a0[1]. In\u00a0contrast, ALS methods are able to\u00a0achieve an\u00a0error of\u00a0only a\u00a0few dozen cm\u00a0[2,\u00a03]. More recent studies report accuracy of\u00a0a\u00a0few cm\u00a0[4]. Scanners with a\u00a0beam with a\u00a0wavelength close to\u00a0the\u00a0infrared spectrum are most often used for scanning the\u00a0Earth\u2019s\u00a0surface. A\u00a0specific feature of\u00a0using infrared beams is\u00a0the\u00a0inability to\u00a0measure under the\u00a0water surface because water surface absorbs them. In\u00a0such a\u00a0location, the\u00a0beam does not return to\u00a0the\u00a0measuring device, and thus does not determine the\u00a0height mode. The advantage is\u00a0a\u00a0clear distinction between the\u00a0water surface and the\u00a0solid Earth surface\u00a0[5]. However, there are variants of\u00a0ALS that combine laser beams with different wavelengths (infrared and blue-green), which can also be\u00a0used for scanning the\u00a0terrain under the\u00a0water surface\u00a0[6].<\/p>\n

The ALS method was used to\u00a0survey the\u00a0entire Czech Republic in\u00a02009\u20132013. The measurement was carried out using LiteMapper 6800 device from IGI mbH using RIEGL LMS \u2013 Q680 aerial laser scanner. The measuring equipment was placed in\u00a0a\u00a0special L 410 FG\u00a0aircraft. Scanning was done from an\u00a0average height of\u00a01,200\u00a0m or\u00a01,400\u00a0m\u00a0[7]. To\u00a0scan the\u00a0surface, RIEGL LMS \u2013 Q680 laser scanner uses a\u00a0beam with a\u00a0wavelength close to\u00a0infrared spectrum\u00a0[8]. The products of\u00a0this focus are DMR 5G, DMR 4G and DMP 1G data sets. The first product available to\u00a0users was DMR 4G data. The data can be\u00a0found in\u00a0the\u00a0form of\u00a0XYZ points at\u00a0a\u00a0regular spacing of\u00a05\u00a0\u00d7\u00a05\u00a0m. The height accuracy of\u00a0this data is\u00a00.3\u00a0m in\u00a0open terrain and 1\u00a0m in\u00a0densely built-up\u00a0areas or\u00a0forest cover. A\u00a0certain limitation of\u00a0this data layer may also be\u00a0the\u00a0reduced ability to\u00a0describe fracture edges, which is\u00a0based on\u00a0the\u00a0minimum spacing of\u00a0points\u00a0[9]. DMR 5G data is\u00a0accessible in\u00a0the\u00a0form of\u00a0irregularly spaced XYZ points. The height accuracy of\u00a0this data is\u00a00.18\u00a0m in\u00a0open terrain and 0.3\u00a0m in\u00a0densely built-up\u00a0areas or\u00a0forest cover. DMR 5G data is\u00a0able to\u00a0better describe terrain breaks and edges. Their disadvantage can be\u00a0their volume, which is\u00a0related to\u00a0their point density\u00a0[10]. DMP 1G data displays a\u00a0digital model of\u00a0the\u00a0surface. This means that they also contain forest stands and buildings (listed in\u00a0the\u00a0real estate cadastre). However, in\u00a0open terrain the\u00a0data is\u00a0identical to\u00a0DMR 5G data\u00a0[11].<\/p>\n

It\u00a0is\u00a0usually not possible to\u00a0compare directly the\u00a0quality of\u00a0the\u00a0representation of\u00a0the\u00a0Earth\u2019s\u00a0surface with DMR 4G and DMR 5G data. The reason is\u00a0a\u00a0different position of\u00a0source points in\u00a0individual data sets. The solution to\u00a0this problem is\u00a0usually the\u00a0use of\u00a0interpolation methods, on\u00a0the\u00a0basis of\u00a0which DMTs with identical resolution are created, which are then compared. Another possibility is\u00a0the\u00a0use of\u00a03D control lines. Commonly used interpolation methods are Delaunay triangulation (TIN), inverse distances (IDW), minimum curvature (Spline), Natural Neighbor, or\u00a0Kriging\u00a0[12]. Evaluating the\u00a0comparison of\u00a0the\u00a0interpolation method effect on\u00a0the\u00a0resulting quality of\u00a0DMT based on\u00a0DMR 5G data shows that different interpolation methods achieve comparable results both in\u00a0open and in\u00a0forested terrain. This is\u00a0due to\u00a0the\u00a0high density of\u00a0DMR 5G data\u00a0[13].<\/p>\n

MATERIAL AND METHODS<\/h2>\n

Study sites<\/h3>\n

Novosedelsk\u00fd stream \u2013 site No. 1<\/strong><\/span><\/p>\n

The site is\u00a0located southwest of\u00a0the\u00a0town of\u00a0Strakonice and is\u00a0part of\u00a0the\u00a0\u0160umava foothills. From a\u00a0morphological point of\u00a0view, the\u00a0area is\u00a0located at\u00a0altitudes ranging from 446.75\u00a0m to\u00a0864.131\u00a0m above sea level (a.s.l.), with a\u00a0total height difference of\u00a0417.38\u00a0m a.s.l. The highest point of\u00a0the\u00a0area is\u00a0in\u00a0the\u00a0south-eastern part and, conversely, the\u00a0lowest point occurs in\u00a0the\u00a0north-eastern part of\u00a0the\u00a0site. The average altitude of\u00a0the\u00a0site is\u00a0636.386\u00a0m a.s.l.<\/p>\n

\u017divn\u00fd stream \u2013 site No. 2<\/strong><\/span><\/p>\n

The site is\u00a0located southeast of\u00a0the\u00a0built-up\u00a0area of\u00a0the\u00a0town of\u00a0Prachatice and is\u00a0also part of\u00a0the\u00a0\u0160umava foothills. The altitude ranges from 546.89 to\u00a01,094.06\u00a0m\u00a0a.s.l., with a\u00a0total height difference of\u00a0547.17\u00a0m a.s.l. The highest point in\u00a0the\u00a0area is\u00a0Lib\u00edn hill in\u00a0the\u00a0eastern part of\u00a0the\u00a0site, and the\u00a0lowest altitudes occur in\u00a0the\u00a0valley where the\u00a0\u017divn\u00fd potok flows. The average altitude of\u00a0the\u00a0site is\u00a0766.67\u00a0m\u00a0a.s.l.<\/p>\n

\u0160irovsk\u00e1 stoka \u2013 site No. 3<\/strong><\/span><\/p>\n

The site is\u00a0located south of\u00a0the\u00a0town of\u00a0Vod\u0148any and is\u00a0part of\u00a0the\u00a0\u010cesk\u00e9 Bud\u011bjovice Basin. From a\u00a0morphological point of\u00a0view, the\u00a0site is\u00a0located at\u00a0altitudes ranging from 388.49 to\u00a0619.36\u00a0m a.s.l., with a\u00a0total height difference of\u00a0230.86\u00a0m a.s.l. The highest point is\u00a0Holi\u010dka hill in\u00a0the\u00a0south-eastern part of\u00a0the\u00a0site. In\u00a0contrast, the\u00a0lowest point occurs in\u00a0its northeastern part. The\u00a0average altitude of\u00a0the\u00a0site is\u00a0451.23\u00a0m a.s.l.<\/p>\n

Vydra \u2013 site No. 4<\/strong><\/span><\/p>\n

The site is\u00a0located south of\u00a0the\u00a0village of\u00a0Modrava, which is\u00a0part of\u00a0\u0160umava National Park. From a\u00a0morphological point of\u00a0view, it\u00a0is\u00a0located at\u00a0altitudes ranging from 1,035.32 to\u00a01,372.32\u00a0m a.s.l., with a\u00a0total difference in\u00a0height of\u00a0336.895\u00a0m\u00a0a.s.l. The highest points of\u00a0the\u00a0area border the\u00a0southern part of\u00a0the\u00a0study site and are formed by\u00a0Blatn\u00fd hill, Studen\u00e1 hora, \u0160pi\u010dn\u00edk, Hrani\u010dn\u00ed hora, and Velk\u00e1 and Mal\u00e1 Mokr\u016fvka. Towards the\u00a0north of\u00a0the\u00a0area there is\u00a0a\u00a0significant decrease in\u00a0altitude parallel to\u00a0the\u00a0Vydra riverbed. The average altitude of\u00a0the\u00a0site is\u00a01,195.12\u00a0m a.s.l.<\/p>\n

The overview of\u00a0watercourses in\u00a0the\u00a0study sites is\u00a0given in\u00a0Tab. 1. The location of\u00a0the\u00a0study sites within the\u00a0Czech Republic is\u00a0shown in\u00a0Fig.\u00a01<\/em>.<\/p>\n

Tab. 1. Specifications of watercourses at study sites<\/em><\/h5>\n\"\"<\/a>\n
\"\"<\/a>Obr.\u00a01. Vybran\u00e1 povod\u00ed IV. \u0159\u00e1du a\u00a0jejich lokalizace v\u00a0r\u00e1mci \u010cR. Barevn\u011b (oran\u017eov\u00e1, modr\u00e1, zelen\u00e1, \u010derven\u00e1) jsou vyzna\u010deny rozsahy (rozvodnice) jednotliv\u00fdch povod\u00ed k\u00a0p\u0159\u00edslu\u0161n\u00fdm vodn\u00edm tok\u016fm. Stejnou barvou, jako maj\u00ed rozvodnice, je pak vyzna\u010dena poloha dan\u00e9ho povod\u00ed v\u00a0r\u00e1mci \u010cR. Tenk\u00e1 tmav\u011b modr\u00e1 linie ukazuje pr\u016fb\u011bh vodn\u00edch tok\u016f v\u00a0r\u00e1mci jejich povod\u00ed<\/h6>\n
Fig.\u00a01. Selected 4th order basins and their localization in the Czech Republic. The extents (watershed boundaries) of individual watersheds to the relevant watercourses are marked in\u00a0color (orange, blue, green, red). The location of the watershed within the Czech Republic is marked with the same color as its watershed boundary. The thin dark blue line shows the\u00a0position of the given watercourse in its catchment<\/h6>\n

Data description<\/h3>\n

When defining a\u00a0suitable digital relief model (DMR) for determining the\u00a0gradient on\u00a0individual watercourses, two basic products of\u00a0the\u00a0Land Survey Office were used \u2013 the\u00a0Digital Terrain Model of\u00a0the\u00a0Czech Republic of\u00a0the\u00a04th generation (DMR 4G) and the\u00a0Digital Terrain Model of\u00a0the\u00a0Czech Republic of\u00a0the\u00a05th\u00a0generation (DMR 5G)\u00a0[7].<\/p>\n

The fourth-generation digital terrain model of\u00a0the\u00a0Czech Republic represents the\u00a0natural or\u00a0human-modified Earth\u2019s\u00a0surface in\u00a0digital form in\u00a0via the\u00a0heights of\u00a0discrete points in\u00a0a\u00a0regular raster (5\u00a0\u00d7\u00a05\u00a0m) of\u00a0points with a\u00a0complete mean height error of\u00a00.3\u00a0m in\u00a0open terrain and 1\u00a0m in\u00a0forested terrain\u00a0[9].<\/p>\n

The fifth-generation digital terrain model of\u00a0the\u00a0Czech Republic represents natural or\u00a0human-modified Earth surface in\u00a0digital form via the\u00a0heights of\u00a0discrete points with a\u00a0total mean height error of\u00a00.18\u00a0m in\u00a0open terrain and 0.3\u00a0m in\u00a0forested terrain\u00a0[10].<\/p>\n

The watercourse axes for the\u00a0study sites were taken from DIBAVOD (DIgit\u00e1ln\u00ed B\u00c1ze VOdohospod\u00e1\u0159sk\u00fdch Dat; Digital Database of\u00a0Water Management Data). This is\u00a0a\u00a0water management extension of\u00a0ZABAGED (Z\u00e1kladn\u00ed b\u00e1ze geografick\u00fdch dat; The Fundamental Base of\u00a0Geographic Data of\u00a0the\u00a0Czech Republic). Specifically, layer A03 \u2013 watercourse (rough sections) was used, last updated on\u00a05th June 2006. It\u00a0is\u00a0a\u00a0section river model of\u00a0main watercourses of\u00a0fourth order catchments. The data is\u00a0vector oriented in\u00a0the\u00a0direction of\u00a0flow and provided in\u00a0ESRI format\u00a0[14].<\/p>\n

All data used in\u00a0this article were in\u00a0the\u00a0S-JTSK \/ Krovak East North coordinate system (EPSG 5514) and the\u00a0Baltic height system after levelling (EPSG 5705).<\/p>\n

METHODOLOGY<\/h2>\n

Creation of\u00a0digital terrain models<\/h3>\n

Terrain models were created in\u00a0the\u00a0ArcGIS Desktop environment. DMR 4G and DMR 5G datasets were used as\u00a0input data for DMT creation. 8 DMTs were created for each site, i.e. 32 DMTs in\u00a0total. The models can be\u00a0divided into four groups based on\u00a0the\u00a0use of\u00a0their data source and the\u00a0interpolation method used for their creation. The first group of\u00a0models was created from the\u00a0DMR 4G dataset. Its representative is\u00a0the\u00a0ras4G_5 model. It\u00a0is\u00a0a\u00a0raster model with a\u00a05\u00a0\u00d7\u00a05\u00a0m raster resolution produced by\u00a0the\u00a0Inverse Distance Weighting (IDW) method. The second group of\u00a0models was also created using the\u00a0IDW method, but from DMR 5G data. DMTs in\u00a0this group differ from each other only in\u00a0raster resolution. Three raster sizes of\u00a01 m, 5\u00a0m and 10\u00a0m are used. The models are then labelled IDW_1, IDW_5 and IDW_10. The third group of\u00a0models consists of\u00a0tin5G. This is\u00a0a\u00a0TIN terrain model created from DMR 5G data. The fourth group of\u00a0models is\u00a0based on\u00a0the\u00a0TIN model from the\u00a0third group, which was subsequently converted to\u00a0rasters using the\u00a0TinToRaster function. The models are labelled TTR_1, TTR_5 and TTR_10. They differ from each other only in\u00a0the\u00a0resulting raster size to\u00a0which the\u00a0models were transformed when they were converted from TIN format to\u00a0raster format. A\u00a0simple overview of\u00a0DMT for each site and their specifications are given in\u00a0Tab. 2.<\/p>\n

Tab. 2. List of terrain models built for each study site<\/em>\u00a0<\/em><\/h5>\n\"\"<\/a>\n

Extracting the\u00a03D axis of\u00a0watercourses<\/h3>\n

The 3D axes of\u00a0the\u00a0most important watercourses in\u00a0the\u00a0study sites (Fig.\u00a01<\/span><\/em>) were extracted from the\u00a0prepared DMTs using the\u00a0Interpolate Shape<\/span> function. The\u00a0identical Sampling Distance<\/span> parameter was set for all extracted 3D line, which guaranteed that the\u00a0height value on\u00a0the\u00a0watercourse line was always determined by\u00a0the\u00a0program for identical stationing. This is\u00a0a\u00a0basic condition for the\u00a0possibility of\u00a0comparing different height lines of\u00a0one watercourse with each other. The 3D lines were exported to\u00a0a\u00a0text file using the\u00a0Profile Graph<\/span> function, where they were prepared for further comparison.<\/span><\/p>\n

Evaluation of\u00a03D lines<\/h3>\n

The actual processing was carried out in\u00a0the\u00a0R program. The 3D lines of\u00a0the\u00a0watercourse were loaded for individual sites and processed. DMT tin5G was always chosen as\u00a0a\u00a0reference for other raster DMTs of\u00a0the\u00a0given site. Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) metrics were used to\u00a0determine the\u00a0degree of\u00a0agreement.<\/span><\/p>\n\"\"<\/a>\n

where:<\/p>\n

ElevDEM\u00a0\u00a0\u00a0 is\u00a0\u00a0\u00a0\u00a0 the elevation value (m) extracted from each DMT (ras4G_5, IDW_1, IDW_5, IDW_10, TTR_1, TTR_5, TTR_10)<\/p>\n

ElevRef\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 \u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 its corresponding elevation from the\u00a0reference DMT\u00a0(tin5G)<\/p>\n

N\u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 number of height records on a given watercourse line<\/p>\n

Comparison of\u00a0DMT size when stored on\u00a0Hard Disk Drive<\/h3>\n

In\u00a0this comparison, the\u00a0size of\u00a0individual DMTs was determined when stored on\u00a0a\u00a0Hard Disk Drive (HDD). Subsequently, the\u00a0relative sizes of\u00a0the\u00a0raster models compared to\u00a0the\u00a0comparative TIN models were calculated.<\/p>\n

RESULTS<\/h2>\n

When comparing the\u00a0quality of\u00a0the\u00a0height representation, it\u00a0was found that the\u00a0DMTs with a\u00a0raster size of\u00a01\u00a0m (DMT TTR_1 and IDW_1) showed the\u00a0lowest mean errors. The worst results were achieved by\u00a0DMTs with a\u00a0raster size of\u00a010\u00a0m (TTR_10, IDW_10), and the\u00a0ras4G_5 model also achieved similarly poor results.<\/p>\n

A\u00a0visual comparison of\u00a0the\u00a0quality of\u00a0the\u00a0topographical description for selected watercourse sections in\u00a0the\u00a0Loc_4 site is\u00a0shown in\u00a0Fig.\u00a02. In\u00a0section U1, all DMTs show a\u00a0similar quality of\u00a0schematization. The only significantly different DMT is\u00a0ras4G. In\u00a0sections U2 and U3, a\u00a0more significant deviation of\u00a0the\u00a010\u00a0m resolution models and the\u00a0ras4G model can be\u00a0seen.<\/p>\n\"\"<\/a>\n

<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
<\/h6>\n
Fig.\u00a02. Visual comparison of the quality of the slope lines extracted from the compared DTMs at the Loc_4 site<\/h6>\n

Models TTR_1 and IDW_1 showed the\u00a0lowest mean absolute error (MAE). The mean error for TTR_1 was 0.02\u00a0m. The range of\u00a0values ranged from 0.01\u00a0m (Loc_1 and Loc_4) to\u00a00.03\u00a0m (Loc_3). IDW_1 achieved a\u00a0mean error of\u00a00.09\u00a0m with a\u00a0range of\u00a00.06\u00a0m (Loc_4) to\u00a00.11\u00a0m (Loc_2). In\u00a0contrast, the\u00a0highest MAEs were found for the\u00a0IDW_10 and TTR_10 models (both identically 0.36\u00a0m); with minimum values of\u00a00.24\u00a0m. IDW_10 and TTR_10, they had almost identical error values even for the\u00a0corresponding sites. The TTR_5 and IDW_5 models achieved an\u00a0MAE of\u00a0around 0.2\u00a0m. In\u00a0contrast, the\u00a0ras4G_5 model (same resolution) gave an\u00a0error of\u00a00.31\u00a0m. The overall overview of\u00a0the\u00a0MAE values is\u00a0shown in\u00a0Tab. 3<\/em>.<\/p>\n

Tab. 3. Summary of achieved MAE values<\/em>\u00a0<\/em><\/h5>\n\"\"<\/a>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n
<\/h5>\n

Also, the\u00a0RMSE results show that the\u00a0best values were achieved for the\u00a0TTR_1 model, where the\u00a0average RMSE error was 0.03\u00a0m. The worst results were again detected for the\u00a0TTR_10 and IDW_10 models. Other values follow similar trends to\u00a0the\u00a0MAE values. The overall overview of\u00a0RMSE values is\u00a0shown in\u00a0Tab. 4.<\/em><\/p>\n

<\/h5>\n
Tab. 4. Summary of achieved RMSE values<\/em>\u00a0<\/strong>\"\"<\/a><\/h5>\n

The physical size of\u00a0individual DMTs when stored on\u00a0a\u00a0HDD was also evaluated. TIN models have the\u00a0biggeststorage requirements. The only exception is\u00a0at\u00a0the\u00a0site LOC_3, where the\u00a0raster models with a\u00a0resolution of\u00a01\u00a0m are larger. The other DMTs with a\u00a01\u00a0m raster have a\u00a0size in\u00a0the\u00a0range of\u00a050\u201375\u00a0%. The 5\u00a0m raster models have a\u00a0consistent size ranging from 1.9\u20135.2\u00a0%. The 10\u00a0m raster models range within 0.5\u20131.3\u00a0%. The complete list of\u00a0absolute and relative values of\u00a0DMT sizes when saved to\u00a0disk is\u00a0shown in\u00a0Tab. 5<\/em>.<\/p>\n

Tab. 5. Comparison of the amount of memory needed to store a given DTM on HDD<\/em><\/h5>\n
\"\"<\/a> <\/em><\/h5>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

<\/h2>\n

DISCUSSION<\/h2>\n

The tin5G model was chosen as\u00a0the\u00a0reference DMT. This model uses the\u00a0maximum potential of\u00a0the\u00a0DMR 5G data set, i.e. all points and their absolute values for the\u00a0creation of\u00a0a\u00a0complete digital terrain model. In\u00a0contrast, raster models average the\u00a0available point values within their raster. The fact that a\u00a0large number of\u00a0fourth order watercourses flow through the\u00a0forested area also contributed to\u00a0this choice. In\u00a0such conditions, the\u00a0TIN terrain model provides the\u00a0best results\u00a0[13].<\/p>\n

There are multiple interpolation methods for creating DMT. In\u00a0this paper, IDW methods and a\u00a0combined DMT creation approach were used, when a\u00a0TIN model was first created which was then transformed into a\u00a0raster of\u00a0the\u00a0given size. The reason for choosing IDW and the\u00a0combined approach was the\u00a0speed of\u00a0creating these terrain models in\u00a0the\u00a0R programming environment, which is\u00a0planned to\u00a0be\u00a0used to\u00a0process data for the\u00a0entire Czech Republic. The fact that interpolation methods with high point density give the\u00a0same results was also taken into account\u00a0[13].<\/p>\n

In\u00a0the\u00a0Czech Republic, in\u00a0addition to\u00a0DMR 4G and DMR 5G data, it\u00a0is\u00a0also possible to\u00a0use 3D contours from the\u00a0ZABAGED data layer to\u00a0create DMT. These data were not included in\u00a0the\u00a0study; the\u00a0decision was based on\u00a0a\u00a0literature search. The basis for ZABAGED 3D is\u00a0the\u00a0ZM\u00a010 map from 1971\u20131988. These data are burdened by\u00a0a\u00a0greater degree of\u00a0obsolescence (although some map sheets have been updated). Another disadvantage is\u00a0systematic overestimation, which on\u00a0average amounts to\u00a00.23\u00a0m compared to\u00a0DMR 5G data. The description using contours also touches on\u00a0the\u00a0issue of\u00a0schematizing small terrain formations (small ridges and valleys)\u00a0[15].<\/p>\n

The results of\u00a0this paper show the\u00a0ability of\u00a0individual DMTs to\u00a0schematize the\u00a0height of\u00a0watercourses (thalwegs). The primary uncertainty of\u00a0height schematization comes from the\u00a0specifications of\u00a0the\u00a0source data\u00a0[7, 9]. The authors are aware of\u00a0these specifications. They are also aware of\u00a0the\u00a0limitations resulting from the\u00a0ALS technology itself that was used for their acquisition (inability to\u00a0scan watercourse beds). Another uncertainty is\u00a0the\u00a0quality of\u00a0watercourse axis schematization in\u00a0the\u00a0DIBAVOD database, especially in\u00a0forested terrain. In\u00a0these places, the\u00a0axis of\u00a0the\u00a0watercourse may be\u00a0guided outside the\u00a0actual watercourse bed. For the\u00a0purpose of\u00a0this study, it\u00a0would be\u00a0possible to\u00a0create the\u00a0watercourse axes manually; however, it\u00a0is\u00a0unrealistic for the\u00a0application in\u00a0the\u00a0entire area of\u00a0the\u00a0Czech Republic.<\/p>\n

When comparing MAEs, it\u00a0can be\u00a0tentatively concluded that raster models with a\u00a0raster size of\u00a01\u00a0m show better results than models with a\u00a0larger raster size. The surprise was that the\u00a0ras4G_5 model, which has a\u00a0raster size of\u00a05, gives similar results to\u00a0models with a\u00a0raster size of\u00a010 (IDW_10, TTR_10). While maintaining this level of\u00a0schematization quality, it\u00a0would be\u00a0worth considering whether to\u00a0choose other models with a\u00a010\u00a0m raster instead of\u00a0the\u00a0ras4G_5 model, which are also smaller (in\u00a0terms of\u00a0disk storage). This results in\u00a0lower requirements for their computer processing. The best results were achieved by\u00a0the\u00a0TTR_1 model, which also outperformed the\u00a0IDW_1 model. In\u00a0this case, the\u00a0method used to\u00a0create the\u00a0given terrain probably plays a\u00a0role, especially the\u00a0very principle of\u00a0the\u00a0IDW technique.<\/p>\n

The RMSE values copy the\u00a0MAE values to\u00a0some extent. This is\u00a0due to\u00a0the\u00a0fact that RMSE is\u00a0based on\u00a0MAE and is\u00a0modified to\u00a0reflect more the\u00a0occurrence of\u00a0extreme deviations\u00a0[16]. In\u00a0our case, it\u00a0can therefore be\u00a0stated that none of\u00a0the\u00a0tested DMTs carries extreme error values when compared.<\/p>\n

A\u00a0comparison of\u00a0the\u00a0physical size of\u00a0individual rasters (i.e. the\u00a0size they occupy on\u00a0disk) shows how changing their spatial resolution (e.g. from 1\u00a0m to 5\u00a0m) dramatically reduces their size on\u00a0disk. The exception is\u00a0the\u00a0IDW_1 and TTR_1 models at\u00a0Loc_3. In\u00a0this case, the\u00a0size of\u00a0the\u00a0raster model exceeds the\u00a0size of\u00a0the\u00a0TIN model, which may be\u00a0caused by\u00a0the\u00a0flat nature of\u00a0Loc_3. In\u00a0the\u00a0case of\u00a0flat sites, DMR 5G provides a\u00a0lower density of\u00a0points than in\u00a0sloping sites\u00a0[10]. Lower point density reduces the\u00a0size of\u00a0the\u00a0TIN model.<\/p>\n

This article was created within the\u00a0TA\u00a0CR\u00a0project TK04030223 and as\u00a0such follows its goals. One of\u00a0them is\u00a0to\u00a0create 3D lines of\u00a0fourth order watercourses for the\u00a0Czech Republic. For this purpose, it\u00a0is\u00a0necessary to\u00a0use the\u00a0available datasets covering the\u00a0entire Czech Republic, process and evaluate them appropriately. Due to\u00a0the\u00a0scope of\u00a0processing and evaluation, machine data processing is\u00a0then necessary. It\u00a0is\u00a0also necessary to\u00a0take into account the\u00a0physical size of\u00a0the\u00a0produced DMTs due to\u00a0their subsequent storage. Thus, this article is\u00a0primarily intended to\u00a0help answer the\u00a0questions of\u00a0which available data sets are the\u00a0most suitable for the\u00a0needs of\u00a0the\u00a0project and what spatial resolution of\u00a0the\u00a0rasters produced by\u00a0DMT will be\u00a0appropriate, especially with regard to\u00a0their accuracy and storability.<\/p>\n

CONCLUSION<\/h2>\n

The results of\u00a0comparing the\u00a0height schematization quality of\u00a0the\u00a0watercourse line, produced by\u00a0different DMTs, show that models based on\u00a0DMR 4G data achieve worse results than models with the\u00a0same resolution based on\u00a0DMR 5G data. When comparing models with the\u00a0identical spatial resolution, based on\u00a0DMR 5G data and created with a\u00a0different interpolation method, it\u00a0is\u00a0evident that the\u00a0choice of\u00a0method for creating DMT plays a\u00a0role, especially for rasters with a\u00a0higher resolution. As\u00a0resolution decreases, the\u00a0importance of\u00a0the\u00a0interpolation method influence declines. The best MAE values were achieved by\u00a0the\u00a0TTR_1 model, with MAE of\u00a00.02\u00a0m. The worst results were equally achieved by\u00a0the\u00a0TTR_10 and IDW_10 models, with MAE of\u00a00.36\u00a0m. The\u00a0RMSE values are only slightly different from the\u00a0MAE values. It\u00a0can therefore be\u00a0assumed that none of\u00a0the\u00a0DMTs contain extreme values of\u00a0residual errors.<\/p>\n

Comparing the\u00a0physical size of\u00a0DMTs on\u00a0disk shows how the\u00a0size of\u00a0raster DMTs increases with their resolution. The rasters with 1\u00a0m resolution reach 50\u201370\u00a0%, with 5\u00a0m 1.9\u20135.2\u00a0%, and with 10\u00a0m 0.5\u20131.3\u00a0% of\u00a0the\u00a0size of\u00a0the\u00a0corresponding TIN DMT. However, for rasters with a\u00a0resolution of\u00a01\u00a0m, this reduction does not always apply \u2013 it\u00a0especially applies to\u00a0flat basins, where the\u00a0point density of\u00a0DMR 5G data is\u00a0low.<\/p>\n

Acknowledgements<\/h3>\n

This article was created with the\u00a0support of\u00a0the\u00a0Technology Agency of\u00a0the\u00a0Czech Republic as\u00a0part of\u00a0the\u00a0project No. TK04030223 \u201cDetermination of\u00a0hydropower potential of\u00a0Pico-Hydropower in\u00a0current and predicted climatic conditions of\u00a0the\u00a0Czech Republic\u201d.<\/span><\/span><\/em><\/p>\n

The Czech version of this article was peer-reviewed, the English version was translated from\u00a0the Czech original by Environmental Translation Ltd.<\/p>\n","protected":false},"excerpt":{"rendered":"

Determining the\u00a0gradient of\u00a0watercourses in\u00a0the\u00a0case of\u00a0local applications is\u00a0a\u00a0common problem, which is\u00a0most often dealt with by\u00a0geodetic surveying. However, determining the\u00a0gradient of\u00a0all watercourses in\u00a0the\u00a0Czech Republic is\u00a0a\u00a0challenge. The use of\u00a0geodetic methods on\u00a0such a\u00a0scale is\u00a0usually unrealistic. Therefore, it\u00a0is\u00a0necessary to\u00a0choose a\u00a0different approach, such as\u00a0the\u00a0extraction of\u00a0the\u00a0gradient lines from other already existing elevation data. The\u00a0DMR\u00a04G and DMR 5G are elevation models currently available for the\u00a0Czech Republic. For\u00a0the\u00a0extraction of\u00a0gradient lines, it\u00a0is\u00a0necessary to\u00a0create a\u00a0digital terrain model (DTM) from the\u00a0available datasets. Various interpolation methods are used for this. But which of\u00a0the\u00a0available interpolation methods is\u00a0the\u00a0most suitable? What role does the\u00a0size of\u00a0spatial resolution play in\u00a0the\u00a0quality of\u00a0altimetry representation and subsequent sizes of\u00a0the\u00a0stored DTMs? To\u00a0find answers to\u00a0these questions, we\u00a0chose four study sites (fourth order catchments) in\u00a0the\u00a0Otava river basin. Eight different DTMs were then created at\u00a0each site, which were then compared. The results show that choice of\u00a0raster size has a\u00a0significantly greater influence on\u00a0the\u00a0resulting quality of\u00a0the\u00a0gradient lines than the\u00a0choice of\u00a0interpolation method in\u00a0the\u00a0case of\u00a0DTM creation from DMR 5G data. DTM from DMR 4G data gives worse results than from DMR 5G at\u00a0the\u00a0same raster resolution.<\/p>\n","protected":false},"author":8,"featured_media":28757,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[2,92],"tags":[3450,3451,3452,839],"coauthors":[1578,1414,3453,1410,1409,3454],"class_list":["post-29093","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-from-the-world-of-water-management","category-main","tag-dmt","tag-sklon","tag-spad","tag-vodni-tok"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/29093","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/comments?post=29093"}],"version-history":[{"count":10,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/29093\/revisions"}],"predecessor-version":[{"id":29094,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/29093\/revisions\/29094"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media\/28757"}],"wp:attachment":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media?parent=29093"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/categories?post=29093"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/tags?post=29093"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/coauthors?post=29093"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}