Soil loss research work was carried out on three experimental plots at Kamundi catchment area, which is situated in Mthiramanja EPA under Mangochi RDP in Machinga ADD in the Southern Region of Malawi.

These experimental plots were treated differently as follows: plot 1 was tilled, bare and had no ridges, plot 2 was tilled, mulched and had no ridges, and plot 3 was tilled and had maize planted on ridges. Collection tanks were used to measure the amounts of runoff and eroded soil sediments from each plot. Sediment samples were analyzed at Chitedze Research Station. Rainfall data was collected using automatic and improvised rain gauges. Slope steepness values were also recorded.

Soil losses caused by individual storms, monthly rainfall and seasonal (December to March) were computed using both RUSLE and SLEMSA models. A Sensitivity analysis was also carried out on all the parameters of SLEMSA model. Significantly higher soil sediments were collected from plot 1 followed by plot 2 and the lowest soil sediments were collected from plot 3. The results showed that using individual storms and monthly rainfall, RUSLE over estimated while SLEMSA under estimated the values of soil loss. On the other hand, using seasonal rainfall, the RUSLE model gave best prediction on plot 2 better on plot1 but overestimated on plot 3 while SLEMSA gave better predictions on all plots.

Based on these findings, both models RUSLE and SLEMSA have proved to be suitable for estimating soil loss using seasonal rainfall i.e. estimate long term soil loss rather than individual storm and monthly rainfall.

Soil erosion by water is a major threat to sustained land and crop production and causes degradation of water resources. In 1994, soil erosion was estimated to range from 13mt\ha\yr to 29mt\ha\yr averaging to 21mt\ha\yr. This results in yield loss of between 4% to 11% per year. The cost associated with erosion has been estimated at K1,155 million per year corresponding to about 8% of the country's GDP of 1994 (VISION 2020,1997).

Since soil loss is a continuing threat to agricultural systems, it is important to promote adoption of conservation practices by farmers. The effectiveness of conservation practices depends upon number of factors. Before planning any conservation work, it is helpful to assess the amount of soil loss from the agricultural lands. In addition, it is useful to determine the effects of different conservation strategies on soil erosion rates. Hence there is a need to develop methods that can predict soil loss under a wide range of conditions.

The predictive techniques ideally should satisfy the conflicting requirements of the reliability, universal applicability, easy usage with a minimum data, comprehensiveness in terms of the factors included, and the ability to take into account of the changes in land use and conservation practices. Because of the complexity of the soil erosion process with its numerous interacting factors, the most promising approach for developing a predictive procedure lies in formulating a conceptual model of the erosion process.

Soil erosion models are used for indicating the severity of soil erosion under the existing land use practices. Soil erosion models are design tools used in formulating soil conservation practices. Therefore the importance of an erosion model in land use planning can be measured by the ability of the model to provide acceptable predictions of the soil loss for a given set of conditions.

Consequently, the objectives of this research were to evaluate the RUSLE and SLEMSA models and determine the model that has more predictive potential under Malawi conditions. The second objective was to identify the most sensitive parameters of each model.

Various soil erosion models have been developed and used worldwide. These models include Revised Universal Soil Loss Equation (RUSLE) and Soil Loss Estimation Model for Southern Africa (SLEMSA).

The Revised Universal Soil Loss Equation (RUSLE) model is a revised version of the USLE model developed in the U.S. for predicting water erosion. RUSLE computes sheet and rill erosion from rainfall and the associated runoff. RUSLE is principally used to estimate the rate that erosion is removing soil from critical parts of the landscape and to guide the choice of conservation practices that will control erosion to a "soil loss tolerance level" (RUSLE user guide, 1995). Soil loss tolerance, or T, values have been assigned by Natural Resources Conservation Society (NRCS) to major soils in the U.S.. Typical values of T range from 3 to 5 ton /ac/yr. with some values as low as 1 ton/ac/yr. for range and other land uses where the soil loss is very fragile. If the computed value is less than T-value, control of sheet and rill erosion is assumed to be adequate. If the computed soil loss exceeds the T-value, sheet and rill erosion is considered to be excessive and improved erosion control is needed.

According to RUSLE user guide (1995), soil loss values computed by RUSLE should be used as a guide rather than being as considered absolute values and that soil loss values computed for a profile is representative of an area to the degree that the profile represents the area. It does not compute average sheet and rill erosion for a field unless soil loss is computed for several profiles and the results weighted according to the fraction of the field that each profile represents. RUSLE does not compute sediment yield (Haan et.al.1994).

RUSLE was developed as a means to estimate soil loss, to point out factors that cause erosion and to present alternatives for controlling erosion. It was developed using a more than 40 year data measured from small plots located in many parts of the United States (Wischmeier and Smith, 1978). The RUSLE model has been widely used for planning purposes to predict the impact of land use on soil erosion. The model is data intensive because its relationship and parameter values are derived almost entirely from experimental data. This model has been used for soil conservation planning in Malawi despite limited availability of input data (Mwendera, 1988). It reflects the influence of all the major factors known to influence water erosion. RUSLE or USLE is a multiplicative relationship of factors and is expressed as follows:

Figure 1 RUSLE framework

A=RKLSCP (1)

Where:

• A is the average soil loss per unit of area, expressed in the units selected for K and the time period specified by R. Normal English units are tons/acre/year but other units can be used like tons/ha/yr.
• R is the rainfall/runoff factor, which is the number of rainfall units for rainfall energy and runoff, plus a factor for runoff from snowmelt.
• K is the soil erodibility factor, which is the rate of soil loss per unit of R (erosion index units) for a given soil under continuos fallow with up and downhill cultivation on a slope length of 72.6ft (22.1m)
• L is the slope length factor, which is the ratio of soil loss from a defined slope length relative to that from a slope length of 72.6-ft (22.1 m).
• S is the slope steepness factor, which is the ratio of soil loss from a slope with a given steepness relative to that from a 9% slope.
• C is the cover and management factor, which is the ratio of soil loss from an area with a given cover and management relative to that from an identical area in continuos fallow.
• P is the supporting conservation practice factor, which is the ratio of soil loss from a field with a conservation support practice such as contouring relative to that with straight row farming up-and downhill.

R, K, L, and S are factors which indicate the susceptibility of an area to erosion while L, S, C, and P factors are each dimensionless which allow comparison of the site being estimated with the standard conditions of the database (Bennema and De meester, 1980). RUSLE factors are soil loss measurements derived directly from field plots. This is expensive because the experiment must be continued for 30 years before reasonable estimates of the mean soil loss are obtained. Consequently treatments can be rendered obsolete by changes in agricultural techniques before the measurements are completed. And that it requires many thousands of plots to measure from every possible combination of climate, soil, crop, slope, and tillage and management practices (Elwell, 1978).

The Erosivity factor is a product of rainfall energy and maximum 30-minute intensity divided by 100 for numerical convenience known as the EI30 index. On annual basis, the EI30 value is the sum of values over the storms in an individual year. The erosivity of rainfall varies greatly by location because of the effects of elevation in rainfall. Wischmeier & Smith (1958) observed that there's a high correlation between rainfall kinetic energy and its maximum intensity, EI30 and the amount of soil eroded. Erosivity for a single storm is the product of the single storm's energy E and its maximum 30 minute intensity (I30) for qualifying storms. Thus erosivity factor is an indication of the two most important characteristics of a storm determining its erosivity: amount of rainfall and peak intensity sustained over an extended period (RUSLE user guide, 1995).

Calculations for rainfall energy require an algorithm relating energy to some measurable parameter. Up to an intensity of 3 in/hour, rainfall energy increases with increase in storm intensity because drop size and fall velocity increases with increase in intensity above 3 in/hr. When the drop size reaches its maximum size, however, rainfall energy remains constant (Haan et.al 1994).

Morgan (1979) criticized the EI30 index for basing its value on estimates of kinetic energy and that the index is of suspect validity for tropical rainfall of high rainfall intensity. The EI₃₀ index assumes that erosion occurs even with light intensity rains whereas Hudson (1965) has shown that erosion is almost entirely caused by rain falling at intensities greater than 25 mm/hr. The inclusion of I30 in the index is an attempt to correct the overestimation of the importance of light intensity rain but it is not entirely successful because the ratio of the intense erosive rainfall to non-erosive rainfall is not well correlated with I30. Stocking and Elwell (1973a) recommended its use only for bare soil conditions.

Ideally, soil erodibilty is a measure of a soil resistance to the erosive powers of rainfall energy and runoff. K factor reflects the susceptibility of a soil type to erosion i.e. it is the reciprocal of soil resistance to erosion (FAO, 1993). Thus K is the average soil loss and is expressed in tons per hectare. The larger the value of K, the easier that particular soil erodes. Practically, soil erodibility is an integration of the impacts of rainfall and runoff on soil loss for a given soil. Experimentally, soil erodibility is the soil loss per unit rainfall index on a standard erosion plot, i.e., a plot under fallow conditions on a slope of 9% with a slope length of 72.6 ft with up- and downslope tillage. In RUSLE, K is assumed to be constant throughout the year. In the absence of published data, nomographs are widely used for predicting erodibility where K values are predicted as a function of five soil and soil profile parameters: %silt, %very fine sand, %sand, %organic matter, soil structure and permeability. K values vary with texture, aggregate stability, shear strength, infiltration capacity and organic matter and chemical content. Some researchers have shown that K also varies with antecedent moisture and with freezing and thawing.

The effects of topography on soil erosion are determined by the dimensionless L and S factors which account for both rill and interill erosion components. LS factor expresses the ratio of soil loss under a given slope steepness and slope length to the soil loss from the standard conditions of a 5% slope, 22 m long for which LS =1.0

L is a ratio which compares the soil loss with that from a specified length of 22.6m (FAO, 1993). S is a ratio which compares the soil loss with that from a field of specified slope of 9% (FAO, 1993) and is used to predict the effects of the slope gradient on soil loss.

According to Morgan (1990), C is the ratio of soil loss under a given crop to that from bare soil. Soil loss varies with erosivity and the morphology of the plant cover. Therefore changes in plant morphology during the year are taken into account while estimating annual C value. The C factor also accounts for the effects of cover above the ground (ground cover), root mass, incorporated residues, surface roughness and soil moisture on soil erosion. The C factor also includes the effects of crop sequence, productivity level, length of growing season, tillage practices, residue management and the expected time distribution of erosive rainstorms (Schwab et al, 1981). The tabulated values for C were meant to determine crop rotation and management practices in the USA. As a result, for other countries average annual C values are used. C factors change with land use and cover resulting in varying erosion rates over the course of a season even with invariant month-to-month rainfall energy.

P is the ratio of soil loss from any conservation support practice to that with up- and downslope tillage. It is used to evaluate the effects of contour tillage, strip cropping, terracing, subsurface drainage and dry land farm surface roughening. If the surface is tilled with up- and downslope row orientation or relatively smooth, a drainage pattern that allows eroded sediment to be readily transported downslope develops. If tillage is on the contour, however, excessive rainfall is stored in the furrows between the tillage ridges allowing significant amounts of depression storage and sedimentation. The effectiveness of contouring depends on the ability of the tillage marks to store runoff and is obviously imparted by the size or roughness of the tillage system, the amount of runoff, and the peak intensity of the rainfall. As concentrated flow moves downslope, the quantity of runoff increases thus reducing the effectiveness of a given contour tillage system. The P values are obtained from tables of the ratio of soil loss with and without the practice. P values vary with the slope steepness.

The Soil Loss Estimation Model for Southern Africa (SLEMSA) was developed largely from data from Zimbabwe highveld. SLEMSA was used for evaluating soil erosion resulting from different farming systems (Morgan, 1990). The model was derived to estimate soil losses from sheet erosion arising from agricultural practices on the lands between terraces. According to Elwell (1996), SLEMSA framework is a systematic approach for developing models for estimating sheet erosion from arable lands in Southern Africa. Mwendera (1988) wrote that the model considers soil erosion process as an interaction of energy, resistance and protection factors.

The procedure adopted in building the SLEMSA model was to divide the soil erosion environment into four physical systems: -climate, soil, crop and topography. Out of these four physical systems, five control variables were identified namely energy interception (I), rainfall energy (E), soil erodibility (F), slope steepness (S), and slope length (L). These are the major overriding factors controlling soil losses in each system.

Then these control variables were arranged into three submodels namely: -

· Principle sub model (K) to estimate soil loss from a bare soil

· a sub model to account for cropping practices (C), and one

· to account for differences in topography (X)

A correction factor is applied to the topographic sub model X to account for ridging practices. The main model expresses the relationship between the sub models. These sub models were formulated to interact as simple products in the main models as follows:

FIGURE 2. SLEMSA framework

Z = K C X (2)

Where:

• Z is the predicted mean annual soil loss (ton/ha/yr.) from the land under evaluation.
• K is the annual soil loss (ton/ha/yr.) from a standard conventionally tilled field plots 30m by 10m at a 4.5% slope for a soil of known erodibility (F) under a weed free fallow.
• X is the ratio of soil loss from a field slope of length (L) in meters and slope percent (S) to that lost from the standard plot.
• C is the ratio of soil loss from a cropped plot to that from bare

This is the mean annual soil loss t/ha/yr. from a conventionally -tilled field plot (30m by 10m at 4.5% slope), for a soil of known erodibility (F) under a weed free bare fallow. K is a function of mean seasonal rainfall energy (E) and soil erodibility (F). Rainfall energy (E) is influenced by rainfall characteristics such as raindrop size, duration and intensity. Several authors investigated the effects of rain energy on erosion processes and concluded that E is dependent on the soil type, crop type and crop stages. Elwell and Stocking (1973) found that cumulative rainfall momentum above intensities of 2.1 mm/hr are better estimation parameter for rainfall. Armstrong (1980) stated that there are differences in the results found by several researchers of different places which is supporting Hudson’s caution that the application of an energy relationship developed for USA may not be appropriate to other geographical areas.

Determination of soil erodibility (F) is dependent on the soil type, and a wide range of field conditions. On the soil type, erodibility is the function of infiltration and runoff, which affects the aggregate stability and particle dispersion. Tillage treatments and some cropping practices also influence soil erodibilty (Elwell, 1977). The input values for soil erodibility are obtained from the attached tables.

The Topographic factor (X) adjusts the value of soil loss calculated for a standard condition to that for the actual conditions of slope steepnees and slope length. Slope steepness influence soil loss and soil loss increases exponentially with slope. This is generally acceptable with an exception of mulch cover on Nigerian alfisols where soil loss and slope were not related (Mutchtler & Murphree, 1980).

The Crop factor (C) is derived from the energy interception factor (i) which is determined by the crop type, yield and emergence date for crops, natural grasslands, dense pastures and mulches. C is a function of percent energy intercepted and is determined by seasonal percent crop cover and energy distributed curves. This method separates treatments that influence crop canopy. The value of C is dependent upon the percentage of the rainfall energy intercepted by the crop (i). The value of i is obtained by weighing the percentage crop cover in each ten day period by the percent of the mean annual energy (E) occurring in that period and summing the values.

· >Both models, were designed to estimate soil loss for seasonal rainfall and not single or monthly rainfall, however there are some places where annual rainfall amounts are very low or very high hence they may not give good predictions of soil loss in such areas. For instance SLEMSA was designed to predict soil loss for areas with annual rainfall ranging from 800-1200 mm implying that areas with less than 800mm and above 1200mm can not use SLEMSA for predicting soil loss.

· In both models, the effect of antecedent moisture content is not considered explicitly, This however, affects the erodibility of soil for a particular rainfall.

· RUSLE estimated some soil loss values even where there were no actual sediments collected from the experimental plots.

· In both models, runoff to which soil loss is closely related is not considered explicitly.

· RUSLE is restricted to slopes where cultivation is permissible, normally 0-7o and to soils with low content of montimorillonite (clay minerals); RUSLE does not provide erodibility values for sandy soils.

• RUSLE assumes that the values of cover factor does not vary over the season i.e. it is constant and it also assumes that all crop have that the same value of C.
• SLEMSA requires the knowledge of the past land management practices in determining the F value. This might limit its use for a person that is new to an area.

This research work was done on already existing experimental plots at Kamundi catchment in Mthiramanja EPA, Mangochi RDP in the Machinga ADD. It was an on-going project under Malawi Environmental Monitoring Program (MEMP). This study was carried out during the months of December 1997 to March 1998. Each plot was surrounded by iron sheets raised about 30cm above the ground surface to contain the runoff within the designated plot area. Runoff flowed downslope to the lower end of the plot where there was a receiving channel leading into the collection pit with dimensions of 1m by 1m by 1m (1m3). The layout of the experimental plots and their respective treatments are shown below:

Figure 3. Layout of the experimental plots

PLOT 1

This plot was tilled, had no contour ridges and it was left bare. In case of growing weeds, careful hand weeding was done and compaction was avoided or minimized.

PLOT 2

This plot was under flat tillage and however it had residue cover treatment (10 Kg of maize stalks). In case of weeds, careful hand weeding and compaction was avoided or minimized.

PLOT 3

Plot 3 had contour ridges and maize crop was planted on top of the ridges. The ridges were 30 cm high and 90 cm apart. Weeding and banking were done. The following parameters were recorded: ridge height, ridge interval, number of plants per planting station and per row.

Improvised rain gauges were installed at the experimental plot site. Using improvised rain gauges and recording rain gauges, rainfall data was collected. Measuring rulers were used to measure the runoff data collected in the plot pits after each rainfall event. After each storm that caused runoff, sediments and water samples were collected from the experimental pits as follows:-

After the storm, water collected in the pits was stirred. This was done to make sure that all sediments were in suspension state. Then two water samples were drawn. These water samples were sent to the Central Water Quality Laboratory for analysis.

Firstly, suspended sediments and water samples were removed using buckets until water was approximately 2cm at the surface of the deposited material. Then, using a flexafoam, the remaining water was sucked. The sludge was leveled and the depth of the sediments was measured using a ruler. Sediments of known area were taken, dried and put in plastic bags. These soil samples were analyzed at Chitedze Research Station (Soil Chemistry Lab.). The pits are cleaned after each rainfall event in preparation for the next rainfall event.

Total soil losses were calculated using RUSLE and SLEMSA models as given in appendix I using a single storm, monthly rainfall, and seasonal rainfall (December to March). The calculated soil loss was compared to the actual sediment collected from the pits. The input variables used in each model are also given in appendix III.

Table 1 shows the amounts of sediments collected from the each plot. These plots were located under the same slope length and steepness, soil characteristics and rainfall amount and intensities. Therefore, the differences in the soil loss are attributed to the plot treatment only.

Table 1 Total sediments (tons/ha/yr.) collected per month

Substantially higher soil losses were recorded on the bare plot. This is due to higher degradative impacts of raindrop, rainfall intensity and high amounts of runoff which caused detachment and transportation of the soil. As flow moves downslope, the quantity and velocity of runoff increases thereby eroding more soil. A 51% decrease in soil loss was recorded in plot 2 as compared to plot 1despite having the same treatments as plot 1 except for 10 kg of mulch on plot 2. From the conservation point of view, mulch simulates the effects of a plant cover. Thus mulches on plot 2 provided protection to the soil against the blasting action of falling raindrops thereby preventing the soil from being eroded. Mulches absorb the energy of raindrops which cause soil detachments. Mulches also slow down the flow of runoff (velocity) thereby allowing for some runoff to infiltrate into the soil as well as some eroded sediment to be re-deposited within the plot due to reduced runoff velocity. The lowest soil loss was obtained from maize plot. This was due to the effect of contour ridges and crop cover. The cover protected the soil from the direct impact of raindrop by absorbing rainfall energy. It was observed that more soil sediments were recorded in December and January. This was due to small percentage maize cover because the crop had not yet grown to provide enough cover to protect the soil. Plot 3 had the lowest amount of sediments collected because of the effects of contour ridges and maize cover. Contour ridges allowed for more runoff to be served between the ridges thereby allowing more of it to infiltrate into the soil thus protecting the soil from being eroded

These results agree with Hudson (1971), that differences in soil erosion caused by different land management practices on the same soil is much more greater than the differences in erosion from different soil types given the same management. It is also found that the effectiveness of management practices in preventing erosion is proportional to the amount of cover present when the rain falls. This is considering the level with which the land surface is covered from the rain drops and the bulk or weight of the cover present to absorb the energy of the raindrops. It is important to note that sediment yield is not the same as gross erosion occurring within the catchment surface. The sediment yield is generally low because the soil particles eroded from the catchment do not all pass immediately out of the catchment. Some eroded materials or sediments may be re-deposited within the plot (Hudson, 1981).

Table 2 shows some selected rainfall events, which caused highest soil losses using the two models, RUSLE and SLEMSA. During these days there was high rainfall which contributed to the high soil loss. Soil loss caused by individual storms for the whole season are given in Appendix III.

Table 2: Selected soil losses (tons/ha/yr.) caused by individual storm events

Based on individual storm calculations, out of these 8 selected events, RUSLE overestimated the values of soil loss approximately 83% of the selected events. These findings agree with Morgan who said that the USLE model was developed to estimate long-term mean annual soil loss such that it can not be used to predict erosion from an individual storm. If applied in this way, it gives an estimate of the average soil loss expected from a number of storms and this may be quite different from the actual soil loss in any single storm (Morgan, 1995).

The overestimation might also be due to the effect of the EI30 index. This index exaggerates the erosivity value using individual storms. Morgan (1979) criticized the use of the EI30 index for assuming that erosion occurs with light intensity rain. Hudson (1965) also showed that erosion is almost entirely caused by rain falling at intensities greater than 25mm/hr. Hence the use of the EI30 index on individual storms overestimates the value of erosivity resulting in high soil loss values. Stocking and Elwell (1973a), however, recommended its use only for bare soil conditions.

SLEMSA underestimated all the amount of soil loss from the individual storm. The low values of soil loss obtained using SLEMSA might be due to the effect of K values calculated using the individual storms which were very small as compared to the rainfall amount SLEMSA was designed for. As a result the low values of K lowered the soil loss using the individual storm. Both models were not suitable for predicting individual storms because these were not intended for that purpose.

Table 3 shows the soil loss computed based on the monthly rainfall using RUSLE and SLEMSA models.

Table 3:Soil loss (tons/ha/yr.) computed using monthly rainfall.

The results indicate that RUSLE overestimated the amount of soil loss from all plots in December and plot 3 for every month as compared to the actual sediment collected. The values of soil loss obtained using SLEMSA were small as compared to the actual values of sediment obtained from the pit. Similar to individual storms, the underestimation in SLEMSA may be due to the effects of rainfall amount on erodibility, K factor. This indeed is in agreement with Morgan (1995) who stated that both models are suitable for prediction of long term soil loss.

Table 4 shows total soil loss computed using RUSLE and SLEMSA models for the whole season (December to March).

Table 4: Seasonal soil loss (tons/ha/yr.)

The results indicate that SLEMSA estimated the values of soil loss very close to the amount of sediment collected from all plots. In plot 3 the crop factor was varying depending on the stage of crop growth thus the effect change in percent cover on soil loss was taken into account. As percent cover was increasing due to the growth of the maize crop, the value of the cover factor was greatly reduced as a result of high energy intercepted. As for plot 2 the value of cover factor was assumed to be constant which is not true in real conditions. Reduction in the Cover factor is expected due to the effects of decomposition of mulch thereby reducing the value of cover factor hence high soil loss. Since the effect of cover variation was not taken into account, it might be the reason for the slight difference (overestimation) with the actual sediment collected. Thus the use of a constant cover factor might not give the right prediction of soil loss.

On the other hand, RUSLE gave a value of soil loss close to the actual sediments for plot 1 though slightly higher than SLEMSA while the value of soil loss for plot 2 was very close to the actual amount of sediments collected. RUSLE however greatly overestimated the amount of soil loss for plot 3. This overestimation of soil loss from plot 3 might be due to the effect of the crop factor that was assumed to be constant throughout the season.

By varying the control variables of each submodel factor a sensitivity analysis on all parameters in SLEMSA model was done. Table 5 shows the results of sensitivity analysis done on plot 2 by increasing each control variable (F, i and S) by 15%.

Table 5: Sensitivity analysis

Figure 4: Sensitivity analysis

From the sensitivity analysis, the most sensitive parameter in SLEMSA is the K factor and mainly the soil erodibility value (F value). A 15% increase in the F value caused a 22.16% decrease in soil loss. This implies that as the value of F value increases, it reduces the K value, which in return reduced the amount of soil loss calculated. Therefore the higher the f value the less erodible the soil is and the vice versa. The second sensitive parameter was found to be the topographic ratio especially the slope steepness where a 15% increase in the slope steepness caused 15% increase change in the soil loss. This implies that as the slope steepness increases, the amount of soil eroded from the land also increases and this might be due to increased runoff velocity.

The primary aim of this research was to evaluate the suitability of RUSLE and SLEMSA models for estimating soil loss under Malawi conditions. The other objective was to identify sensitive factors of soil erosion for SLEMSA model.

The results showed that using the individual storms and monthly rainfall RUSLE overestimated while SLEMSA underestimated the values of soil loss. On the other hand, using seasonal rainfall SLEMSA model gave better predictions of soil loss as well as RUSLE on plot 1 and 2 but greatly overestimated the soil loss on plot 3.

Therefore it can be concluded that both models do not have the potential predictive power of estimating soil loss using individual storms as well as monthly rainfall. But for long-term soil loss both models are suitable. Basic soil erodibility and slope steepness factors are the most sensitive parameters in SLEMSA. should be However it should be noted that crop management factor in RUSLE should be modified in order to give better predictions for land grown to crop.

• Both models can be used to predict long-term soil loss in Malawi.
• A crop management factor in RUSLE should be modified and also the C factors should be developed for different crops.

1. Armstrong C.L, Mitchell J.K and Walker P.N 1980 Soil loss Estimation in Africa -a review. Assessment of Erosion (1980). Dee Boodt & Gabriels (ed.) Illinois. USA.
2. Bennema J. & De Meester T.1980. The role of soil erosion and land degradation in the process of land evaluation. Pp77-86. Assessment of Erosion (1980). Dee Boodt & Gabriels (ed.) Illinois. USA.
3. Elwell H.A.1991. Erosion and sediment yield. Some methods and measurement and modeling. England. Britain.
4. Elwell H.A.1996. Guideline for SADC Region. Environmental monitoring of land degradation and soil erosion methods and techniques
5. Hudson N.W 1971. Soil conservation. Billing and sons. London.
6. Hudson N.W 1981. Soil conservation. Batsford Academic and Educational Ltd. London.
7. Hudson N.W 1985 Soil conservation. New edition. Onchan. Great Britain.
8. Hudson N.W 1990. Erosion prediction with insufficient data. Silsoe. Bedford. England.
9. Hudson N.W 1993 Field measurements of erosion and run off. Soils bulletin 68, FAO Rome. Pp121-127.
10. Hurni H. 1980. Labour intensive soil conservation measures. Pp185-210. Assessment of Erosion (1980). Dee Boodt & Gabriels (ed.) Illinois. USA.
11. Morgan, R.P.C. 1995. Soil erosion and conservation. 2nd ed. Silsoe.
12. Mutchtler C.K. & Murphree C.E.1980. Prediction of soil loss on flatland. Pp321-326. Assessment of Erosion (1980). Dee Boodt & Gabriels (ed.) Illinois. USA.
13. Mwendera, E.J. 1988. Preliminary Evaluation of soil loss estimation model for Southern Africa under Malawi conditions. Lilongwe. Malawi.
14. Paris, S.1990. Erosion Hazard Model (Modified SLEMSA). Land resources evaluation project Malawi. FAO.
15. Schwab, G.O., Frevert, R. K., and Barnes, K.K. 1981. Soil and Water Conservation Engineering 3rd ed. John Wiley, New York.
16. Stalling, J.H., 1957. Soil conservation. Prentice-Hall, inc. Englewood cliffs, N.J. USA.
17. Steichen J. M., Science and Technology Guide. University of Missouri-Columbia Extension Division.
18. Stocking, M.A 1988. Assessing vegetative cover and management effects. Chapter 8. In: Soil erosion reset methods. R.Lal (ed.). Soil and water conservation society. Ankeny. IOWA.
19. Vision 2020 bulletin, 1997.

These computations were done at different levels: - using individual rainfall, monthly rainfall, and seasonal rainfall. On individual rainfall, soil loss caused by each storm was computed and the soil loss values were summed up for each month and for the whole season. The monthly soil loss values were also summed to find for the whole season.

The procedure followed when calculating SLEMSA factors were as follows: -

(1) E was estimated from the daily and monthly rainfall using equation (3) because Kamundi area was identified as a Non-Guti area.

E is given by: -

E = 18.846P (3) for the Non-Guti areas (does not experience significant amounts of low intensity.

Where: P is the rainfall amount

(2) Soil Erodility (F);

This was derived from the original tables of SLEMSA which take into account soil group, soil family, sol texture, and land management factors and crop management e.g. crop rotation. The assumptions made was that soil loss from the previous year were:

Plot 1 over 20 tons/ha/yr.

Plot 2 over 20 tons/ha/yr.

Plot 3 less than 10 tons/ha/yr.

(3) The value of K is determined by relating mean annual soil loss to mean annual rainfall energy (E) using the exponential relationship:

K = exp (0.4681 + 0.7663F) In E + 2.884 - 8.2109F (4)

Where E is in J/mm2

(4) Determination of (i) values:

The i value for maize crop was determined from the crop type, yield, and emergence date and it varied from 43% to 53%. The variety of maize grown was NSCM 41 whose potential yield is 6000kg\ha. The i value for bare plot was estimated at 0% because there was no energy interception. The i value for mulch plot was estimated at 5% because the mulches partially covered the soil surface.

(5) The Crop factor was calculated using either equation (6) or (7) depending on the value of i: -

C = (2.3 - 0.01i) / 30 (5)

(for crops and natural grassland when i is greater than or equal to 50%)

C = e(-0.06i) (7)

(for crops and natural grassland when i less than 50 for dense pastures and mulch when i is greater than or equal to 50%).

Where: i am the energy intercepted.

(6) The value of X is obtained from the following equation: -

X=S (L)^1/2/(10.742S+8.038) for slopes less than 4% (7)

Where: L is slope length S is slope steepness (%)

(7) Then the total soil loss (Z) was determined by multiplying K by C by X

The procedure followed when calculating parameters for RUSLE factors were as follows: -

(1) Using the EI30 index, the R factor is given by:

R = EI30 / 100 (8)

Where: E is the total energy in a storm in ft.tonsf/acre and

I30 is the maximum 30-minute intensity (mm/hr).

The expression of how rainfall energy is related to intensity is given by:

e = 1099(1-0.72exp(-1.27i)) (9)

where: e is the kinetic energy in ft.tonsf/acre in.

i is the average intensity of the storm in in/hr.

Tonsf refer to the force of rainfall impact as opposed to tons mass of sediments being eroded.

In order to convert kinetic energy to total energy of a storm the depth of rainfall P multiplies e.

E = e P (10)

Where: E is the total energy in a storm in ft.tonsf/acre and

P is the total storm depth of rainfall in inches.

(2) The value of soil erodibility K was read from the nomograph

The analytical relationship for the nomograph is based on this equation:

K=(2.1*10-4(12-OM)M1.14+3.25(S1-2)+2.5(P1-3)) /100 (11)

Where: OM is %organic matter

S1 is structure

P1 is permeability

M is a function of the primary particle size fractions given by:

M=(%MS + %VFS)(100-%CL) (12)

Where: MS is %silt

VFS is %very fine sand

CL is %clay

(3)The appropriate values can be obtained from the Nomograph or from the equation

LS = ( L/22.13 )n (0.065+ 0.045S + 0.0065S2) (13)

where LS is the slope length and slope steepness factor

L is the slope length

S is the slope steepness

Soil Physical data for Kamundi