IDENTIFYING GULLY EROSION AND DETERMINING ITS RELATIONSHIP TO
ENVIRONMENTAL FACTORS USING GIS IN THE ZHULUBE MESO-
CATCHMENT
Farai Dondofema 1 Amon Murwira 2 and Alexandra Mhizha 3
1 Strategic Environmental Focus, PO Box 7485, Lynwood ridge 0040, Pretoria, South Africa
2 Department of Geography, University of Zimbabwe, Box MP167, Mt. Pleasant, Harare, Zimbabwe
3 Department of Civil Engineering, University of Zimbabwe, Box MP167, Mt. Pleasant, Harare, Zimbabwe
Contact: farai@sefsa.co.za
Abstract
Geographical Information Systems (GIS) and remote sensing techniques were used to identify and map gully erosion extend in the Zhulube Meso-Catchment area of Zimbabwe.
Additionally a comparison was done between gully erosion characteristics and environmental factors involved in this soil degradation process. The objectives of this study were (1) to determine the accuracy of using GIS and remote sensing techniques in gully identification and
(2) to test for significant relationships between gully characteristics and environmental factors.
GIS, remote sensing and field surveys techniques were used to identify gullies within the
Zhulube Meso-catchment of Zimbabwe. The data collection involved analysis of soil core samples and measurement of gully characteristics. GIS and remote sensing techniques were also used to determine the sedimentation and stream power indices with the study area. The statistical analysis focused on the correlation between gully, soil and vegetation characteristics as a means of identifying areas susceptible to gully erosion. The results from this study illustrate that 36% of major gullies are discernible using Landsat TM imagery, 56% using
Spot panchromatic imagery and 77% using Orthophotos. A significant relationship between gully depth and bulk density was evident at r
2
= 0.87 (p<0.05) with the soil clay content and its related minerals showing a significant relationship with gully development. Other significant relationships were evident between gully depths, stream erosive power and slope gradient at r
2
= 0.62 (p<0.05), while streams sediment loadings showed a non-significant effect on the gully depth with r 2 = 0.02 at (p<0.05). It can therefore be concluded that GIS and remote sensing techniques are applicable in gully identification, with variable accuracy levels depending on the spatial, spectral and temporal resolution of the imagery. It was also evident that the inherent susceptibility of soils to detachment and transport by various erosive agents is a function of topography, vegetation and soil properties (physical and chemical). In addition, Soil erodibility assessment using simulated stream erosive forces and sediment loadings revealed that sediment yield and erosive power of the streams in the study area increased with increasing slope gradient depending on the clay content of the soils.
Index Terms— Erosion, Gully, GIS, Environment, Satellite, Water
1

1. Introduction
Inventorying and monitoring of eroded areas at catchment scale (Mm
2
) can be very useful for environmental planning and can help to reduce land degradation and sediment yield to streams. Remote sensing and Geographic Information System (GIS) techniques are common tools for time change analysis. However in most cases satellite imagery or small-scale aerial photography is used as the increased resolution of large-scale aerial photos helps in identifying small features on the ground and is highly useful in the assessment of riparian areas. Combined use of remote sensing images and auxiliary attribute data has been widely used for mapping various environmental features, including surface erosion. Several combinations of radiometric data (a sequence of images from different seasons of the year) and other spatially related information, including topographical (altitude and slope) and geological maps, were compared in their ability to predict previously identified erosive features. The Spearman rank order correlation coefficient calculated and used as the classification method. The datasets were compared in terms of classification error statistics
(sensitivity and specificity) using independent random samples collected along transect laid within the study site.
Soil erosion is approximated to have occurred on 1 .
1 billion hectares of land globally Brown and Wolf (1984). In the Zhulube Meso-catchment of Zimbabwe, soil erosion and sedimentation are key limitations to achieving sustainable utilisation of land and water resources. The most prominent type of erosion in Zhulube is gully erosion which is geographically a widespread problem (Cooke & Reeves, 1976; Lal, 1992) and the worst stage of soil erosion. Anderson et al ., (1993) state that in Zimbabwe alone an estimated 1 million hectares of land are affected by severe gully erosion with the figures rising to 29 million hectares of land in Africa at continental level. SARCCUS (1981) states that gully erosion is a permanent form of erosion which is difficult and expensive to control resulting in a depreciation of land value due to a lowered water table and depleted water reserves. Dregne
(1990), in his survey of Zimbabwe highlights more than half of Zimbabwe’s communal land as from suffering irreparable gully erosion damage, mainly because of Zimbabwe’s land tenure policies.
Zimbabwe inherited a highly skewed pattern of land distribution, with 1% of farmers holding nearly half the available agricultural area and the bulk of the fertile land. The natural resources on any area of land will support in perpetuity only a limited number of people and if this limit
2

is exceeded, without a compensating change in the system of land usage, then a cycle of degenerative changes is set in motion, which must result in deterioration or destruction of the land. As in the case of communal areas in Zimbabwe a decline in resources status, in turn has adverse effects on living standards. As poverty becomes worse, so the ability to effect changes in land use to reverse the cycle of degenerative changes is reduced. In this way a vicious cycle of poverty and land degradation is established.
GIS and remote sensing techniques are most suitable for gully identification and the assessment of an area’s vulnerability to erosion. These techniques thus present an appropriate method of quantifying and combating the gully erosion which is a challenge to the concept of
Intergrated water resources management (Morgan, 1993).
2. Description of the study area.
The study area covered the Zhulube Meso-catchment, which falls within the Mzingwane catchment, in the Limpopo river basin with coordinates of 20° 47′ south and 29° 22′ east. The basin is located within four Southern African countries namely, Zimbabwe, Botswana, South
Africa and Mozambique. The Mzingwane catchment is part of the seven catchments in
Zimbabwe. Four sub-catchments covering 20% of the Matabeleland South and Midlands
Provinces were created from the main catchment. The study focused on ward 1 of Insiza district, which is located in the Upper Mzingwane sub-catchment. The Insiza district is one of the six districts in Matabeleland North province. There are 18 wards in Insiza district, 11 of which fall within Mzingwane catchment. The research work was conducted in the Zhulube village as shown in Figure 1.
3

Bulawayo
#
#
#
Fort Rixon
Agro-ecological
Region IV
#
Matopos
#
Esigodini
#
Mbalabala
#
Filabusi
Zhulube
# #
Kezi
#
Gwanda
# Mphoeng
Shashe
River
Agro-e cologic al Reg ion V
Thuli
River
100 km
M zin gw an e R iv er
#
Mberengwa
West Nicholson
Bubye
River
#
Beitbridge
Rutenga
Mwenezi
Mwenezi
River
Figure 1: Zhulube Meso-catchment, Zimbabwe.
3. Methods
The data acquisition process consisted of stereoscopic interpretation of multi-temporal satellite images and aerial photographs. Sequences of satellite images provided a time-series
(Whitlow, 1986) from which gully changes were mapped while aerial photographs increased the spatial resolution. For the purposes of gully identification using remotely sensed imagery, the smallest scale images were selected from sequential aerial photographs and satellite images of various scales, which were used as the threshold for data extraction (Watson, 1990).
Smaller scales lose detail and larger scales involve a sacrifice in terms of the synoptic view
(Gelmroth, 1981 and Rossouw, 1997).
3.1 Selection of sampling transects
GIS DNR sampling tools were used to select nine transects incorporating the major gullies, discernible range in plant species composition and physiognomy. The sampling tools were designed by the Department of Natural Resources in Minnesota to assist biologists in using
ArcView to generate spatially explicit random or systematic sampling schemes to support resource monitoring, mapping, and research needs. The tool works either with polygons in a
4

theme or with graphics that have been added to a View. Samples can be entirely within a single polygon or shape, or distributed among several disjoint polygons or shapes. A number of user defined constraints and settings are offered as input options. All the transects generated from this option are random both in the placement of the starting point and direction although these can be random placed within a specified direction by setting lower and upper azimuth limits for transects.
3.2 Soil sampling and erosion mapping
The soils in the Zhulube Meso-catchment vary from clayey loams in the north to sandy soils in the south with a rugged high ground. Thus the area was subdivided into homogeneous sections, from which nine sampling sites were located for soil samples. Between 10 and 20 sub-samples where collected and sub-sampled from each transect. The Sub-samples were small enough that the composite sample was of a size that can be completely processed for analysis. Factors like vegetation cover, type and ground hardness determined the sampling depth. An excavated profile pit at the Zhulube dam site, collapsed gully walls and supplement auger borings were used to demonstrate lateral and vertical changes in the soil properties, and for the full description of type soils and for the taking of soil samples for chemical and physical analysis. Soils were mapped at a scale of 1:25,000 by, which the pattern of soils in individual fields could be identified. Figure 2 shows the sampling sites within the study area as selected using GIS and ground truthing methods.
Figure 2: Sampling areas.
3.3 Botanical methods
Discrimination between vegetation classes was based predominantly on the pattern and presence of plant communities. Twenty points were systematically spaced over each transect and the presence of bare or occurrence of an herbaceous plant tuft was recorded. Stunting is a conspicuous feature of the woody species on sodic soils and transects incorporated observed range in height of the dominant woody species. Within each transect all individuals of a
5

species were grouped into 1 meter and greater than 2 meters height classes. Three broad vegetation classes were chosen to facilitate consistent and easy interpretation from the orthophotos and satellite images. Field verification of the mapped vegetation classes involved surveying transects through GPS data collection in the study area and the spatial extent of the area under study restrict the use of Landsat Normalized Difference Vegetation Index (NDVI).
For each height class the bush density, canopy diameter, mean cross sectional area of the stem, that is the breast height of the stem was calculated. Tree height classes lower than two meters were classified as shrubs and the cross sectional area was recorded from the lowest browseable material.
3. 4. Quality control, data analysis and management.
The process of data cleaning and quality control was achieved by making sure the data were in the same formats. The data-handling component of the study involved capturing the mapped vegetation, gully and erosion data into the ESRI ArcView 3.2a, ArcGIS 9.1 and Microimages
TNT MIPS softwares.
3.5. Statistical analyses
Statistical analysis of the research data included correlation determination and regression analysis for vegetation, slope gradient, gully and soil properties. Nineteen variables were collected per site: seventeen soil variables from the collected soil samples, percentage herbaceous cover and depth of sampling. The Spearman rank order correlation coefficient was computed by ranking all values of each variable, then computing the Pearson Product Moment
Correlation coefficient, r , and ranking the parameters.
4. Results and discussion
4.1 Remote sensing of gully development in the Zhulube meso-catchment.
Using a Digital Elevated Model(DEM) the topographic index (TPI) tool was run and areas of notable topographic differences where identified. Depending on the TPI value areas with values below minus one where identified and ground truthing was done in the field. The gullies identified in during the field work where then overlaid onto the TPI results and the results of matching low TPI values and identified and measured gully depths where scored resulting in expression of the accuracy of identification as a percentage of TPI obtained values and groundtruthing values.
6

Figure 3 (a, b and c) shows gullies identified from three images, Landsat TM, SPOT and an
Orthophoto. Figure 3(a), illustrates the gullies that were identified from the Landsat TM image. The smallest gully width was six meters. Figure 3(b) illustrates the gullies that were identified from the SPOT Panchromatic image. It can be observed that the SPOT interpretation achieved 56% accuracy (explain how you get this figure), with a smallest gully width of six meters. Figure 3(c), illustrates the gullies that were identified from the
Orthophoto. Figure 3(c) also shows a gully identification accuracy of 77% (explain how you get this figure), more information on the occurrence of the gullies becomes evident mainly due to its fine spatial resolution of two and a half meters, with much is less than the average gully width of 17m within the study area. When carrying out the groundtruthing exercise the gullies identified by the satellite imagery are shown as in Figure 3.
7

(a) (b)
Figure 3(a, b & c): Gully identification using Landsat satellite imagery (a), Spot imagery (b) and Orthophoto (c).
(d)
8

4.2 Relationships between gully erosion and vegetation characteristics.
Statistical analysis using the Spearman Correlation Matrix between the gully characteristics
(gully depth, gully length, gully area, gully width), with vegetation characteristics was done.
There was a significant relationship (p < 0.05) between the most of the gully characteristics and the vegetation components with the exception of gully depth and basal cover and tree density and gully volume which did not show any significant relationships.
Table 1 shows the Spearman Correlation Matrix.
TREE BASAL
DENSITY COVER
GULLY
DEPTH
BASAL COVER 0.217
GULLY
WIDTH
0.345 0.092
TREE DENSITY
GULLY GULLY
LENGTH VOLUME
-0.110
0.217 -0.143 -0.084 0.521
0.018
-0.008
GULLY
AREA
-0.202
0.101
4.3 Relationships between gully depth and other properties.
The Pearson correlation analysis showed that the soil chemical characteristic, bulk density had a highly significantly (p < 0.05) effect on gully depth as shown in Figure 4. The relationship shows linear increase in gully depth as the bulk density of a soil increases due to the vertical removal of soil mass by flowing water. The results also show that site 3 had a higher tree density than other sites; a look at the collected vegetation data shows that even though there are adequate woody species around the gully with a tree density of 600 plants per hectare and a tree index of 3.6 is not enough to hold the soil from further erosion.
1.40
1.35
y = 0.1056x + 1.1502
R
2
= 0.619
p=0.002
1.30
1.25
1.20
1.15
1.10
0.00
0.20
0.40
0.60
0.80
1.00
1.20
Gully Depth(m)
Figure 4: Relationships between gully depth and bulk density
1.40
1.60
1.80
2.00
Figure 5 shows illustrates that a slight increase in the proportion of clay within the soils causing significant (p <0 .05) declines in gully depth across the sampling sites. The clay
9

content showed great inter-sites variation ranging from 2 % to 8 % between the sampling sites, indicating that the study area was dominated by sandy soils, which are weak and prone to erosion. The trend depicted by the four minerals that had a significant effect on the gully depth indicate a similar pattern, with all four chemical components presenting a negative significant correlation coefficient with the exception of sodium. This can be explained by the fact that magnesium, iron, manganese and sodium are major constituent of clay minerals, which include the layer silicates, the metal oxides and crystalline chain silicate minerals like kaolinite, chlorite, illite and montimorilonite.
5
4
3
2
7
6
9
8 y = -3.0764x + 8.5551
R
2
= 0.528
p=0.02
1
0
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
Gully Depth (m)
Figure 5: Relationships between gully depth and clay properties
The possible severity of erosion can be estimated roughly by considering the degree of the slope and other related environmental factors. Figure 6 illustrates a positive significant relationship with a correlation coefficient R 2 = 0.62. This can be interpreted as a resultant increase in gully depth as the slope gets steeper, as the erosive power of a flowing water body increases linearly with an increase in slope.
10

5
4
7
6
9
8
3
2
1 y = -3.0764x + 8.5551
R
2
= 0.528
p=0.02
0
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
Gully Depth (m)
Figure 6: Relationships between gully depth and clay properties
Stream power varied within the study area appearing to be more depending on the gully width, as illustrated in Figure 7. There is a high positive correlation of stream power with gully depth
(R
2
= 0.612), meaning that if the stream flow increases the gullies increase in their widths.
1040
1030
1020
1010
1000 y = 1.1319x + 982.07
R
2
= 0.612
p=0.042
990
980
970
0 5 10 15 20 25 30 35 40 45
Guly Width (m)
Figure 7: Stream Power Index, significant (p<0.05) relationship between gully width and stream power index
The sediment transport index (Figure 8) illustrates the potential sources and transportation of sediments. The dark colours indicate potential sources of sediments while the lighter toned colours show areas of potential sediment deposit within the catchment. Low sediment loads are indicated within the greater parts of the Meso-catchment but it is evident that the final sampling transects fall in close proximity to areas of high sediment transport indices, meaning around these areas the erosivity of the streams is high thus resulting in relatively deep gullies.
The is a significant (p<0.05) increase in sediment transport values as you move closer to site
10, which show very high sediment loading values and these tally with the field collected data as these were some of the deepest and broadest gullies in the study area. The regression of the gully depth and the sediment transport index show a weak correlation between the depth of the gullies and the amount of sediment that is transported.
11

80
60
40
20
120
100 y = 12.509x + 42.946
R
2
= 0.5052
p=0.04
0
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
Gully depth (m)
Figure 8: Sediment transport index, a non-Significant (p> 0.05) relationship between Gully Depth and
Sediment Transport Index.
5. Conclusions and recommendations
The general conclusions that can be drawn from this study is that significant relationships exist between vegetation types, soil characteristics (physical and chemical), and topography and gully development. The use of high resolution imagery and GIS techniques can aid in gully identification with high and desirable levels of accuracy. The inherent susceptibility of soils to detachment and transport by various erosive agents was a function of the physical and chemical properties of soil.
The interactive effects of the topographic, vegetation cover and rainfall factors in turn influence these environmental factors and gully characteristics. Sampling sites at lower slope gradients had reduced stream power and sediment loads than those located at areas with higher slope gradients like site thus making slope an important factor in gully erosion. Soil erodibility assessment using simulated erosive force (stream power index) and sediment loadings (stream power index) in the study site revealed that sediment yield or the erosive power of the streams in the study area increased with increasing slope gradient depending on the clay and other minerals content of the soils meaning the richer the mineral content the lower the erodibility factor as shown by the relationship between clay content and the gully depths.
Understanding how the land use change influence the river basin hydrology will enable planners to formulate policies to minimize the undesirable effects of future land use changes.
Land cover changes increase impervious ground surfaces, decrease infiltration rate and increase runoff rate, hence causing low base flow during the dry seasons. Efficient tools such as satellite remote sensing and Geographic Information System (GIS) are currently being used to manage the limited water resources.
12

References
Anderson, I.P., Brinn, P.J., Moyo, M. and Nyamwanza, B. 1993. Physical resource inventory of the communal lands of Zimbabwe – An overview. NRI Bulletin 60. Chatham,
UK: Natural Resources Institute.
Brown, L. R., and E. C. Wolf. 1984. Soil erosion: quiet crisis in the world economy
Worldwatch Paper 60, Worldwatch Institute, Washington, DC.
Cooke RU and Reeves RW, Arroyos and environmental change in the American Southwest.
Oxford, UK: Oxford University Press. 1976
Dregne H.E, "Erosion and Soil Productivity in Africa", JSWC, 45: 431-36. 1990
Gelmroth, H., 1981: Mapping of Soil Erosion from Aerial Photographs, Ministry of
Agriculture, Division of Land Utilisation, Gaborone, Botswana.
Keech, M.A: Remote sensing in planning the control of erosion, in De Boodt, M. and
Gabriels, D. (eds), Assessment of Erosion, John Wiley and Sons, Chichester,
419-425. 1980
Lal R. 1992. Restoring land degraded by gully erosion in the tropics. Advances in Soil
Science 17:123–152.
Morgan, R.P.C., 1993: Soil erosion assessment, Proceedings of a Workshop on Soil Erosion in Semi-Arid Mediterranean Areas, Taormina, Italy, 28-30 October 1993,
Consiglio Nazionale delle Ricerche, Roma, 3-17.
Rossouw, N. J., 1997: Mapping vegetation and erosion changes on the northern slopes of
Table Mountain using multi-temporal aerial photography and GIS, 1944-1992.
South African Geographical Journal 79(2): 136-146.
SARCCUS, 1981: A System for the Classification of Soil Erosion in the SARCCUS Region,
Department of Agriculture and Fisheries, Pretoria.
Watson, H.K, A Comparative Study of Soil Erosion in the Umfolozi Game Reserve and
Adjacent Kwazulu area from 1937 to 1983, Unpublished PhD thesis, University of Durban-Westville. 1990.
Whitlow, R., 1986: Mapping erosion in Zimbabwe: a methodology for rapid survey using aerial photographs, Applied Geography, 6, 149-162.
13