The\u00a0delimited catchments were further classified according to\u00a0a number of\u00a0characteristics related to\u00a0the\u00a0risk of\u00a0direct runoff. A\u00a0cluster analysis was performed in\u00a0order to\u00a0classify these catchments. The\u00a0catchment characteristics that influence the\u00a0hydrological response were included in\u00a0the\u00a0analysis. These are mainly rainfall data, hydro-morphological characteristics of\u00a0the\u00a0relevant basin, land use, and soil hydrological characteristics. One negative impact of\u00a0direct runoff is\u00a0erosion. Erosion monitoring can be indirectly used as\u00a0an indicator of\u00a0the\u00a0state of\u00a0a\u00a0specific area in\u00a0terms of\u00a0the\u00a0occurrence of\u00a0direct runoff (https:\/\/me.vumop.cz). As\u00a0part of\u00a0this initiative, which completed ten years of\u00a0operation in\u00a02022, erosion events are recorded. The\u00a0database contains more than two thousand records. However, the\u00a0records within the\u00a0Czech Republic are inconsistent, which is\u00a0due to\u00a0the\u00a0involvement of\u00a0branches of\u00a0the\u00a0State Land Office (St\u00e1tn\u00ed pozemkov\u00fd \u00fa\u0159ad, SP\u00da). However, it is\u00a0a\u00a0relatively extensive evidence of\u00a0erosion.<\/p>\n
INTRODUCTION<\/h2>\n
River catchments in\u00a0the\u00a0Czech Republic are divided into four levels by default. However, the\u00a0smallest of\u00a0them, the\u00a04th order basins, are quite different in\u00a0terms of\u00a0size \u2013 from basins with a\u00a0size exceeding 20 km2<\/sup> to\u00a0additional basins with an\u00a0area of\u00a0less than 1 km2<\/sup>. Fourth order basins were categorized in\u00a0terms of\u00a0their potential hydrological response according to\u00a0the\u00a0method described in\u00a0[1]. Categorization of\u00a0the\u00a04th order basins in\u00a0terms of\u00a0hydrological response is\u00a0influenced specifically by the\u00a0different size of\u00a0the\u00a0area. Another factor is\u00a0the\u00a0combination of\u00a0headwater (upstream) and flow catchments.<\/p>\n Upper \u2013 non-flow catchments form a\u00a0specific group of\u00a0catchments, sometimes they are called \u201cfirst order catchments\u201d [2], other times these catchments are referred to\u00a0as headwater catchments. These catchments form the\u00a0basis of\u00a0the\u00a0hydrographic network and are the\u00a0primary areas for capturing or reducing flood damage. Simultaneously, these upper basins provide often diverse ecosystem services to\u00a0the\u00a0areas below them [3]. They tend to\u00a0be very sensitive to\u00a0changes and are the\u00a0fastest developing parts of\u00a0the\u00a0landscape. For these reasons, planning and management within these areas is\u00a0a\u00a0complex task [4]. In\u00a0the\u00a0past, a\u00a0number of\u00a0authors dealt with the\u00a0similarities, characteristics and response of\u00a0basins from different points of\u00a0view. For example, [5] are motivated to\u00a0classify basins rather with regard to\u00a0long-term processes in\u00a0basins. on\u00a0a\u00a0similar principle, basin attributes are defined, classified and shared within CAMELS [6] and others. Data sets are created for basins that describe six main groups of\u00a0attributes \u2013 topography, climate, hydrological characteristics, land cover, and soil and geological data.<\/p>\n There are relatively few verification data of\u00a0the\u00a0hydrological response in\u00a0the\u00a0Czech Republic. CHMI operates less than twenty observation profiles on\u00a0small catchments of\u00a0up to\u00a010 km2<\/sup>. In\u00a0addition, the\u00a0creation of\u00a0most of\u00a0these basins was motivated by the\u00a0monitoring of\u00a0water in\u00a0the\u00a0basins of\u00a0water supply reservoirs, and thus they are mainly forest basins. Geochemical monitoring of\u00a014\u00a0small forest basins is\u00a0dealt with by the\u00a0GEOMON project [7]. It is\u00a0focused primarily on\u00a0the\u00a0material composition of\u00a0precipitation, soil and runoff, but it also records precipitation totals and flow values in\u00a0the\u00a0closing profiles. The\u00a0monitored basins of\u00a0GEOMON overlap in\u00a0some places with the\u00a0basins operated by CHMI. on\u00a0agricultural land, the\u00a0number of\u00a0monitored basins operated by a\u00a0professionally oriented organization is\u00a0incomparably smaller, and the\u00a0time series of\u00a0data are also significantly shorter.<\/p>\n Practically the\u00a0only tool for designing objects on\u00a0small watercourses and modifications in\u00a0the\u00a0catchment area are hydrological models. They are most often based on\u00a0the\u00a0empirically derived SCS-CN method [8], which is\u00a0constantly being developed and tested; from recent works, for example [9, 10]. The\u00a0sensitivity of\u00a0the\u00a0method to\u00a0available data for the\u00a0Czech Republic was dealt with by Strouhal [11, 12]. By default, design data is\u00a0provided according to\u00a0\u010cSN 75\u00a01400.2014. Hydrological data of\u00a0surface waters. In\u00a0the\u00a0lowest class, which includes small catchments, data are also derived using a\u00a0model based on\u00a0the\u00a0SCS-CN method. In\u00a0addition to\u00a0this regulation, TNV 75 2102 \u2013 Modification of\u00a0streams from 2010 states that modelling can be used for proposals for modification of\u00a0small watercourses in\u00a0catchments of\u00a0up to\u00a05 km2<\/sup>. A\u00a0boundary of\u00a05 km2<\/sup> was adopted for the\u00a0derivation of\u00a0the\u00a0upper basins that this article is\u00a0presenting. For a\u00a0more detailed description of\u00a0the\u00a0runoff response, it is\u00a0also possible to\u00a0use physically based models such as\u00a0SMODERP [13] and EROSION 3D [14].<\/p>\n A specific feature of\u00a0small catchments is\u00a0the\u00a0speed of\u00a0their hydrological response. The\u00a0speed of\u00a0response to\u00a0causative precipitation and the\u00a0associated risks are influenced by a\u00a0number of\u00a0parameters. The\u00a0biggest threat in\u00a0terms of\u00a0flows and associated risks in\u00a0these catchments is\u00a0torrential rainfall. Ka\u0161par [15] recently dealt with the\u00a0distribution of\u00a0precipitation in\u00a0the\u00a0Czech Republic.<\/p>\n The\u00a0most frequently used tool for describing rainfall are IDF (Intensity-Duration-Frequency) curves, which describe the\u00a0relationship between rain intensity, its length and recurrence time [16]. on\u00a0a\u00a0global scale, e.g. Courty [17] deals with the\u00a0distribution of\u00a0the\u00a0above mentioned statistical attributes of\u00a0rain. In\u00a0addition to\u00a0the\u00a0intensity of\u00a0the\u00a0precipitation event, its shape also significantly affects the\u00a0hydrological response at\u00a0the\u00a0local scale of\u00a0small catchments [18, 19].<\/p>\n Other important parameters that influence the\u00a0runoff response of\u00a0individual basins include properties of\u00a0soils, soil cover, and morphological characteristics. The\u00a0properties of\u00a0the\u00a0hydrographic network, described by the\u00a0number of\u00a0coefficients, also play a\u00a0role. The\u00a0nature of\u00a0the\u00a0terrain \u2013 morphology \u2013 primarily affects the\u00a0shape of\u00a0the\u00a0runoff wave, and thus the\u00a0overall response of\u00a0the\u00a0basin to\u00a0increased runoff, including soil erosion. There are several parameters for describing the\u00a0morphology of\u00a0the\u00a0basin; the\u00a0most frequently reported values are the\u00a0average slope, the\u00a0length of\u00a0the\u00a0slope, and the\u00a0topographic index [20].<\/p>\n One of\u00a0the\u00a0negative impacts of\u00a0the\u00a0surface component of\u00a0direct runoff is\u00a0erosion. on\u00a0the\u00a0scale of\u00a0source areas from 0.3 km2<\/sup> to\u00a010 km2<\/sup>, the\u00a0so-called critical points are determined, which are defined as\u00a0points of\u00a0entry of\u00a0concentrated runoff paths into the\u00a0urban areas [21]. The\u00a0critical points are determined on\u00a0the\u00a0DMR derived from the\u00a0ZABAGED contour model and the\u00a0risk rate is\u00a0determined based on\u00a0the\u00a0ratio of\u00a0arable land, average slope and using CORINE Land Cover. A\u00a0different approach to\u00a0the\u00a0threat not only to\u00a0the\u00a0urban areas, but also to\u00a0other elements of\u00a0critical infrastructure was assessed within the\u00a0project VG20122015092 \u2013 \u201cErosion \u2013 increased risk of\u00a0endangering the\u00a0population and water quality in\u00a0connection with expected climate change\u201d implemented in\u00a02012 and 2015. The\u00a0resulting map of\u00a0points is\u00a0available at\u00a0https:\/\/heis.vuv.cz. In\u00a0both cases, it is\u00a0a\u00a0certain view of\u00a0the\u00a0riskiness of\u00a0the\u00a0points, which is\u00a0based on\u00a0the\u00a0characteristics of\u00a0the\u00a0contributing small headwater catchments, however, these are still model situations. Another perspective can be the\u00a0recorded occurrence of\u00a0an erosion event, which is\u00a0part of\u00a0erosion monitoring [22] and the\u00a0map portal (https:\/\/me.vumop.cz\/).<\/p>\n As part of this initiative, which completed ten years of operation in 2022, erosion events are recorded. The database contains over two thousand records. Although the records within the Czech Republic are spatially uneven, which is due to the involvement of branches of the State Land Office, it is nevertheless a relatively extensive record of erosion manifestations.<\/p>\n Small headwater catchments (SHC) [23] are so-called non-flow catchments that have no tributaries, and thus correspond to\u00a0the\u00a0definition of\u00a0\u201cfirst order catchments\u201d [2]. This property was used in\u00a0deriving their borders in\u00a0the\u00a0whole of\u00a0the\u00a0Czech Republic based on\u00a0DMR 4G data at\u00a0a\u00a0resolution of\u00a05 \u00d7 5 m [24], water courses and water reservoirs. The\u00a0parameters that influence the\u00a0hydrological response were subsequently determined for the\u00a0areas of\u00a0the\u00a0basin defined in\u00a0this way, especially with regard to\u00a0the\u00a0possible risk of\u00a0runoff from short-term extreme rainfall.<\/p>\n SHC according to\u00a0[23] are not only catchments with a\u00a0size of\u00a0only 5 km2<\/sup>, but also all smaller catchments. This means, for example, that two catchments with a\u00a0size of\u00a03 km2<\/sup> after the\u00a0confluence already exceed 5 km2<\/sup>, but separately they are two catchments that fall within the\u00a0SHC. to\u00a0define the\u00a0SHC, six size categories listed in\u00a0Tab. 1 were chosen. Basins have been derived for all these classes, which allows their further mutual comparison.<\/p>\n Areas smaller than category 005 can be considered elementary runoff areas and are not evaluated as\u00a0separate catchments. At\u00a0the\u00a0same time, the\u00a0lower limit of\u00a00.3 km2<\/sup> corresponds to\u00a0the\u00a0lower limit of\u00a0the\u00a0derivation of\u00a0critical points [21].<\/p>\n To define the\u00a0SHC, three data sources were used \u2013 a\u00a0digital model of\u00a0the\u00a0terrain, watercourse axes, and water reservoir axes. The\u00a0main input for determining the\u00a0SHC was the\u00a0DMR 4G with a\u00a0resolution of\u00a05 \u00d7 5 m. Since in\u00a0some places the\u00a0watercourse axes, due to\u00a0human intervention and changes in\u00a0the\u00a0landscape, do not correspond to\u00a0the\u00a0runoff lines generated on\u00a0the\u00a0terrain model itself, the\u00a0current watercourse axes that are part of\u00a0ZABAGED\u00ae are included in\u00a0the\u00a0solution. These are based on\u00a0measurements of\u00a0detailed scales and reflect the\u00a0current state of\u00a0the\u00a0water network. When creating the\u00a0SHC, these lines are taken as\u00a0more accurate than DMR-based runoff routing. When deriving basins, these watercourse axes must be included in\u00a0the\u00a0solution. Watercourse axes were projected into the\u00a0DMR. The\u00a0value of\u00a0the\u00a0pixels of\u00a0the\u00a0terrain model through which the\u00a0axis of\u00a0the\u00a0watercourse passes has been reduced so that the\u00a0resulting direction of\u00a0the\u00a0runoff corresponds to\u00a0the\u00a0axes of\u00a0the\u00a0current watercourses. In\u00a0the\u00a0following step, any non-runoff areas were removed on\u00a0the\u00a0terrain model modified in\u00a0this way and a\u00a0runoff routing layer was created. A\u00a0one-way runoff routing tool (D8) was used to\u00a0route the\u00a0runoff. Accumulation in\u00a0each pixel was then derived from the\u00a0runoff directions.<\/p>\n For each catchment category (see Tab. 1), the\u00a0accumulation layer was reclassified so that the\u00a0values of\u00a0the\u00a0accumulation area outside the\u00a0group boundaries have the\u00a0NoData value and the\u00a0values of\u00a0the\u00a0accumulation area corresponding to\u00a0the\u00a0given category the\u00a0value 1. In\u00a0cases where the\u00a0runoff lines classified in\u00a0this way end or intersect with water reservoirs, the\u00a0runoff lines were shortened to\u00a0the\u00a0point where the\u00a0drain line crosses the\u00a0water reservoir. In\u00a0these cases, therefore, basins at\u00a0the\u00a0entrance to\u00a0water reservoirs are considered. For\u00a0modified lines in\u00a0individual categories, the\u00a0endpoints of\u00a0these lines were determined, which form the\u00a0closing profile of\u00a0the\u00a0basin. on\u00a0the\u00a0basis of\u00a0the\u00a0runoff direction derived above, the\u00a0boundary of\u00a0the\u00a0basin was derived for these points, taking into account the\u00a0axes of\u00a0the\u00a0watercourses.<\/p>\n The\u00a0hydrological response from the\u00a0SHC is\u00a0determined by its morphological characteristics, soil properties, land use, and causative precipitation. It can be assumed that the\u00a0hydrological response of\u00a0similar basins will be similar. Therefore, for the\u00a0above SHC categories, parameters were derived for their classification in\u00a0terms of\u00a0possible hydrological response.<\/p>\n The\u00a0morphological characteristics were determined based on\u00a0the\u00a0model of\u00a0the\u00a0terrain and watercourses. In\u00a0particular, these are the\u00a0characteristics of\u00a0altitude, slopes and length of\u00a0runoff paths, as\u00a0well as\u00a0several shape coefficients.<\/p>\n Average width of\u00a0the\u00a0basin<\/p>\n where\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 A\u00a0\u00a0\u00a0 is\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 the\u00a0area [m2<\/sup>]<\/p>\n L\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 maximum length of\u00a0the\u00a0runoff path [m]<\/p>\n <\/p>\n <\/p>\n Basin shape<\/p>\n <\/p>\n where\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 A\u00a0\u00a0\u00a0 is\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 the\u00a0catchment area [m2<\/sup>]<\/p>\n L\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 maximum length of\u00a0the\u00a0runoff path [m]<\/p>\n <\/p>\n Shape coefficient according to\u00a0Gravelius [25]<\/p>\n where\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 O\u00a0\u00a0\u00a0 is\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 the\u00a0circumference [m]<\/p>\n A\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 area [m2<\/sup>]<\/p>\n <\/p>\n All three shape coefficients describe the\u00a0shape of\u00a0the\u00a0basin. In\u00a0the\u00a0case of\u00a0Gravelius coefficient, it is\u00a0a\u00a0comparison of\u00a0the\u00a0shape of\u00a0the\u00a0basin to\u00a0a circle. The\u00a0average width and shape coefficient of\u00a0the\u00a0basin determines the\u00a0extent to\u00a0which the\u00a0shape of\u00a0the\u00a0basin departs from a\u00a0square, or its power.<\/p>\n The\u00a0standard description is\u00a0the\u00a0parameter of\u00a0the\u00a0water network density. This parameter determines the\u00a0ratio of\u00a0the\u00a0total length of\u00a0watercourses to\u00a0the\u00a0catchment area.<\/p>\n where\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 LT<\/sub>\u00a0\u00a0 is\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 the\u00a0length of\u00a0the\u00a0watercourse [m]<\/p>\n A\u00a0\u00a0\u00a0 \u00a0\u00a0\u00a0\u00a0\u00a0 area [m2<\/sup>]<\/p>\n <\/p>\n One of\u00a0the\u00a0parameters that are influenced by the\u00a0morphology and affect the\u00a0course of\u00a0the\u00a0runoff is\u00a0the\u00a0time lag (Tlag). The\u00a0Tlag value is\u00a0used to\u00a0describe the\u00a0unit hydrograph according to\u00a0the\u00a0SCS-CN method [26]. Tlag is\u00a0then calculated using [27].<\/p>\n where\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 Tlag<\/sub> is\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 the\u00a0time lag [hours]<\/p>\n L\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 length of\u00a0the\u00a0longest runoff path [foot]<\/p>\n Y\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 average slope of\u00a0the\u00a0basin [%]<\/p>\n S\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 maximum potential retention [inch]<\/p>\n The\u00a0direct runoff volume potential of\u00a0a given basin can be described by the\u00a0average value of\u00a0CN. CN integrates information about surface properties and soil infiltration properties. In\u00a0the\u00a0example given here, the\u00a0CN values are taken from the\u00a0derivation within the\u00a0Strategy of\u00a0protection against the\u00a0negative effects of\u00a0floods and erosion phenomena by semi-natural measures in\u00a0the\u00a0Czech Republic [28].<\/p>\n The\u00a0last group of\u00a0parameters is\u00a0precipitation data. Since short-term precipitation is\u00a0the\u00a0dominant source of\u00a0runoff in\u00a0small headwater catchments, six-hour design precipitation derived from rain radars with a\u00a0spatial resolution of\u00a01 \u00d7 1 km were selected [19, 15]. These data are available at\u00a0rain.fsv.cvut.cz.<\/p>\n Tab. 2 includes an overview of\u00a0monitored parameters. These are values describing the\u00a0mean value, variance, and minimum or maximum value, depending on\u00a0the\u00a0type of\u00a0parameter.<\/p>\n In total, there are 28 parameters that were subsequently tested in\u00a0all size categories in\u00a0terms of\u00a0mutual dependence using regression analysis. The\u00a0aim was to\u00a0obtain a\u00a0set of\u00a0independent parameters and classify the\u00a0basins into groups according to\u00a0their similarity using cluster analysis.<\/p>\n Delineation of\u00a0small catchments, assignment and calculation of\u00a0characteristics from the\u00a0DMR, and CSC-CN basins was processed in\u00a0the\u00a0ESRI environments (ArcGIS and ArcGIS Pro), subsequent statistical analyses were processed in\u00a0the\u00a0R\u00a0environment. Descriptive statistics and regression analysis tools were used for the\u00a0solution. Cluster analysis was performed using the\u00a0K-mean method. The\u00a0individual clusters were subsequently aggregated in\u00a0terms of\u00a0the\u00a0relative riskiness of\u00a0key parameters to\u00a0create the\u00a0direct runoff component into five risk classes. The\u00a0verification was carried out using recorded erosion events. The\u00a0aim was to\u00a0monitor whether the\u00a0classification of\u00a0the\u00a0basin in\u00a0terms of\u00a0risk coincides with the\u00a0location of\u00a0erosion events.<\/p>\n Basic data on\u00a0SHC derived according to\u00a0the\u00a0methodology described above are shown in\u00a0Tab. 3. As\u00a0the\u00a0SHC categories are always derived separately, the\u00a0resulting catchments overlap between the\u00a0categories \u2013 a\u00a0smaller catchment may be part of\u00a0a larger one in\u00a0the\u00a0parent categories. Therefore, in\u00a0addition to\u00a0the\u00a0described categories, a\u00a0group of\u00a0basins was created in\u00a0which only the\u00a0largest basins are preserved. Interconnected catchments have been eliminated. In\u00a0this way, catchments smaller than 5 km2<\/sup> in\u00a0the\u00a0monitored area of\u00a0the\u00a0Czech Republic are preserved. This group of\u00a0catchments is\u00a0referred to\u00a0as \u201cSet of\u00a0Largest Catchments\u201d \u2013 \u201cSoLC\u201d and is\u00a0also listed in\u00a0Tab. 3. For clarity, it is\u00a0added how much representation individual size categories have in\u00a0the\u00a0resulting SoLC group; the\u00a0table contains data on\u00a0the\u00a0number of\u00a0elements of\u00a0the\u00a0given category that are part of\u00a0it. The\u00a0representation of\u00a0the\u00a0areas of\u00a0individual catchment categories in\u00a0the\u00a0SoLC group is\u00a0shown in\u00a0Fig.\u00a01.<\/p>\n Tab. 3. Number and the total area of catchments in each category. For individual categories (1st column) the number of elements (2nd column) and the total area of the given class (3rd column) are given. The 4th and 5th columns show the representation of the elements of the given class in the SoLC class and the\u00a0per-centage expression For individual basins in\u00a0all size categories, parameters were derived according to\u00a0Tab. 2. For the\u00a0needs of\u00a0cluster analysis, representative and independent parameters are sought in\u00a0the\u00a0first step. Dependent parameters must be discarded. The\u00a0search for the\u00a0degree of\u00a0agreement between monitored parameters was carried out both for individual categories (including SoLC) and for all basins together. From the\u00a0point of\u00a0view of\u00a0the\u00a0groups of\u00a0dependent parameters, the\u00a0individual categories do not differ from each other. It is\u00a0therefore the\u00a0fact that the\u00a0links between monitored parameters are similar for all size categories. A\u00a0visually adjusted parameter match calculated using Pearson\u2019s correlation coefficient is\u00a0in\u00a0Fig.\u00a02.<\/p>\n Five parameters were chosen from groups of\u00a0elements grouped according to\u00a0their mutual dependence, which can be considered independent and sufficiently representative. Appropriate representative parameters were selected using principal component analysis (PCA), namely:<\/p>\n The\u00a0stream network density (SND) and shape coefficient alpha (\u03b1) parameters are jointly correlated with the\u00a0surface runoff path standard deviation (FSSTD) parameter. The\u00a0SND is\u00a0also related to\u00a0the\u00a0slope characteristics and the\u00a0parameter \u03b1 is\u00a0related to\u00a0the\u00a0altitude. At\u00a0the\u00a0same time, SND directly describes the\u00a0characteristics of\u00a0the\u00a0watercourse network. For this reason, these two parameters were used.<\/p>\n To classify basins into groups in\u00a0terms of\u00a0potential response, it is\u00a0important to\u00a0compare the\u00a0distribution of\u00a0classification parameters between individual basin categories. If\u00a0the\u00a0chosen classification parameters had a\u00a0different distribution for individual groups of\u00a0basins, it would mean that different size categories have a\u00a0different character of\u00a0the\u00a0hydrological response to\u00a0precipitation. The\u00a0aim was to\u00a0compare the\u00a0differences between individual size categories.\u00a0 Article [29] deals with this issue in\u00a0more detail.<\/p>\n Since the\u00a0parameter distribution differences between categories are not significant and do not differ from SoLC, cluster analysis was performed only on\u00a0the\u00a0SoLC group, in\u00a0which all size categories are represented by at\u00a0least 20 %. Cluster analysis using the\u00a0K mean method was performed in\u00a0the\u00a0R environment, in\u00a0the\u00a0range of\u00a0clusters from two to\u00a0eight with a\u00a0setting of\u00a025 initial training points. Each catchment in\u00a0the\u00a0SoLC was assigned to\u00a0a group according to\u00a0five selected parameters each time the\u00a0clusters were created. The\u00a0formation of\u00a0individual groups of\u00a0basins is\u00a0described in\u00a0Fig.\u00a03. The\u00a0groups are marked with letters. If\u00a0a\u00a0group is\u00a0formed only by separating from a\u00a0previously formed group, a\u00a0numerical designation is\u00a0added.<\/p>\n The\u00a0groups formed during the\u00a0gradual creation of\u00a0clusters can be characterized by the\u00a0following description. The\u00a0geographical clustering is\u00a0then shown in\u00a0Fig.\u00a04.<\/p>\n The\u00a0gradually formed groups of\u00a0basins are characterized by the\u00a0mean values of\u00a0the\u00a0selected five parameters mentioned above. Parameters and cluster analysis are discussed in\u00a0more detail in\u00a0article [29].<\/p>\n The\u00a0created basin clusters are further classified according to\u00a0the\u00a0possible risk of\u00a0direct runoff. From the\u00a0point of\u00a0view of\u00a0the\u00a0influence of\u00a0the\u00a0parameters on\u00a0the\u00a0risk associated with the\u00a0emergence of\u00a0direct runoff, the\u00a0following applies for individual parameters:<\/p>\n SND \u2013 The\u00a0higher the\u00a0value, the\u00a0denser the\u00a0permanent water network, any runoff will tend to\u00a0concentrate in\u00a0these paths where the\u00a0runoff is\u00a0expected. A\u00a0larger value therefore means a\u00a0lower level of\u00a0risk.<\/p>\n Tlag<\/sub> \u2013 The\u00a0longer the\u00a0time lag, the\u00a0lower peak flows can be expected.<\/p>\n \u03b1 \u2013 The\u00a0more complex the\u00a0shape of\u00a0the\u00a0catchment area, the\u00a0more the\u00a0runoff paths are lengthened, and thus also the\u00a0culmination is\u00a0reduced.<\/p>\n CNM<\/sub> \u2013 The\u00a0smaller the\u00a0mean CN value, the\u00a0greater the\u00a0retention rate in\u00a0the\u00a0basin and the\u00a0lower the\u00a0potential risk of\u00a0threat.<\/p>\n P20<\/sub> \u2013 The\u00a0higher the\u00a0rainfall, the\u00a0higher the\u00a0risk of\u00a0possible runoff response.<\/p>\n For individual parameters, a\u00a0mean value was calculated in\u00a0the\u00a0SoLC category, which is\u00a0taken as\u00a0medium risk. The\u00a0degree of\u00a0risk was determined for individual parameters relative to\u00a0this mean value of\u00a0the\u00a0given parameter. For each value of\u00a0the\u00a0parameter corresponding to\u00a0the\u00a0centre of\u00a0gravity of\u00a0the\u00a0individual clusters, the\u00a0proportion with this mean value was determined, thereby determining the\u00a0riskiness of\u00a0each parameter in\u00a0the\u00a0given cluster. Those combinations of\u00a0five parameters are considered at\u00a0risk where a\u00a0negative assessment prevails, and vice versa. The\u00a0level of\u00a0overall risk is\u00a0divided into five categories from low to\u00a0high risk as\u00a0described below:<\/p>\n Low risk \u2013 the\u00a0combination of\u00a0possible runoff response parameters assumes a\u00a0low risk in\u00a0terms of\u00a0direct runoff affecting the\u00a0basin. These areas appear to\u00a0be unproblematic from a\u00a0direct response point of\u00a0view and the\u00a0need The\u00a0parameter values for determining the\u00a0risk level are shown in\u00a0Tab. 4.<\/p>\n The\u00a0classification of\u00a0the\u00a0groups of\u00a0catchment areas of\u00a0the\u00a0cluster analysis according to\u00a0the\u00a0level of\u00a0risk is\u00a0shown in\u00a0Tab. 5, where groups from the\u00a0number of\u00a0clusters 2\u20138 are included.<\/p>\n The\u00a0geographical expression of\u00a0the\u00a0level of\u00a0risk is\u00a0then shown in\u00a0Fig.\u00a05. Groups A2 and C together form a\u00a0group with a\u00a0high risk, groups A11, B1, B3 a\u00a0group with a\u00a0medium risk, and A12 and D a\u00a0group with a\u00a0lower risk.<\/p>\n The\u00a0classification of\u00a0small catchments in\u00a0terms of\u00a0the\u00a0risk of\u00a0direct runoff is\u00a0expressed relatively between individual parameters. A\u00a0certain validation criterion of\u00a0the\u00a0results can be a\u00a0comparison of\u00a0the\u00a0classification according to\u00a0the\u00a0degree of\u00a0risk with recorded erosion events in\u00a0the\u00a0Monitoring of\u00a0agricultural soil erosion. Monitoring has been ongoing since 2012, and by the\u00a0end of\u00a02021, over 2,200 erosion events have already been recorded [30].<\/p>\n The\u00a0intersection of\u00a0the\u00a0affected land listed in\u00a0the\u00a0monitoring with the\u00a0boundaries of\u00a0the\u00a0defined small catchments is\u00a0shown in\u00a0Fig.\u00a06. to\u00a0assign the\u00a0event to\u00a0the\u00a0relevant basin, the\u00a0centre of\u00a0gravity of\u00a0the\u00a0polygon delimiting the\u00a0recorded event was taken.<\/p>\n Of the\u00a0total number of\u00a02,220 recorded events until 2021, half of\u00a0them were in\u00a0high-risk catchments. Most of\u00a0the\u00a0recorded erosion events are recorded in\u00a0Vyso\u010dina Region and South Moravia. In\u00a0other regions, where erosion events are not recorded, it is\u00a0more about the\u00a0completeness of\u00a0the\u00a0database of\u00a0erosion events than about parts of\u00a0the\u00a0Czech Republic without occurrence of\u00a0events.<\/p>\n Basin classification is\u00a0more commonly used in\u00a0experimental hydrology. Basins are also classified in\u00a0the\u00a0expanding CAMELS database. In\u00a0these cases, the\u00a0list of\u00a0parameters is\u00a0larger. In\u00a0contrast to\u00a0the\u00a0selection of\u00a0parameters presented here, it includes hydrological data of\u00a0long-term balances and parameters that affect long-term runoff and other components of\u00a0the\u00a0balance [6]. In\u00a0most cases, it monitors larger basins. Long-term series of\u00a0observations in\u00a0small basins are significantly less frequent than in\u00a0larger basins. The\u00a0small catchments presented here bring information about the\u00a0upper non-flow basins to\u00a0the\u00a0standard classification. At\u00a0the\u00a0same time, these upper catchments are classified according to\u00a0key characteristics affecting the\u00a0direct component of\u00a0runoff.<\/p>\n Input data with different spatial resolutions were used to\u00a0create basin boundaries and their properties. The\u00a0delineation of\u00a0the\u00a0basin boundaries was created on\u00a0the\u00a0basis of\u00a0a terrain model with a\u00a0resolution of\u00a05 \u00d7 5 m, which is\u00a0sufficiently detailed even for the\u00a0delineation of\u00a0small catchments in\u00a0the\u00a0considered size category 005. Based on\u00a0the\u00a0terrain model, other morphological characteristics were then derived at\u00a0the\u00a0same resolution. When using the\u00a0D8 method at\u00a0a\u00a0lower resolution, the\u00a0creation of\u00a0basin boundaries could be affected, especially for the\u00a0smallest category.<\/p>\n The\u00a0parameters that enter the\u00a0cluster analysis do not differ significantly in\u00a0terms of\u00a0the\u00a0distribution of\u00a0values between the\u00a0categories. Smaller basins are also part of\u00a0larger basins and together they form SoLC, where at\u00a0least 20 % of\u00a0the\u00a0number of\u00a0basins from each category is\u00a0represented. The\u00a0total area of\u00a0the\u00a0upper basins included in\u00a0the\u00a0SoLC is\u00a063,000 km2<\/sup>, which is\u00a0about 80 % of\u00a0the\u00a0area of\u00a0the\u00a0Czech Republic (78,000 km2<\/sup>).<\/p>\n Several of\u00a0the\u00a028 parameters considered are mutually correlated. The\u00a0first group (A, see Fig.\u00a02) of\u00a0mutually correlated parameters are the\u00a0geometric parameters of\u00a0the\u00a0basin (size, area, runoff path length, runoff accumulation) with Tlag. The\u00a0shape coefficients (group B) are linked in\u00a0a mutual correlation with the\u00a0SND and the\u00a0length of\u00a0the\u00a0runoff paths outside the\u00a0watercourse. From this group, the\u00a0parameters SND and \u03b1 are the\u00a0least interconnected. Another important group (C) is\u00a0the\u00a0intercorrelated parameters describing the\u00a0slope ratios of\u00a0the\u00a0basin, the\u00a0slopes of\u00a0the\u00a0watercourses, the\u00a0altitude in\u00a0relation to\u00a0the\u00a0land use and soil characteristics of\u00a0the\u00a0CNM. This connection corresponds to\u00a0the\u00a0use of\u00a0land in\u00a0mountainous, mostly steeper sloping areas, which are mainly forested. A\u00a0separate group of\u00a0parameters is\u00a0precipitation (D), which have a\u00a0mutually strong link. They do not show a\u00a0significant link with the\u00a0other parameters.<\/p>\n From the\u00a0point of\u00a0view of\u00a0response and, possibly, from the\u00a0point of\u00a0view of\u00a0the\u00a0risk of\u00a0increased flows, mainly short-term rains are key in\u00a0small headwater catchments. The\u00a0occurrence of\u00a0a flood and possible threat is\u00a0a\u00a0combination of\u00a0the\u00a0current conditions of\u00a0the\u00a0basin and the\u00a0course of\u00a0the\u00a0causative precipitation. Especially short-term torrential rains are difficult to\u00a0predict. However, it\u00a0is\u00a0a\u00a0fact that two differently classified basins, which will have the\u00a0same initial state and will be loaded with the\u00a0same rainfall, will have a\u00a0different response to\u00a0causative rainfall. The\u00a0classification of\u00a0basins according to\u00a0parameters has a\u00a0practical impact on\u00a0possible prioritization in\u00a0terms of\u00a0the\u00a0implementation of\u00a0measures.<\/p>\n The\u00a0subsequent cluster analysis of\u00a0the\u00a0basins from the\u00a0point of\u00a0view of\u00a0their hydrological response shows that, according to\u00a0the\u00a0selected parameters, there is\u00a0a\u00a0basic division of\u00a0the\u00a0basins into two groups, in\u00a0which the\u00a0categories A2 and B2 are gradually separated, which according to\u00a0their parameters fall into the\u00a0group with the\u00a0risk of\u00a0increased runoff from torrential rains. Above all, risk group C is\u00a0then separated from these two groups. The\u00a0independently created group D is\u00a0created from the\u00a0previously created groups a\u00a0and B, and the\u00a0basins with the\u00a0lowest risk in\u00a0terms of\u00a0threat are separated within it. The\u00a0creation of\u00a0two clusters C1 and C2 from group C and partly from group A2, with a\u00a0total number of\u00a0clusters of\u00a0eight, no longer brings new information in\u00a0terms of\u00a0possible threat. For the\u00a0classification of\u00a0SHCs in\u00a0terms of\u00a0their potential threat, it\u00a0is\u00a0therefore appropriate to\u00a0classify them into seven clusters.<\/p>\n Clusters of\u00a0small catchments were assigned a\u00a0risk level value on\u00a0a\u00a0five-point scale. When divided into seven clusters, the\u00a0lowest risk is\u00a0in\u00a0group A12. Together with D, it falls into the\u00a0\u201creduced risk\u201d category, however, it is\u00a0on\u00a0the\u00a0borderline of\u00a0values for inclusion in\u00a0the\u00a0\u201clow\u201d category. Groups A11, B1 and B3 have a\u00a0medium risk. High risk A2 and C, where A2 is\u00a0the\u00a0highest risk of\u00a0all groups and C is\u00a0on\u00a0the\u00a0borderline for inclusion in\u00a0the\u00a0\u201cincreased risk\u201d group.<\/p>\n Some validation of\u00a0the\u00a0resulting risk can be done by comparing the\u00a0locations of\u00a0the\u00a0actual observed erosion events and the\u00a0boundaries of\u00a0the\u00a0resulting SHC. The\u00a0result shows that more than half of\u00a0the\u00a0recorded events are in\u00a0the\u00a0high risk class. Less than 15 % are in\u00a0the\u00a0medium and reduced risk classes, and 20\u00a0% of\u00a0the\u00a0recorded events are on\u00a0land outside the\u00a0SoLC, i.e. in\u00a0inter catchments.<\/p>\n The\u00a0presented derivation and subsequent classification of\u00a0SHC (small headwater catchments) from the\u00a0point of\u00a0view of\u00a0the\u00a0level of\u00a0threat bring insight into their possible hydrological response. It can be said that up to\u00a0the\u00a0number of\u00a0five clusters, the\u00a0primary division into two groups a\u00a0and B is\u00a0preserved, which are already created during the\u00a0creation of\u00a0the\u00a0first two clusters. In\u00a0both, two groups are gradually formed, which are rather risky. We can consider seven clusters a\u00a0sufficiently explanatory classification of\u00a0SHC, where both group D (a very low risk), consisting of\u00a0elements of\u00a0groups a\u00a0and B, and a\u00a0group C (very threatened) by delineation from groups A2 and B2, will be formed. With seven clusters from the\u00a0area of\u00a0the\u00a0Czech Republic, this approach assesses 28.5 % of\u00a0the\u00a0area as\u00a0at\u00a0risk, 29.4 % of\u00a0the\u00a0area with medium risk, 22 % with below average risk, and 20 % of\u00a0the\u00a0area of\u00a0the\u00a0Czech Republic is\u00a0not assessed \u2013 it does not fall into the\u00a0SHC category.<\/p>\n Headwater catchments cover a\u00a0significant part of\u00a0the\u00a0Czech Republic. With the\u00a0selected limit of\u00a0up to\u00a05 km2<\/sup>, the\u00a0SoLC (Set of\u00a0Largest Catchments) make up about 80\u00a0% of\u00a0the\u00a0Czech Republic. SHC are a\u00a0space for the\u00a0primary accumulation of\u00a0rainwater. At\u00a0the\u00a0same time, these basins are most affected by direct runoff, which subsequently reduces the\u00a0availability of\u00a0water in\u00a0their area. The\u00a0classification of\u00a0small headwater catchments in\u00a0terms of\u00a0potential threat from torrential rains is\u00a0one of\u00a0the\u00a0possible perspectives. Another use of\u00a0the\u00a0spatial delineation of\u00a0these basins can be subsequent classification, for example, from the\u00a0point of\u00a0view of\u00a0water availability for irrigation, or\u00a0the\u00a0application of\u00a0other adaptation measures with expected climate change.<\/p>\n Within the\u00a0Czech Republic, it is\u00a0possible to\u00a0consider the\u00a0agriculturally used parts of\u00a0South Moravia and the\u00a0western part of\u00a0the\u00a0Bohemian-Moravian Highlands and north-western Bohemia to\u00a0be more at\u00a0risk. Alternation of\u00a0high-risk and lower-risk basins is\u00a0also typical in\u00a0these areas. The\u00a0region of\u00a0South Moravia and Western Bohemia is\u00a0a\u00a0typical agricultural landscape. Areas with medium risk are mainly mountainous (\u0160umava, Krkono\u0161e, Jizersk\u00e9 hory, Jesen\u00edky), which are characterized by increased precipitation totals, the\u00a0impact of\u00a0which is\u00a0reduced by\u00a0increased afforestation. This group also includes the\u00a0Beskydy Mountains and northern Moravia. The\u00a0Ore Mountains fall within an area with a\u00a0lower risk, which is\u00a0due to\u00a0lower precipitation totals. Areas with a\u00a0lower risk include foothill areas, with the\u00a0exception of\u00a0Orlick\u00e9 hory foothills and south-western Pilsen Region, which fall within areas with an increased risk. The\u00a0largest area of\u00a0the\u00a0basin with a\u00a0reduced risk is\u00a0Polab\u00ed, partly T\u0159ebo\u0148 Region, and the\u00a0hilly areas of\u00a0Brdy and western Bohemia.<\/p>\nMETHODOLOGY FOR DERIVATION OF CATCHMENT BOUNDARIES AND THEIR CLASSIFICATION<\/h2>\n
Definition of\u00a0catchment boundaries<\/h3>\n
Tab.\u00a01. SHC categories (a\u00a0range of\u00a0area sizes was chosen for each category)
![\"\"](\"https:\/\/www.vtei.cz\/wp-content\/uploads\/2023\/02\/Kavka-tab-1-N.jpg\")
<\/a><\/h5>\nCharacteristics of\u00a0small catchments<\/h3>\n
<\/a>\n<\/a>\n<\/a>\n<\/a>\n<\/a>\nTab.\u00a02. List of\u00a0parameters that enter the\u00a0SHC cluster analysis. Parameters 1\u201316 were derived from DMR and vector lines of\u00a0water courses, catchment slope was derived using unconditioned DMR. Parameters 20\u201324 derived according to\u00a0Eq. (1\u20134)
<\/a><\/h5>\nRESULTS<\/h2>\n
SHC definition<\/h3>\n
<\/a><\/p>\n<\/a>Fig.\u00a01. Summary of\u00a0the\u00a0area of\u00a0the\u00a0catchments in\u00a0SoLC categories<\/h6>\nSelection of\u00a0parameters<\/h3>\n
<\/a>Fig.\u00a02. Visualization of\u00a0the\u00a0correlation between individual parameters for all catchments regardless of\u00a0size category. A\u00a0negative correlation is\u00a0shown in\u00a0red, and a\u00a0positive correlation is\u00a0shown in\u00a0blue. The\u00a0stronger the\u00a0bond between two parameters, the\u00a0darker and larger the\u00a0symbol. Similar parameters are grouped together to\u00a0visualize groups of\u00a0similar parameters A\u00a0to D [29]<\/h6>\n\n
\nsix-hour rainfall values due to\u00a0the\u00a0derivation of\u00a0this data.<\/li>\n
\nof the\u00a0runoff hydrograph, and thus the\u00a0size of\u00a0the\u00a0peak flow.<\/li>\n
\nof the\u00a0length of\u00a0all watercourses in\u00a0the\u00a0basin and the\u00a0area of\u00a0the\u00a0basin. Together with the\u00a0shape coefficient alpha (\u03b1), they include both the\u00a0characteristics of\u00a0the\u00a0length of\u00a0the\u00a0runoff paths and the\u00a0shape of\u00a0the\u00a0basin. These two parameters together represent group B.<\/li>\n<\/ul>\nDistribution of\u00a0parameters<\/h3>\n
<\/a>Fig.\u00a03. The\u00a0Sankey diagram shows the\u00a0evolution and regrouping of\u00a0SoLC classes with increasing number of\u00a0clusters. The\u00a0number of\u00a0elements in\u00a0a\u00a0given group corresponds to\u00a0a\u00a0belt. At\u00a0the\u00a0same time, the\u00a0diagram shows how individual watersheds are oversubscribed according to\u00a0the\u00a0number of\u00a0clusters. The\u00a0basic division is\u00a0already visible in\u00a0the\u00a0formation of\u00a0two clusters (A, B). From the\u00a0number of\u00a0clusters six to\u00a0the\u00a0development of\u00a0groups that are created by combining the\u00a0basic division into A, B and subgroups. At\u00a0six, group D is\u00a0formed, which is\u00a0a\u00a0combination of\u00a0all previously formed groups. With the\u00a0number of\u00a0clusters 7 and 8, groups C are created, which are a\u00a0combination of\u00a0parts of\u00a0groups A2 and B2 [29]<\/h6>\n\n
\nis characterized by a\u00a0higher CNM with a\u00a0lower volume of\u00a0precipitation P20.
\nGroup B is\u00a0characterized by higher precipitation P20 and a\u00a0larger CNM value (Fig.\u00a04a).<\/li>\n
\nof the\u00a0basin, in\u00a0terms of\u00a0stream network density (SND) and in\u00a0terms of\u00a0lag time (Tlag) (Fig.\u00a04b).<\/li>\n
\ninto groups C1 and C2. The\u00a0newly formed group C1 is\u00a0also made up of\u00a0a part of\u00a0the\u00a0catchment area of\u00a0group A2 and is\u00a0characterized, like the\u00a0original group C, by higher values of\u00a0P20 and CNM. It is\u00a0distinguished from group C2 by the\u00a0difference in\u00a0the\u00a0SND and (\u03b1) parameters; this division no longer brings new information to\u00a0the\u00a0basin classification.<\/li>\n<\/ul>\n<\/a>Fig.\u00a04. Geographical representation of\u00a0group evolution when forming clusters from two (a) to\u00a0seven (f)<\/h6>\n
\nfor measures in\u00a0these areas is\u00a0not anticipated.<\/p>\n\n
\nby\u00a0direct runoff. These areas are unproblematic from the\u00a0point of\u00a0view of\u00a0direct response and taking measures in\u00a0these areas is\u00a0not needed.<\/li>\nTab. 4. Individual parameters used to express the degree of risk in relation to the mean values of the parameters
<\/a><\/h5>\n<\/a>Fig.\u00a05. Distribution of\u00a0the\u00a0area of\u00a0\u200b\u200bthe\u00a0Czech Republic according to\u00a0the\u00a0identified
\nlevel of\u00a0risk in\u00a0the\u00a0case of\u00a0dividing the\u00a0basin into seven clusters<\/h6>\n<\/a>Fig. 6. Recorded erosion events with the risk level of the respective SoLC indicated (high risk in\u00a0red, medium risk in\u00a0yellow, reduced risk in\u00a0green, events outside the\u00a0SoLC in\u00a0white)<\/h6>\n<\/a>Fig.\u00a07. Number of\u00a0recorded erosion events in\u00a0monitoring classified by SoLC risk<\/h6>\nTab.\u00a05. Development of\u00a0the\u00a0risk classification of\u00a0catchment groups produced by cluster analysis
<\/a><\/h5>\nDISCUSSION<\/h2>\n
CONCLUSION<\/h2>\n