Runoff controlling factors in various sized catchments in a semi-arid Mediterranean environment in Spain

Abstract

Understanding land degradation in a semi-arid Mediterranean environment is very difficult because of the contributing factors: precipitation, infiltration vegetation cover and discontinuity of flow and the temporal and spatial levels of resolution at which these factors are acting. Therefore it is sensible to attempt to linearize these relations to make them more amenable to research. One way, used in this thesis is the Representative Elementary Area (REA). The value of the REA-concept was investigated as a tool to overcome the problem of different levels of resolution of the factors contributing to the runoff in various sized catchments. Based on the REA, it is possible to define a resolution at which runoff can be described by simple equations because; The variability in runoff response of the catchment size has reached a minimum compared to smaller sized areas with a more variable runoff response. The runoff response of this area is independent the spatial distribution of the characteristics that control runoff and only dependent of its statistical distribution. The study area in Southern Spain was selected because its geomorphological setting, soil types and vegetation patterns are representative for large areas in the Mediterranean basin. The 9 studied catchments were selected in an area that is known to channel runoff though well-developed streambeds in limestone. The measurement set-up was nested: first order catchments with in them second and third order catchments were selected varying in size between 0.4 and 111 ha. They are covered by natural vegetation with an open structure. In some parts they are covered by afforested terraces, which consist of disturbed soils planted with indigenous pine trees that have not been developed well so far. At some parts, the terraces are not installed according the contour lines. The vegetation cover of the catchment was parameterised at two levels of resolution by the use of satellite images; As classified land cover units. As percentage surface cover per unit area. The land cover units ‘natural woodland’, ‘grassland’, ‘afforested terraces’ and ‘bare soil’ were obtained by conventional supervised classification of the available Landsat imagery. The images were used for Spectral Mixture Analysis (SMA) by which information was obtained on the percentage of cover of the dominant vegetation species Pinus halepensis and Stipa tenacissima per pixel. The Landsat TM images have only six broad wavelength bands. Furthermore the spectral reflectance of the dominant vegetation species Pinus halepensis, Stipa tenacissima and the unvegetated soil surface resemble each other. This lead to limited results of SMA. The most important components of the water balance, precipitation, infiltration and runoff, were studied by detailed fieldwork. The precipitation was characterised by a large number of small rainfall events. Both the storm duration and the rainfall intensity had a skew distribution. The log-transformed rainfall intensity and storm duration resulted in a power-law relation by which the threshold conditions for the occurrence of runoff in second and third order streambeds could be defined. The spatial distribution of rainfall could not be related to topographic features such as aspect and altitude but showed variability within the studied catchments. Infiltration was characterised by the estimation of the saturated hydraulic conductivity (Ks). Proxies were used to find a proper method to map the Ks over the study area. The Ks values were significantly related to the classified units of the land cover map. The variability of Ks within a land cover unit was larger than the variability of the Ks between plant and interplant areas, so it was concluded that Ks was not related to surface cover. Runoff was recorded at the outlet of 9 catchments of which several were nested within each other. The runoff coefficients were very small and decreased with increasing catchment surface which indicated the loss of runoff during transmission in the streambed. Based on the grouped rainfall-runoff recordings it was not possible to define a rainfall-runoff relation which can be used for all catchments. When the rainfall-runoff recordings were grouped into levels of resolution based on the catchment sizes, rainfall-runoff relations were found for each level of resolution which means that the rainfall runoff relation depends on the catchment size. Based on the decreasing variability of runoff with increasing catchment size, the catchments varying from 9.2 – 111 ha. were defined as the catchments in which the variability of runoff had reached a low value, which was one of the constraints for the definition of a REA. This implies that the size of a REA in the study area would be 9.2 ha. For the all catchments ranging from 0.4 – 111 ha. log-linear equations were found by which a rough but quick prediction of runoff was possible based on the channel bottom width, the catchment size and the percentage of the catchment surface that consisted of ‘bare soil’ and ‘afforested terraces’. The relation between the catchment size and the runoff amount or the peak discharge was a power-law relation which indicated that the variation of runoff was self-similar at all levels of resolution, also in the catchment sizes ranging from 0.4 – 2.2 ha. that did not match the REA-constraints. A distributed hydrological model was developed to test the influence of the spatial distribution of land cover units on the runoff amount and peak discharge. The results showed that the peak discharge and runoff amount would increase if ‘bare soil’ and ‘afforested terraces’ were to be located nearer the streambed than in the present situation. When the land cover units ‘bare soil’ and ‘afforested terraces’ cover 50 % or more of the catchment, a random spatial distribution of these land cover units also results in an increasing runoff amount and peak discharge. Obviously, an increase of the areas of ‘afforested terraces’ and ‘bare soil’ will result