256 AE. GILLEY ET AL. Shear Stress (Pa) 8000 coos onto 001s ona ons on ‘Sediment Detachment Capacity dopa ajqussodsyp A210, Joo prepung: 7 sjwapnig Jone paspung J syuopmg dé ns —_— ‘uorwunuoyep 9 @ a Jo qwa1oyjaog, ‘uoHenbs uojssasdoy son]wa ssoms soys JeonuD [ENIDe snsiaA payoipord Jo sasK[EUR [eONSTEIS vy aTavLINITIATION OF RILLING 261 005 © Sol Water Content at 0:3 MPa 223.0% 4 Sol Wer Content a1 0.3 Pa > 23.0% 004 b - — Une ot Pecect Agreement 003 02 Predicted Rill Erodibility Factor (s/m) 000. ‘a0 ‘002 (003 ‘o08 005 Actual Rill Erodibility Factor (s/m) Fig, 5, Predicted vs, actual rill erodibility factors, The regression coefficients shown in eqn. (9) were significantly different from zero at the 95% confidence level. Soil survey data and eqns. (8) and (9) were used to calculate values of K, for cach of the experimental soils. Results of the analyses are presented in Fig, 4, Predicted and actual values of K, shown in Fig. 5 can be seen to be similar. Linear regression analyses were used to compare predicted and actual values of K,, Results of the statistical analyses are presented in Table 5. Equations (8) and (9) produced coefficient of determination values of 0.915 and 0,865, respectively, ‘The hypotheses that the regression coefficients shown in Table 5 equal unity and the intercepts equal zero were evaluated at the 95% confidence level using the students s-test. The slopes were not significantly different from unity nor were the intercepts significantly different from zero. Thus, analyses of the experimental data suggests that eqns. (8) and (9) can be used to estimate rill soil erodibility factors. CRITICAL FLOW RATES. Hdentifying critical flow rates Flow rate versus slope-adjusted detachment capacity for the Barnes soil in North Dakota is shown in Fig. 6. Results obtained for this soil are also representative of the other experimental locations. It can be seen from Fig. 6 that the relationship between flow rate and slope-adjusted detachmentTE. GILLEY ET AL, 262 ‘vol url 960°0 tl S980 TL + Fenaoe 998°9 = paroipag HOE < PAW ED 10 1aIH0> so1DM pos eel L680. 6100 68L0— Ort $160 PUL + fenoe gee'g = pororpary MOE S LEH £0 1 twaIN0d soIDH Yos Jous prpung =; uepmg —foLe prepueIS 7 sqtIOpME. 2 ‘wonreuruuoiep t ‘o a Jo TURK UFB0D, woHTenba woyssautoy *y yuorp4yaoo worssouax jomoe “sa parotpaad Jo soskjeue jwonsnels $ 2TaVLINITIATION OF RILLING 263 a5 16 15 20 Slope Adjusted Detachment Capacity (Kalmés) Fig. 6, Flow rate vs. slope adjusted sediment detachment capacity for a Barnes soil in North Dakota. capacity can be well represented by a linear equation. Critical flow rate for initiation of rilling for the Barnes soil in North Dakota was 0.1731s~'. Linear regression analyses were used to identify the critical flow rate values shown in Table 6. Flow rates required to initiate rilling varied widely between study locations. Critical flow rates ranged from 0.002471s~' for the Miami soil to 0.2171"! for the Pierre soil. Relating critical flow rates to soil properties Critical flow rate values identified in Table 6 were related to soil properties presented in Table 2 using step-wise multiple regression analyses. Statistical analyses suggested that critical flow rates were significantly correlated to silt content. For sites with silt content of 37.0% or less, Q, = 0.222 (aluminum) + 0.0171 (cation exchange capacity) ~ 0.0223 (magnesium) — 0.00984 (soil water content at 0.3MPa) + 0.00204 (very fine sand) + 0.108 (10) where aluminum concentration, soil water content at 0.3 MPa and very fine sand are given as a percentage, and cation exchange capacity and magnesium. content are in centimoles per kilogram. The regression coefficients shown in sgn. (10) were significantly different from zero at the 90% confidence level. For sites with silt content greater than 37.0% Q, = 0.00373 (calcium) + 0.0189 (magnesium) +0.0966 (organic carbon) — 0.0537 (potassium) +0.00462 (soil water content at 0.3 MPa) — 0.130 ay264 JE, GILLEY ET AL. TABLE 6 Statistical analyses of equations used to relate rill low rate to detachment capacity Soil oe Ke Coefficient of determination, as") erm) e 0.177 9,23 0.658 0,164 29.3 0.613 0.152 6.67 0.704 0.173 3.57 0.754 0.198 1.22 0.624 0.0767 3.28 0.693 0.0747, 7.07 0.809 0.0912 1.63 0.709 0.143 496 0657 0.0947 836 0614 1.500 97 0.638 0.00432 6 0.731 0.110 4.55 0.752 0.122 4.76 0.855 0.132 2.31 0.678 0.00247 642 0.730 0.0955 480 0.601 0.106 16.2 0.615 0.147 245 0.639 0.115 6.00 0.731 0217 B8 0.609 0.140 133 0.669 Sharpsburg 0.153 3.53 0734 Sverdrup 0.166 10.5 0.612 Tifton 0.125 4.58 0.703 Whitney 0.0880 9.90 0.647 Williams 0.203 3.70 0.629 Woodward 0.0602 761 0.653 Zahl 0.184 12.3 0.640 “Regression coefficients Q. and KX; are used in the equation -1(2 ont(B) +0 where fow rate and detachment capacity are in liters per second and kilograms per square meter per second, respectively, and slope is given as a fraction.INITIATION OF RILLING 265 925 © sines70% A S237 — Uno af Pereet Agroomant Prodiicted Critical Flow Rate (L’s) 2 2 e 3 & 8 8 005 210 O15 020 025 ‘Actual Critical Flow Rate (L/s) ig, 7, Predicted vs, actual critical flow rates, where calcium, magnesium, and potassium are in centimoles per kilogram, and organic carbon and soil water content at 0.3MPa are given as a percentage, All the regression coefficients shown in eqn. (11) were significantly different from zero at the 90% confidence level. Values of Q, were calculated for each of the experimental soils using soil survey data and eqns. (10) and (11). Results of the analyses are shown in Fig. 7, Predicted and actual values of Q, shown in Fig. 7 can be seen to be similar. Linear regression analyses were used to compare predicted and actual values of Q,. Results of the statistical analyses are shown in Table 7. Coeffi- cient of determination values of 0.878 and 0.882 were found for eqns. (10) and (11), respectively. ‘The Students t-test was used to evaluate the hypotheses that the regression coefficients shown in Table 7 equal unity and the intercepts equal zero at the 95% confidence level. The slopes were not significantly different from unity nor were the intercepts significantly different from zero. Thus, analyses of the experimental data suggests that eqns. (10) and (11) can be used to estimate Q,. FLOW-RELATED RILL SOIL ERODIBILITY FACTORS Identifying flow-related rill soil erodibility factors Figure 6 shows flow rate versus slope-adjusted detachment capacity for the Barnes soil in North Dakota. The flow-related rill soil erodibility factor is represented by the inverse of the slope of the regression line presented in Fig. 6, For the Barnes soil in North Dakota, the slope of the regression line wasJB. GILLEY ET AL. 266 z100 ol £60°0 sI- 06 288'0——E10'0 + Jenne gg'9 = porsiparg MOLE < INS £100 wl 1600 weI- £6 8480 910°0 + reTMoe Ogg’ = porDIporg HOLE > INS ous paepuyg 2 syuspnig Jol prepuers: 4a sjuepmg d ‘worreuyeuy0p og iy a JO WIsTOWFeOD, uonenba uossordoy sonjea ayer MOY [DHL [enaH “sa payorpard Jo sesdpoue FeONSNEIS LaTavLINITIATION OF RILLING 267 0.280, which corresponds to a flow-related rill soil erodibility factor of 3.57 kgh-'m™?, Flow-related rill soil erodibility factors were determined for each of the other study locations using linear regression analyses. Results of the regression analyses are shown in Table 6. Flow-related rill soil erodibility factors were found to vary from 1.22kg]~'m~? for the Caribou soil to 29.3kg1~'m~? for the Amarillo soil. Relating flow-related rill soil erodibility factors to soil properties Step-wise multiple regression analysis was used to relate the flow-related rill soil erodibility factors identified in Table 6 to soil properties presented in Table 2, The flow-related rill soil erodibility factors were found to be signifi- cantly correlated to sand content. For sites with sand content of 40.0% or less, K, = ~—245 (coefficient of linear extensibility) + 0.971 (clay) +-0.336 (silt) + 0.275 (soil water content at 0.3 MPa) + 1.04 (very fine sand) — 43,9 (12) where coefficient of linear extensibility is in centimeters per centimeter and clay, silt, soil water content at 0.3 MPa, and very fine sand are given as a percentage, All the regression coefficients shown in eqn. (12) were significantly different from zero at the 90% confidence level. For sites with sand content greater than 40.0%, K, = 329 (coefficient of linear extensibility) +- 76.3 (sodium) = 20.3 (potassium) — 0.743 (water dispersible clay) + 11.7 (13) where coefficient of linear extensibility is in centimeters per centimeter, sodium and potassium are in centimoles per kilogram, and water dispersible clay is given as a percentage. All the regression coefficients shown in eqn. (13) were significantly different from zero at the 90% confidence level. Values of K; were calculated for each of the experimental soils using soil survey data and eqns, (12) and (13). Results of the analyses are shown in Fig. 8, Predicted and actual values of X; shown in Fig. 8 can be seen to be similar. Linear regression analyses were used to compare predicted and actual values of K;. Results of the statistical analyses are presented in Table 8. Equations (12) and (13) produced coefficient of determination values of 0.906 and 0.837, respectively. ‘The Students /-test was used to evaluate the hypotheses that the regression coefficients shown in Table 8 equal unity and the intercepts equal zero at the 95% confidence level. The slopes were not significantly different from unity268 4UE. GILLEY ET AL. 8 Predicted Rill Erocibility Factor (kg/L-m*) 25 30 5 10 16 20 Actual Rill Erodibility Factor (kg/L-m) Fig. 8, Predicted vs, actual flow related rill erodibility factors. nor were the intercepts significantly different from zero. Thus, analyses of the experimental data suggests that eqns. (12) and (13) can be used to estimate flow-related rill soil erodibility factors. LIMITATIONS OF THE REGRESSION EQUATIONS Although reasonable estimates of critical shear stress and critical flow rates were provided by the regression equations, other factors may limit their application in field situations. For example, crop residues and rock fragments were absent from the sites where the experimental data used to parameterize the regression equations were obtained. Crop residue cover could serve to protect rill and interrill areas from soil detachment. Soil detachment may also decrease as channel armoring develops, if soils contain rock fragments. Critical shear stress and critical flow rates for sites where crop residue or rock fragments are present may be substantially larger than estimates obtained using equations presented in this study. Temporal variations in critical shear stress and critical flow rates may also occur, These differences have been attributed to changes in soil cohesion, rainfall-induced soil consolidation, and development of root fabric. Critical shear stress and critical flow rates identified in this study are strictly applicable only for conditions immediately following tillage. At present, procedures for estimating temporal effects on critical shear stress and critical flow rates have not been identified. Values of K, and K; were determined for bare unconsolidated soil conditions existing immediately after tillage. Soil erodibility may be affectedINITIATION OF RILLING vrE el toro wie Oo L€8'0 Ty'l + lems geg'o = paripard mO'Ob < puns Or9'0 9160 180° svi Stl 9060 $790 + [NIE 906 = PRPAA 00 > pos sous prepunig 7 siwopmg JOU prepuerg 7 sywepng: Zé ‘aopeupurap of y a Jo waryjeod uopenbs worssardoy, ‘8107084 Azifiqiposo [os [I Pareas MOY yeMDe “SA parorpeud Jo sesspeue TeORSHEIS saTavl270 4B. GILLEY ET AL. by soil consolidation, below-ground residue, and freeze-thaw conditions. Methods for calculating the effects of soil consolidation on erodibility were identified by Nearing et al. (1988). Brown et al. (1989) developed relationships for adjusting rill erodibility for below-ground residue. This study was conducted on cropland soils where surface residue had been removed and the areas were maintained free of vegetation for several months. The sites were tilled immediately before testing and were, therefore, in a highly erosive condition. The regression equations developed in this study should not be applied to pasture or rangeland areas with much different soil and vegetative characteristics. SUMMARY AND CONCLUSIONS Hydraulic conditions required to initiate rilling were identified in this study. Linear equations were used to relate shear stress values and flow rates to rill sediment detachment capacity. The intercept of the linear equations provided estimates of critical shear stress and critical flow rate. Runoff and erosion data collected on soils located throughout the USA were used to determine critical shear stress values and critical flow rates. The experimental sites were selected to cover a broad range of soil properties. On cach of the sites, crop residue had been removed, moldboard plowing and disking had occurred, and preformed rills had been constructed. Selected soil properties were measured at each of the experimental locations. Critical shear stress values and critical flow rates were related to site-specific soil properties using multiple regression analysis. Close agreement was found between predicted and actual values. Process-based models for predicting runoff and erosion on upland areas require information on flow hydraulics and soil erodibility. Procedures for estimating hydraulic conditions required to create rills were identified in this investigation. This information will improve our ability to understand and properly model upland runoff and erosion processes. ACKNOWLEDGMENTS This paper is a contribution from USDA-ARS, in cooperation with the Agricultural Research Division, University of Nebraska, Lincoln, and is published as Journal Series No. 9647. REFERENCES Brown, L.C., Foster, G.R. and Beasley, D.B., 1989. Rill erosion as affected by incorporated crop residue and seasonal consolidation. Trans. ASAE, 32(6): 1967-1978.INITIATION OF RILLING m Elliot, W.J., Licbenow, A.M., Lafien, ILM. and Kobl, K.D., 1989. A compendium of soil erodibility data from WEPP cropland soil field erodibility experiments 1987 and 88. NSERL Rep. No. 3, USDA-ARS, West Lafayette, IN, 317 pp. Foster, G.R., 1982. Madeling the erosion process. In: C.T. Haan, H.P. Johnson and D.L. Brakensick (Editors), Hydrologic Modeling of Small Watersheds. American Society of Agricultural Engineers, St. Joseph, MJ, pp. 297-380. Laursen, EM., 1958. The total sediment load of streams. J. Hydraul. Div., ASCE 123(1): 195-206, Nearing, M.A., West, L-T. and Brown, L.C., 1988. A consolidation model for estimating changes in rill erodibitity. Trans, ASAE, 34(3): 696-700. Nearing, M.A., Foster, G.R., Lane, LJ. and Finkner, $.C., 1989. A process-based soil erosion model for USDA-Water Erosion Prediction Project technology. Trans. ASAE, 32(5): 1587— 1593. Schoklitsch, A., 1957. River bed degradation below large capacity reservoir, Trans. ASCE, 122: 695. Soil Survey Staff, 1984. Procedures for collecting soil samples and methods of analysis for soil survey, USDA-SCS Soil Survey Investigations Rep. No. 1., U.S. Government Printing Office, Washington, DC. Swanson, N.P., 1965. Rotating-boom rainfall simulator. Trans. ASAE, 8(1): 71-72.Journal of Hydrology, 142 (1993) 273-299 273 Elsevier Science Publishers B.V., Amsterdam on Changes in the hydrological response of a small Mediterranean basin a year after a wildfire Jacques Lavabre', Daniel Sempere Torres* and Flavy Cernesson* °CEMAGREF, Le Tholonet B.P., 31 F-13612 Aix-en-Provence, France *Untversitat Politeentca de Catalunya, Dpt. Enginyeria Hidraulica i Hidrélogica, Gran Capita sin MODUL DI, E-08034, Barcelona, Spain (Received 18 February 1992; revision accepted 8 Jume 1992) ABSTRACT Lavabre, J., Sempere Torres, D. and Carnesson, F., 1993. Changes in the hydrological response of a small Moditerrancan basin 2 year after a wildfie. J, Hydrol., 142: 273-299. ‘This paper presents the results of a study of the hydrological consequences of a witdfire in a smell Mediterranean basin during the first 12 months after the event. The Rimbaud basin is a nested sub-basin of the Réal Collobrier cesearch and experimental network, which has been continuously monitored since 1967, In August 1990, 85% of its surface area was burnt providing an interesting case study to analyse the Inydrotogical consequences of fire. Models have been calibratéd using a 23 year period before the fre and they have been used to simulate what the response would have been if the wildfire had not taken place. Differences between observed and simulated data are analysed, giving special importance to the changes in the annual runoff yield and in the flood regimes. There is a 30% increase in the annual runoff yield related to the reduction in evapotrat lion due to the destruction of the vegetation cover, On the other hand, there are pronounced changes in the shape of the flood hydrographs, and the flood frequency is greatly increased, The 10-year return period flow estimated before the fire was exceeded three times in the year although the rainfall events did not exceed the L-year return period 12h rainfall. The quantity and quality of the data collected during the pre-fre period and the possibilty of comparing the yydrological responses of similar basins differently affected by fire, give a special relevance to this case study. INTRODUCTION It is widely known and accepted that the destruction of the forested ecosystem of a basin has important consequences for its hydrological behaviour, On the one hand, the disappearance of the vegetal cover directly modifies the processes of interception and evapotranspiration, seriously affecting the hydrological cycle. On the other hand, forest fires can also affect hydrological processes indirectly, but profoundly, altering the physical and Correspondence to: J. Lavabre, CEMAGREF, Le Tholonet B.P., 31 F-13612 Aix-en-Provence, France, 0022-1694/93/S06.00 © 1993 — Elsevier Science Publishers B.V. All rights reserved274 2. LAVABRE ET AL. chemical properties of the soil, converting organic ground cover to soluble ash, modifying the microclimate through removal of overhead foliage, or giving rise to phenomena such as water repellency (Chandler et al., 1983). However, even if the genera] consequences of fire have been largely studied, in particular the ecological and biological impact and the associated soil alteration, there are relatively few studies evaluating the effect of these changes on the hydrological response (see reviews by Anderson et al., 1976, and Tiedemann et al., 1979). This relative scarcity of quantitative hydrological studies merely reflects the difficulty of obtaining good quality data to compare hydrological behaviour before and after a fire, Scientific interest in a burnt area generally starts only after the fire, and therefore it is very difficult to find previous studies and pre-fire data to compare. A possible solution, often used in fire studies, consists of monitoring two twin basins: the first preserved in its original state, the second artificially burnt, However, even if this strategy may be interestingly used in other fields, the hydrological comparison between twinned basins (supposed hydrologic- ally equivalent a priori) is very often difficult and the consequences are usually dubious, Of course, a much better strategy is to monitor an experimental basin, collect data over a Jong period of time (at least ten years), long enough for the characteristics of its hydrological behaviour to become clear, and then produce an experimental and controlled fire. Tn our case a fortuitous wildfire in an area we had been studying for 23 years gave us the chance to study the hydrological effects of fire. In August 1990 a set of Mediterranean experimental research basins which had been studied since 1966 were partially burnt. The consequences of the wildfire, especially the destruction of the vegetation cover and the alteration of soil, were the only changes produced. Thus the comparison between the pre-fire and the post-fire conditions and between the nested sub-basins can be fairly carried out. The profusion and the quality of the data collected during the pre-fire period and the possibility of comparing the hydrological conse- quences on similar basins differently affected by the fire, lends this study a special relevance. In this paper we present the first hydrological results and consequences observed during the first year after the fire. Two aspects, considered the main hydrological changes by other authors (Rowe et al., 1954; Brown, 1972; Campbell et al., 1977; Chandler et al., 1983, Scott and van Wyn, 1990), have been especially analysed; (1) the changes in the annual runoff response, and (2) the changes in the flood regime.HYDROLOGICAL RESPONSE OF A MEDITERRANEAN BASIN 25 THE STUDY AREA The Réal Collobrier is an experimental research basin that has been observed and gauged continuously since 1966 by the CEMAGREF group from Aix-en-Provence, France. This basin (43° 14’N, 6° 13’E), located in the Maures massif (Var Department, France, see Fig. 1) near the Mediterranean sea (about 15km away), is in fact a set of nested sub-basins with marginal human influence. The measurement network consists of 17 pluviographs, spread over an area of 100 km?, a meteorological station and 11 flow recorder stations (with basin areas between 70ha and 70km?), some of which are equipped with a water temperature sensor, conductivity sensor and sampler (Lavabre, 1990; Sempere Torres, 1990). In August 1990 a severe fire partially affected this area (more than 8000ha destroyed in 3 days). Depending on the sub-basin, the portion of burnt surface varies from 0 to 85% (Lavabre et al., 1991). The present study concerns the Rimbaud basin, a 146ha nested sub-basin with elevations ranging from 470 to 622m (see Fig. 1), in which the fire destroyed 85% of the vegetation. - This basin can be considered as geologically homogenous, underlain by slightly altered gneiss. Its thin, sandy soils are of the Ranker type. Shrubby maquis and degraded forest of cork trees, chestnuts and maritime pines had occupied the basin totally before the fire. Table 1 summarises the main characteristics of the basin. The aim of this paper is to analyse and compare the hydrological response of the Rimbaud basin before and after the fire (see Fig. 2). To this end, we considered the 23 year pre-fice period as the reference period defining the normal behaviour for the basin. Some linear regressions and a lumped conceptual model have been calibrated, and used to simulate what the response would have been if the fire had not occurred. Differences between these simulations and the response observed are analysed herein. A com- plementary comparison with the response of the Vaubarnier, a nearby sub- basin of 148 ha which was not affected by the fire (see Fig. 1), is also provided. A complementary analysis of chemical and mechanical erosion has been carried out by Martin and Chevalier (1991). CHANGES IN THE ANNUAL RUNOFF RESPONSE First let us consider the hydrological year Nas the 12 month period starting on I August of the year N. Thus in our study the reference period before the fire comprises the hydrological years 1967-1989 (from August 1967 until July 1990). The study period was the hydrological year 1990 (from August 1990 to July 1991). Note that the fire took place between 22 and 24 August 1990. Nevertheless, no rainfall had been recorded since June and no flow was