Runoff‐induced sediment yield over dune slopes in the Negev Desert. 2: Texture,carbonate and organic matter

Runoff‐induced sediments were collected in the Hallamish dune field for four years (1990–1994). Runoff and consequently water‐transported sediments were generated on the dunes owing to the presence of a thin microbiotic crust. These sediments were analysed for their particle‐size distribution and carbonate content. In addition, the organic matter content was calculated by measuring the chlorophyll content within the runoff. The results were compared to the slope parent material, i.e. the crust and the underlying sand, as well as to playa sediments, which are scattered within the Hallamish interdunal areas, and which were previously hypothesized to originate from runoff‐induced sediments. Higher amounts of fines (silt and clay) and carbonate characterize the footslopes in comparison to the midslopes. Intermediate contents of fines (17 per cent) and carbonate (8 per cent) characterized the sediments in comparison to the fines (27 per cent) and carbonate (15 per cent) of the crust and to the fines (4 per cent) and carbonate (4 per cent) of the underlying sand. The runoff‐induced fines and carbonate contents were significantly different from those of the playas, suggesting that the playa flats do not originate from runoff‐induced sediments. The sediments were enriched with organic matter. Organic matter which originates from the crust amounted to 0·3–0·4 per cent as compared to ≤0·1 per cent in the bare sand. Nevertheless, the crust was found to be relatively resilient to water flow. Only 0·1–0·5 per cent of the crust was annually eroded off the slope by water, with south‐facing crusts showing higher resilience than north‐facing crusts. The data may thus assist in the evaluation of the crust's residence time. Copyright © 2001 John Wiley & Sons, Ltd.     

Forecasting changes of water resources in large rivers following land drainage / Méthode de prévision des modifications des ressources en eau des grands fleuves sous l'effet du drainage

ABSTRACT

Techniques are described for annual forecasts of the water balance after drainage of large river basins. In the development of these techniques precipitation was assumed to be constant and unaffected by drainage. It is shown that the effect of drainage upon the annual runoff of the improved basins is to decrease the groundwater and swamp water resources which leads to evaporation changes. According to experimental data on the hydrophysical properties of peats, mineral soils and subsoils and how they change after drainage, the decrease in the groundwater resources was estimated for each per cent of the basin drained. This allowed account to betaken of this effect while making forecasts of runoff changes. Evaporation changes are computed as the difference between the maximum possible evaporation (potential evaporation) from cultivated areas and that from undisturbed swamps.     

Soil hydrophobic effects on infiltration and catchment runoff

After dry summers or drought, eucalypt forest soils at two sites in southeastern Australia developed hydrophobic or non-wetting surface characteristics that reduced infiltration, measured using a sprinkling infiltrometer. At one site the development of hydrophobic conditions caused the rainfall to runoff conversion efficiency of a forested catchment to increase from 5 per cent to 15 per cent. Under non-hydrophobic conditions at this site, grassland always generated more runoff than forest. However, one major rainfall-runoff was recorded at a time of highly hydrophobic forest soil conditions and this storm generated greater runoff on the forested catchment than the grassland catchment. At the second site forest soils have naturally highly conductive surface layers because of a dense network of macropores and pathways for preferential flow. Hydrophobic conditions produced by drought caused soil water movement to be confined to only a few of the larger macropores exposed to surface ponded water. Even so, infiltration rates remained relatively high so that the impacts of hydrophobic soils were not translated into increased catchment runoff as at the first site.     

Modelling of water quality processes in lakes and reservoirs. II

ABSTRACT

The snowmelt runoff process from small basins is discussed. A differentiation is made between overland flow in the snowpack and groundwater flow induced by infiltrating meltwater. The effect of variations of the snowmelt intensity on streamflow is studied. It is shown that the runoff is high from the first day of snowmelt runoff if the streamflow is caused by overland flow, and that there are pronounced peaks every day, which almost correspond with the snowmelt intensity during daytime. Streamflow originating from groundwater, on the other hand, increases continuously during the snowmelt and shows only small daily peaks in the flow. Simultaneous overland and groundwater flow are also discussed. Observed runoff hydrographs from small basins are analysed in some detail. For the open fields studied the runoff shows the typical character of overland flow. For a rather large forested area the surface runoff also constitutes an important part of the runoff, but the groundwater baseflow is considerable.     

Runoff generative process and runoff yield from arid talus mantled slopes

Previous works dealing with the influence of a stone cover on runoff yield showed that runoff, attributed to the sealing effect of the topsoil by raindrops impact, was negatively related to the per cent of stone cover and stone size. These works were conducted on gentle slopes (3–11·5 degree) with a per cent of stone cover generally lower than 50 per cent, and composed of gravels. The present study deals with the runoff yield of steep talus slopes (26–36 degree) whose per cent of stone cover is very high (90–100 per cent), composed of cobbles and boulders. Three stimulated rainstorms were performed at various rainfall. intensities and durations on each one of six plots representative of the North eastern sector of Sinai. Although the contiguous stony cover prevented surface sealing by raindrops impact, runoff developed quite quickly on most slopes, and reached at peak discharge, after approximately ten minutes, up to 56 per cent of the rainfall. Differences between plots, in time lag, peak discharge and other hydrological variables, could be attributed mainly to the properties of the upper stony layer, with stone size as the most influential factor. Contrary to previous works, a positive relationship, was obtained between stone size to runoff yield. The result is explained by the process of water concentration. Each cobble and boulder behaves, on a smaller scale, like a bare rocky surface and yields per unit area water amounts beyond the infiltration rate of the limited uncovered areas. For a given stony cover the effect of water concentration is quicker with the big blocks than with gravels. A series of graphs trying to relate theoretically the relative importance of sealing and water concentration processes in runoff generation, at various conditions of stone cover and stone size, is proposed. The graphs enable to reconcile the results of the present study with those of previous works.     

A comparison of stream bank erosion processes on forested and moorland streams in the Balquhidder Catchments,central Scotland

Stream bank erosion rates measured over a two-year period on a moorland and a forested stream in the Institute of Hydrology's Balquhidder Paired Catchments in central Scotland were compared. Bank erosion rates are generally higher on the mainstream of the moorland catchment and highest in wintger on both streams. Bank erosion is correlated with the incidence of frost: minimum temperatures measured on stream banks of the forested stream were an average of 3·7°C higher than on stream banks both outside the forest and on the moorland stream. This makes the incidence of frost on forested stream banks half as frequent. Volumes of material eroded from the mainstreams were combined with bulk density measurements and it is estimated that erosion of the mainstream banks is contributing 1·5 and 7·3 per cent of the sediment yield of the forested and moorland catchments, respectively. Analysis of the vertical distribution of erosion on the banks of both streams suggests an undercutting mechanism which is more pronounced in the moorland stream. The influence of trees on bank erosion and possible implications for the management of forest streams are discussed. © 1997 by John Wiley & Sons, Ltd.     

Paired‐basin comparison of hydrological response in harvested and undisturbed hardwood forests during snowmelt in central Ontario: I. Streamflow,groundwater and flowpath behaviour

Impacts of forest harvesting on groundwater properties, water flowpaths and streamflow response were examined 4 years after the harvest using a paired‐basin approach during the 2001 snowmelt in a northern hardwood landscape in central Ontario. The ability of two metrics of basin topography (Beven and Kirkby's ln(a/tan β) topographic index (TI) and distance to stream channel) to explain intra‐basin variations in groundwater dynamics was also evaluated. Significant relationships between TI and depth to potentiometric surface for shallow groundwater emerged, although the occurrence of these relationships during the melt differed between harvested and control basins, possibly as a result of interbasin differences in upslope area contributing to piezometers used to monitor groundwater behaviour. Transmissivity feedback (rapid streamflow increases as the water table approaches the soil surface) governed streamflow generation in both basins, and the mean threshold depths at which rapid streamflow increases corresponded to small rises in water level were similar for harvested (0·41 ± 0·05 m) and forested (0·38 ± 0·04 m) basins. However, topographic properties provided inconsistent explanations of spatial variations in the relationship between streamflow and depth to water at a given piezometer for both basins. Streamflow from the harvested basin exceeded that from the forested basin during the 2001 melt, and hydrometric and geochemical tracer results indicated greater runoff from the harvested basin via surface and near‐surface pathways. These differences are not solely attributable to harvesting, since the difference in spring runoff from the harvested basin relative to the forested control was not consistently larger than under pre‐harvest conditions. Nevertheless, greater melt rates following harvesting appear to have increased the proportion of water delivery to the stream channel via surface and near‐surface pathways. Copyright © 2005 John Wiley & Sons, Ltd.     

The quantity of stormwater runoff from ten stretches of road,a car park and eight roofs in Hertfordshire,England during 1983

Rainfall and runoff were monitored simultaneously for one year from a residential road, a car park, nine sections of road draining to individual gullies, two house roofs, two garage roofs, and three types of factory roof. The sites, which included an automatic weather station, were in Redbourn, Hertfordshire on Flood Studies Report Soil Type 1. The 2906 quality controlled ‘station-storms’ represented 193 rain storms and involved 57.2 per cent of the annual rainfall. 1732 storms were of less than 1.4mm of rain, whilst 77 had over 10mm. The percentage runoff averaged 11.4 per cent for roads and 56.9 per cent for roofs (28.3 per cent and 90.4 per cent respectively for rainfalls >5mm). Percentage runoff from the roads was cyclic with a peak during the summer months but there was a marked variation in monthly percentage runoff within and between sites. Regression analysis to explain percentage runoff was undertaken with various subsets of data for: each site; roads; and roofs. The regression analysis considered all storms; >1 percent runoff events; >5mm rainfalls; and events with > = 4 mm rain and > = 5 per cent runoff. The variable values in percentage runoff could not be explained satisfactorily with statistical methods. Only eight of the 72 equations explained more than 57 per cent of the variance. The most important explanatory variables for roads were short term rainfall intensity and rainfall amount, the former was the most important for roofs. ‘Seasonal’ variables had a positive relation ship for roads which shows that the percentage runoff from roads is higher in summer than winter. The antecedent variables showed that percentage runoff from roads and roofs is increased by antecedent rainfall. Seasonal factors and evaporation were unimportant for the percentage runoff from roofs. Depression storage, assessed by examining rainfalls that did and did not produce runoff, showed a diversity of monthly values. The depression storages derived by the regression intercept method were usually smaller. There were no relationships between depression storage and catchment or roof slope. The mean values for roofs and roads respectively were 0.52 mm and 1.23 mm for the classification method and 0.42 mm and 0.6mm with the regression approach. Peak runoff from the roads showed an attenuation to 12.8 per cent for 1 minute rainfall intensities and 24.2 per cent for 5 minute intensities. For roofs the attenuation averaged 36.8 per cent for 1 minute intensities and 92.6 for 5 minute intensities. Regression for peak runoff coefficients from roofs and roads explained negligible amounts of the variance except when events with 1 minute rainfall intensities of over 30 mm hr^?1 over the roads were analysed. Total rainfall was an important explanatory variable as was the slope of the road. There was evidence that peak coefficients for roads are greater during the summer.     

Rates and mechanisms of discontinuous gully erosion in a red-brown earth catchment,new south wales,australia

Rainfall, runoff, sediment discharge, and gully expansion were measured for five years in a small (45 ha) catchment in central N.S.W. Sixty per cent of the sediment produced from this catchment originated from gully erosion. Gully head erosion was episodic depending on pipe development, cracking, and soil detachment during small runoff events which prepared the heads for rapid soil movement by the infrequent large event.     

Conditions for generation of fire‐related debris flows,Capulin Canyon,New Mexico

Comparison of the responses of three drainage basins burned by the Dome fire of 1996 in New Mexico is used to identify the hillslope, channel and fire characteristics that indicate a susceptibility specifically to wildfire‐related debris flow. Summer thunderstorms generated three distinct erosive responses from each of three basins. The Capulin Canyon basin showed widespread erosive sheetwash and rilling from hillslopes, and severe flooding occurred in the channel; the North Tributary basin exhibited extensive erosion of the mineral soil to a depth of 5 cm and downslope movement of up to boulder‐sized material, and at least one debris flow occurred in the channel; negligible surface runoff was observed in the South Tributary basin. The negligible surface runoff observed in the South Tributary basin is attributed to the limited extent and severity of the fire in that basin. The factors that best distinguish between debris‐flow producing and flood‐producing drainages are drainage basin morphology and lithology. A rugged drainage basin morphology, an average 12 per cent channel gradient, and steep, rough hillslopes coupled with colluvium and soil weathered from volcaniclastic and volcanic rocks promoted the generation of debris flows. A less rugged basin morphology, an average gradient of 5 per cent, and long, smooth slopes mantled with pumice promoted flooding. Flood and debris‐flow responses were produced without the presence of water‐repellent soils. The continuity and severity of the burn mosaic, the condition of the riparian vegetation, the condition of the fibrous root mat, accumulations of dry ravel and colluvial material in the channel and on hillslopes, and past debris‐flow activity, appeared to have little bearing on the distinctive responses of the basins. Published in 2000 by John Wiley & Sons, Ltd.     

Effects of wildfire and logging on streamwater chemistry and cation exports of small forested catchments in Southeastern New South Wales,Australia

In January 1979 four experimental catchments in forests of southeastern New South Wales were burnt by wildfire. Logging before the fire had no detectable effect on concentrations of Ca, K, Mg, Na, Cl, NO₃, nor on pH of stream waters. In all burnt catchments mean K concentrations increased by from 20 per cent to 60 per cent for a 12 month period and nitrate concentrations increased by factors of about ten in severely burnt catchments. In one of the catchments (unlogged) Ca also increased. From one year to four years after the fire, concentrations of all ions were either close to or less than levels predicted from the control but, during the fifth and sixth years, concentrations of Mg and Na were higher by 20 per cent to 60 per cent. In all burnt catchments, cation exports increased considerably during the first three years after the fire but major components of these increases were elevated levels of runoff. Exports of Mg and Na were higher than those of the control during the fifth and sixth years after the fire, although runoff had returned to pre-fire levels in the two unlogged catchments and was 10 per cent to 20 per cent greater than the control in the two logged catchments. During this final period, increased ion concentrations were the main factors which contributed to the elevated exports. Post-fire logging in one catchment had no detectable effects on streamwater parameters measured in the study but was associated with a further increase in runoff.     

Slopewash,surface runoff and fine-litter transport in forest and landslide scars in humid-tropical steeplands,luquillo experimental forest,Puerto Rico

Rainfall, slopewash (the erosion of soil particles), surface runoff and fine-litter transport steepland sites in the Luquillo Experimental Forest, Puerto Rico (18° 20’ N, 65° 45’ W) were measured from 1991 to 1995. Hillslopes underlain by (1) Cretaceous tuffaceous sandstone and silstone in subtropical rain (tanonuco) forest with vegetation recovering from Hurricane Hugo (1989), and (2) Tertiary quartz diorite in subtropical lower mantone wet (colorado and dwarf) forest with undisturbed forest canopy were compared to recent landslide scars. Monthly surface runoff on these very steep hillslopes (24° to 43°) was only 0·2 to 0·5 per cent of monthly rainfall. Slopewash was higher in sandy loam soils whose parent material is quartz diorite (averaging 46 g m⁻² a⁻¹) than in silty clay loam soils derived from tuffaceous sandstone and siltstone where the average was 9 g m⁻² a⁻¹. Annual slopewash of 100 to 349 g m⁻² on the surfaces of two recent, small landslide scars was measured initially but slopewash decreased to only 3 to 4 g m⁻² a⁻¹ by the end of the study. The mean annual mass of fine litter (mainly leaves and twigs) transported downslope at the forested sites ranged from 5 to 8 g m⁻² and was lower at the tabonuco forest site, where post-Hurricane Hugo recovery is still in progress. Mean annual fine-litter transport was 2·5 g m⁻² on the two landslide scars. Copyright © 1999 John Wiley & Sons, Ltd.     

Soil and onsite nutrient conservation potential of aromatic grasses at field scale under a shifting cultivated,degraded catchment in Eastern Ghats,India

Land degradation due to soil erosion is a global problem, especially on cultivated hill slopes. Economically important aromatic grasses can protect degraded hill slopes more effectively than field crops, but little information is available on their performance. This study quantifies runoff, sediment yield,enrichment ratios of soil and nutrients, and sediment-associated organic carbon and nutrients losses under three aromatic grass species: citronella(Cymbopogon nardus), lemon(Cymbopogon flexuosus), and palmarosa(Cymbopogon martini), compared with a traditional field crop, finger millet(Eleusine coracana)grown at three land slopes(4%, 8%, and 12%). It was observed that the degree of slope and type of grass both significantly influenced runoff generation. Runoff and sediment yield(SY) were significantly higher at 12% slope than at 8% and 4% slopes. Relation between rainfall and runoff were significant for all the grass species(p 0.05). Palmarosa, lemon, and citronella grass reduced the SY by 10, 54, and 60%,respectively, over finger millet. SY was also significantly related to rainfall for all the treatments(p 0.05). The threshold runoff values to produce SY were higher for aromatic grasses compared to finger millet. Enrichment of clay, silt, sand, soil organic carbon(SOC), available nitrogen(N), phosphorus(P) and potassium(K) in the sediment were not significantly different between slopes but differed significantly between aromatic grasses and finger millet. Sediment associated nutrient load varied inversely with SY mainly because of the nutrient dissolution effect of high runoff volume. Annual loss of SOC and nutrients varied from 84.7-156.8 kg ha~(-1) y~(-1) for SOC, 4.38-9.18 kg ha~(-1) y~(-1) for available N, 0.35-0.75 kg ha~(-1) y~(-1) for available P, and 2.22-5.22 kg ha~(-1) y~(-1) for available K, with the lowest values for citronella and highest for finger millet. The study found that the aromatic grasses have greater environmental conservation values than finger millet on steep degraded land.     

Analysis of runoff pathways and flow contributions using deuterium and stream chemistry

The two-component hydrograph separation method has been used to investigate the sources of storm runoff at Allt a Mharcaidh during storms in October 1987, September 1988, and June 1989. Results show that for the September and October storms, more than 88 per cent of storm runoff was contributed by pre-event water and 54 per cent of the June storm runoff. Instantaneous contributions of pre-event water at the hydrograph peak vary from 91 per cent in September 1988 to 41 per cent in June 1989. Detailed examination of soil and borehole water, however, shows that these subcomponents of pre-event water do not always have similar chemistry and isotope concentrations. Therefore, information from the two-component separation method alone should not be used to infer processes or pathways, although some broad indications can be obtained through the additional use of soil water and groundwater chemistry.     

Factors controlling sediment yield in China's Loess Plateau

The Loess Plateau in China, an area with some of the highest sediment yield in the world, contributes predominant proportion of the sediments found in the Yellow River. We examined sediment yield and its control variables in the plateau based on a multi‐year dataset from 180 gauging stations in areas varying in size from 10² to 10⁴ km². Various morphometric, hydrologic, climatic and land cover variables were estimated in order to understand and predict the variations in sediment yield. The results show a spatial pattern of sediment yield exhibiting an obvious zonal distribution and a coupling between precipitation and vegetation cover that fits the Langbein–Schumm law. A critical threshold of precipitation and vegetation cover was observed among the relationships of sediment yield and precipitation/vegetation cover. A multiple regression equation with three control variables, i.e. vegetation cover, percentage of cultivated loess and annual runoff, explains 65% of the total variation in sediment yield. For the loess dominated basins, where the cultivated loess accounts for more than 60% of the total area, annual runoff was the dominant variable, explaining 76% of the observed variation in sediment yield. The established equation could be a valuable tool for predicting total sediment yield in the Loess Plateau. Copyright © 2010 John Wiley & Sons, Ltd.     

Developing a sediment budget for a small drainage basin in Australia

A sediment budget was developed for the 1.7 km² Maluna Creek drainage basin located in the Hunter Valley, New South Wales, Australia, for the period 1971-86. the impact of viticulture, which commenced at Maluna in 1971, was studied using erosion plots, with caesium-137 as an indicator of both soil erosion and sedimentation. Two methods were used to estimate vineyard soil losses from caesium-137 measurements. Sediment output from the catchment was measured for three years, and extrapolated from readings taken at a nearby long-term stream flow gauging station for the remaining 13 years. Relative amounts of soil loss from forest (60 per cent basin area), grazing land (30 per cent) and vineyards (10 per cent) were calculated. Soil losses by rain splash detachment were ten times greater from bare/cultivated sufaces than from the forest. Erosion plots of area 2 m² showed no significant differences in soil loss between forest and grassland but, under bare soil, losses were 100 times greater. the ¹³⁷Cs method was employed to calculate net soil loss from all vineyard blocks using both a previously established calibration curve and a proportional model. the latter method gave estimates of soil loss which were 3-9 times greater than by the calibration curve, and indicated that average soil losses from the vineyard were equivalent to 62 t ha^?1 y^?1 (1971-86). It was estimated that the forest contributed 1-8 per cent, the grazing land 1.6 per cent, and the vineyard 96.6 per cent of the total soil loss during that period. Sediment storages within the fluvial system adjacent to the vineyard ws 9460 t for the period, whereas sediment output was equivalent to 215 t km^?1 y^?1. Independent measurements of soil erosion, storage, and output showed that 56 per cent of the eroded sediment remained in the catchment, and 34 per cent was transported out by Maluna Creek. the budget was able to be balanced to within 10 per cent.     

Vegetation influences on water yields from grassland and shrubland ecosystems in the Chihuahuan Desert

This study examines runoff generated under simulated rainfall on Summerford bajada in the Jornada Basin, New Mexico, USA. Forty‐five simulation experiments were conducted on 1 m² and 2 m² runoff plots on grassland, degraded grassland, shrub and intershrub environments located in grassland and shrubland communities. Average hydrographs generated for each environment show that runoff originates earlier on the vegetated plots than on the unvegetated plots. This early generation of runoff is attributed to soil infiltration rates being overwhelmed by the rapid concentration of water at the base of plants by stemflow. Hydrographs from the degraded grassland and intershrub plots rise continuously throughout the 30 min simulation events indicating that these plots do not achieve equilibrium runoff. This continuously rising form is attributed to the progressive development of raindrop‐induced surface seals. Most grassland and shrub plots level out after the initial early rise indicating equilibrium runoff is achieved. Some shrub plots, however, display a decline in discharge after the early rise. The delayed infiltration of water into macropores beneath shrubs with vegetation in their understories is proposed to explain this declining form. Water yields predicted at the community level indicate that the shrubland sheds 150 per cent more water for a given storm event than the grassland. Copyright © 2002 John Wiley & Sons, Ltd.     

Limestone solution rates and processes in the Waitomo District,New Zealand

Karst solution processes are investigated on Oligocene limestones in the Waitomo district, west central North Island, New Zealand. Estimates of the inputs, throughputs and outputs of water and dissolved calcium and magnesium in two drainage basins were used to establish the rate of limestone solution by autogenic waters. The best estimate for solution loss from the basins during the study year is 69 m³/km². The potential measurement errors inherent in each parameter used in the erosion rate computations were assessed and the probable maximum and minimum erosion rates were estimated to be 88 and 61 m³/km². In both basins approximately 67 per cent of the annual solute load is transported by flows greater than the mean annual discharge, over 15 per cent being transported by flood flows that are exceeded only 5 per cent of the time. Almost half of the annual load is transported during the three winter months (June-August), but no one month accounts for more than 18 per cent or less than 2.7 per cent of the annual total. Approximately 37 per cent of solution takes place within the soil profile, and most of the remainder is concentrated in 5–10 m of weathered bedrock (the subcutaneous zone) beneath this. Thus, it is likely that at least 85 per cent of the total solutional erosion contributes to the surface lowering of soil and bedrock.     

An evaluation of the universal soil loss equation and field techniques for assessing soil erosion on a tropical alfisol in western Nigeria

Four techniques for soil erosion assessment were compared over two consecutive seasons for bare-fallow plots and a maize-cowpea sequence in 1985 at IITA, Ibadan, Nigeria. The techniques used were: tracer (aluminium paint), nails (16 and 25), the rill method, and the Universal Soil Loss Equation (USLE). Soil loss estimated by these techniques was compared with that determined using the runoff plot technique. There was significantly more soil loss (P < 0·01) in bare-fallow than in plots under maize (Zea mays) or cowpea (Vigna unguiculata). In the first season, soil loss from plots sown to maize was 40·2 Mg ha^?1 compared with 153·3 Mg ha^?1 from bare-fallow plots. In the second season, bare-fallow plots lost 87·5 Mg ha^?1 against 39·4 Mg ha^?1 lost from plots growing cowpea. The techniques used for assessing erosion had no influence on the magnitude of soil erosion and did not interfere with the processes of erosion. There was no significant difference (P < 0·05) between soil erosion determined by the nails and the runoff plot technique. Soil loss determined on six plots (three under maize, three bare-fallow) by the rill technique, at the end of the season, was significantly lower (P < 0·05) than that determined by the runoff plot technique. The soil loss estimated by the rill method was 143·2, 108·8 and 121·9 Mg ha^?1 for 11, 11, and 8 per cent slopes respectively, in comparison with 201·5, 162·0, and 166·4 Mg ha^?1 measured by the runoff plot method. Soil loss measured on three bare-fallow plots on 10 different dates by the rill technique was also significantly lower (P < 0·01) than that measured by the runoff plot. In the first season the USLE significantly underestimated soil loss. On 11, 11, and 8 per cent slopes, respectively, soil loss determined by the USLE was 77, 92, and 63 per cent of that measured by the runoff plot. However, in the second season there was no significant difference between soil loss determined by the USLE and that determined by the conventional runoff plot technique.