Hydrology South Africa

[Shima Costar]

Given the many users in South Africa of the SCS (Soil Conservation Service) hydrograph-generating technique, which requires key soils information, this paper presents a methodology to develop a detailed map (at the spatial resolution of South Africa's 27 491 terrain units) of the SCS runoff-related soil groups. In order to achieve this objective, background is first provided on the soils context within the SCS technique, as well as on the approaches to defining SCS hydrological soil groups, by assigning SCS soil groups to South African soil series. Steps when mapping SCS soil groups across South Africa at terrain unit spatial resolution are then outlined. This is followed by the presentation and interpretation of the detailed terrain unit level map of SCS soil groups for South Africa. The detailed new map illustrates the finer detail across contributing catchments countrywide now obtained from the terrain unit approach, compared to the previous map of SCS soil groups based only on the coarser soil land types.

There is a frequent need for hydrological information in the planning and design of water-related structures such as culverts, stormwater systems or spillways on small catchments, i.e. with areas < 30 km2. Runoff volumes and peak discharge rates are commonly required for selected design return periods, and in the absence of long-term measurements at the locations where such structures are required, these have to be estimated by simulation modelling. Such simulation models inherently account for rainfall, land use and catchment hydraulic characteristics, but the importance of soil characteristics is often underestimated. The reason for this is a lack of appreciation that it is the soil that absorbs, retains and releases water following a rainfall event, and that soil is thus a prime regulator of the catchment's response to rainfall.

One rainfall-runoff model which has become accepted and established worldwide, also in South Africa (SA), for use on small catchments is the USDA's (United States Department of Agriculture) Soil Conservation Service (SCS) technique for the estimation of design flood volume and peak discharge (USDA 1972 and updates).

Previous soil-related mapping across SA for use with the SCS technique was based on the hydrological characteristics of so-called soil land types, of which just over 6 000 had been mapped for the country (Schulze 2012). This spatial resolution is, however, relatively coarse when used across small catchments, for which the SCS technique had been designed.

The objective of this paper is to present a methodology by which a finer detailed map of SCS soil-related information for SA has been developed at the spatial resolution of "soils terrain units" (TUs), of which over 27 000 were identified across the country. The reason for the detailed mapping is that many users of the SCS in SA may not have the time, nor the background, to undertake the necessary soils fieldwork when applying the SCS-SA model for small catchments. The detailed SCS soils mapping outlined below is anticipated to reduce errors in simulated runoff volumes and peak discharges, compared with the coarser soils information previously available.

Thereafter the outcomes of the study are presented in the form of maps of SCS soil groups across SA at terrain unit spatial resolution, an interpretation of the maps, and a section on applications, followed by some concluding thoughts.

Crucial to this paper from Equation 1 is the potential maximum soil water retention S, because it includes the catch-ment's soil moisture status just prior to a rainfall event, with S related, inter alia, to hydrological soil properties, categorised into seven hydrological soil groups. S finds expression through a dimensionless runoff response index termed the catchment's curve number (CN), which is related to S by:

Before any adjustments for a catchment's wetness condition just prior to a runoff-producing rainfall event are made, an initial CN is determined according to soils and land-cover properties. These initial CNs have values given in SCS literature for a wide range of land cover and treatment classes, and hydrological soil groups. Important in the context of this paper is that the initial curve numbers, and hence S in the runoff equation (Equation 1), vary by hydrological soil groups, which for South African conditions have been categorised into seven groups (see later).

Soils in South Africa were assigned to the seven hydrological soil groups based on texture, compaction (bulk density) and strength of soil structure (Schulze 1995) for use in the estimation of soil water movement, which is controlled by its infiltration rate at the surface and its permeability. The seven hydrological soil groups are termed as follows:

During the course of the development of the first South African SCS User Manual (Schulze & Arnold 1979), and in subsequent research on runoff responses to soils properties (e.g. Schulze et al 1985), an expert group of soil and hydrological scientists classified the country's 501 soil series into SCS soil groups (A, A/B, ... D). This was done using the South African Binomial Soil Classification (MacVicar et al 1977) on the basis of soil texture, leaching, crusting, water table height and clay distribution models, which are related to clay properties with soil depth.

While a soil land type is defined as "a homogeneous, unique combination of terrain type, soil pattern and macroclimate zone" (SIRI 1987), based primarily on agricultural rather than hydrological properties, the polygons of these land types, of which there are over 6 000 across SA, then determined the spatial resolution at which SCS soil groups across SA were mapped (Schulze 2012). See Figure 1.

With soil classifications in SA continually evolving, this study, however, retained the SCS soil groupings according to the SA Binomial Classification by MacVicar et al (1977), rather than soil groupings from more recent South African soils classifications, e.g. the Taxonomic Classification (SCWG 1991). The reason for this is that the Binomial Classification is the classification at which soils of SA were spatially defined. An attempt to match soil properties between the two classifications can be found in Schulze and Schtte (2020).

As defined in SIRI (1987), a terrain unit (TU) is a subdivision of land types with homogeneous slope and form, e.g. a convex or a concave slope, or a near-vertical cliff. Terrain can be thought of as being made up of all, or some, of the following kinds of TUs: a crest, scarp, midslope, footslope, and the valley bottom, as shown in an idealised representation in Figure 2. Each of these TUs will likely have different hydro-logical soil properties, and hence different runoff and peak discharge responses, which will be better reflected in the TU-based SCS-SA classes than in the coarser land-type-based classes. More detailed information on the TUs is given later.

Assigning an SCS soil group to each of the 501 Binomial Classification soil series identified in SA (Schulze 1985) proved only a first step towards mapping the SCS soil groups at the TU spatial resolution. The steps which needed to be taken in order to map SCS soil groups per TU were as follows:

The (now digitally) delineated and mapped 6 000+ soil land types are accompanied by detailed tabulated inventories of information on each land type which were compiled using data collected during the terrain, soil and climate survey phases (Land Type Survey Staff undated). Each land type is made up of a number of soil series, with the inventory of each land type containing information regarding, inter alia:

More importantly, in the inventory of each of the 6 000+ land types there is information on the soil series which are found in up to 5 TU classes making up a land type (see Figure 2). These TUs, which had been identified in the field and were named in the section above, are the crest (defined as being convex-shaped), the scarp (akin to a cliff), the midslope (concave in shape), the footslope (also concave in shape) and the valley bottom (also known as the wet area adjacent to a stream).

Because each TU within a land type comprises different (but usually related) individual soil series, each with different soil thicknesses and hydrological properties, it stands to reason that each TU would have a different hydrological response. This is illustrated in Figure 3 in which the different soil water contents of the topsoil of the crest (TU 1), of the footslope (TU 4) and the topsoil of the valley bottom (TU 5) of land type Ac207, all modelled by the ACRU daily time-step and process-based model (Schulze 1995 and updates), are shown for an identical climate time series.

While still at the Agricultural Research Council (ARC), Hein Beukes (Beukes 2018, personal communication) superimposed a Digital Elevation Model (DEM) with a 90 m resolution (pixel) over each mapped soils land type identified in SA, and then used Neighbourhood Analysis to compute the slope to the north, east, south and west pixel from the pixel of interest in order to determine convexity or concavity to the neighbouring pixels. On the premise that the crest was convex-shaped, but that the midslope was concave in shape, he then applied zonal statistics to delineate and then map the individual TUs within a land type. Following that procedure, he then superimposed satellite imagery at a 30 m resolution over the map of TUs to refine the delineations of the TUs, especially the valley bottoms in hilly areas. This was because the 90 m resolution of the DEM, while being at a satisfactory spatial resolution in flattish terrain, was found to be too coarse where it was hilly. From Beukes' ARC Database of Terrain Units per Soils Land Type, based on the refined 90 m DEM, 27 491 spatially defined TUs were identified across SA by the techniques described above.

To illustrate the spatial detail from mapping SA's soils at the TU level, an area to the north and west of Durban in KwaZulu-Natal was selected (Figure 4). The figure shows the delineated crests in dark brown, midslopes in light brown, footslopes in green and the valley bottoms in dark blue (Figure 4). Scarps, taken to be vertical to near-vertical, are assumed to have no spatial extent when mapped. When the rectangular inset in Figure 4 is zoomed in upon in Figure 5, the degree of detail of the TU delineations is even more evident, with individual pixels of a specific TU visible in places. It is at this level of spatial detail that the SCS soil groups were mapped.

c80f0f1006