# Spatial and seasonal distribution of rainfall erosivity in Australia.

Introduction

Erosion risk assessment is required for land management and conservation planning. The most commonly used method for predicting the average soil loss rate at large-scale remains the universal soil loss equation (USLE, Wischmeier and Smith 1978) and its recent modification, the Revised USLE (RUSLE, Renard et al. 1997). To use the USLE/RUSLE for soil loss prediction or to determine soil erodibility for the USLE/RUSLE at a particular site, the numerical value of a rainfall and runoff factor, known as the R-factor, is needed. The R-factor is a measure of rainfall erosivity and is defined as the mean annual sum of individual storm erosion index values, E[I.sub.30], where E is the total storm kinetic energy and [I.sub.30] is the maximum 30-min rainfall intensity. When factors other than rainfall are held constant, soil losses are directly proportional to the magnitude of rainfall erosivity (Wischmeier and Smith 1958, 1978). The R-factor represents the climatic influence on water-related soil erosion, and therefore can be used to quantify broad-scale, climate-driven, soil erosion potential. Monthly distribution of the rainfall erosivity is needed to determine a weighted cover factor for the RUSLE.

To compute storm E[I.sub.30] values, continuous rainfall intensity data at time intervals of less than 30 min are needed. Wischmeier and Smith (1978) recommended that at least 20 years of rainfall intensity data at short time intervals be used so that the natural climatic variations can be accommodated. Rainfall intensity data at short time intervals are available either in digitised pluviographs (commonly known as break-point data) or discrete rainfall rates associated with tipping bucket rain gauge. For simplicity, we call, henceforth, rainfall intensity data at short time intervals (<30-min) pluviograph data, as distinct from daily rainfall data. Spatial and temporal coverage of pluviograph data is usually limited. When available, pluviograph data are often incomplete and the recorded period is mostly short. Daily rainfall data, in contrast, are more widely available and for longer periods. It is therefore desirable to be able to estimate the R-factor and its monthly distribution, needed to apply the USLE/RUSLE for soil erosion prediction, from daily rainfall amounts.

In areas where long-term pluviograph data are not available, the R-factor may be estimated using mean annual rainfall or the Modified Fournier Index (e.g. Stocking and Elwell 1976; Arnoldus 1977; Roose 1977; Renard and Freimund 1994; Yu and Rosewell 1996c). Alternatively, 2-year, 6-h rainfall intensity probability values may be used to estimate the R-factor (Ateshian 1974; Wischmeier 1974; Wischmeier and Smith 1978; Rosewell 1993a). These approaches, however, do not allow determination of the seasonal distribution of rainfall erosivity. Information on seasonal distribution is needed to calculate average annual cover and management factors in the USLE/RUSLE (Renard et al. 1997). Seasonal distribution of rainfall erosivity is important for assessing erosion hazards. When peak rainfall erosivity coincides with exposure of bare soils through, for example, bare fallow, forest harvesting, or land clearing at construction sites, soil erosion risk is increased considerably. To adequately represent the erosive potential of rainfall for each temporally distinctive period, it is recommended that the USLE/RUSLE cover and management factor needs to be calculated on 15-day or monthly basis (Renard et al. 1997). At the continental scale, problematic to the annual-based application of the RUSLE is the pronounced wet-dry precipitation regime in the tropics and in regions with a Mediterranean climate.

Some work on the estimation of R-factor and its monthly distributions has been done in the past for most States of Australia. Isoerodent maps showing lines of equal exosivity were published for Victoria (Garvin et al. 1979), Queensland (Rosenthal and White 1980), Western Australia (McFarlane et al. 1986), New South Wales (Rosewell and Turner 1992), and South Australia (Yu and Rosewell 1996b). In addition, a relationship between the R-factor and 2-year, 6-h rainfall intensity was developed for a number of sites in Australia (Rosewell and Turner 1992; Rosewell 1993a). Such a relationship has been used to estimate rainfall erosivity for all States and territories based on rainfall intensity data published in Australian Rainfall and Runoff (Pilgrim 1987; Rosewell 1993b, 1997). The quality of the estimated R-factor depends on the quality of the limited pluviograph data and it is also known that rainfall intensity data cannot be used to estimate the seasonal distribution of rainfall erosivity. To overcome the problem with limited pluviograph data and the inability to estimate the seasonal distribution of rainfall erosivity, a model using daily rainfall data has been tested for both temperate and tropical regions of Australia (Yu and Rosewell 1996a, 1996b; Yu 1998). Regional relationships of model parameters were also developed to allow prediction of the R-factor and its monthly distribution from daily rainfall anywhere in Australia (Yu 1998). However, despite the increasing demand for environmental management, an up-to-date, high spatial resolution and consistent national digital map of R-factor and its seasonal distribution is currently lacking.

There are 2 major objectives of this study. The first is to assess the accuracy of estimating the R-factor and its monthly distribution from daily rainfall amounts. The second is to provide high resolution, up-to-date maps of R-factor and its monthly distribution across Australia that can be readily used by soil conservationists and environmental managers. In this paper, the daily erosivity model of Yu and Rosewell (1996a, 1996b) is tested for 132 sites around Australia by comparing the modelled R-factor with that compiled from literature based on pluviograph data. R-factor and its monthly distribution are further evaluated at 43 sites where long-term pluviograph data are used. The 43 sites cover a wide range of climates across Australia. No calibration of model parameters was attempted so that model errors could be independently assessed. An isoerodent map and maps showing monthly variation in rainfall erosivity are then produced using high-resolution daily rainfall data at 0.05[degrees] for Australia.

Methods

Data

Grid daily rainfall data

The model was applied to the Australian continent using 0.05[degrees] resolution daily rainfall data interpolated by Queensland Department of Natural Resources and Mines. The description of daily rainfall interpolation can be found in Jeffrey et al. (2001). Ordinary kriging was first used to spatially interpolate monthly rainfall values. Then, for each grid cell, the daily distribution of rainfall in the month was calculated by accessing the rainfall record from the station nearest to the point of interest, and partitioning the interpolated monthly rainfall onto individual days according to the historical record of daily rainfall at the nearest station. In this study, 20 years of grided daily rainfall data from 1 January 1980 to 31 December 1999 were used. Mean monthly rainfall, needed by the erosivity model to estimate model parameters, was calculated using the same daily rainfall data.

Pluviograph data

Two types of pluviograph sites were used in this study. The first consisted of 132 sites around Australia (Fig. 1). These R-factor values were compiled from previous studies (Rosenthal and White 1980; McFarlane et al. 1986; Yu and Rosewell 1996a, 1996b; Yu 1998). The R-factor was calculated by different researchers using pluviograph data available at the time. Different storm energy equations and algorithms were used to calculate the R-factor. Multiple R-factor values at the same sites were used as either different energy equations or period of rainfall data were used by different researchers. In some cases, the pluviograph data were incomplete or the record length was particularly short (Rosenthal and White 1980; McFarlane et al. 1986). In total, 153 R-factor values were compiled. All these R-factor values were used in this study to compare with the R-factor predicted using the 20-year grid daily rainfall data to assess conservatively the magnitude of model errors. No attempt was made to compare seasonal R-factor distribution for these sites as insufficient information about seasonal distribution was available from previous studies.

[FIGURE 1 OMITTED]

The second group contained 43 pluviograph sites distributed throughout Australia (Fig. 1). They were selected not only for their spatial coverage across Australia but also for their record length (at least 20 years) and their period of operation (mostly covering the period from 1980 to 1999). These sites cover all the major climate zones in Australia with the mean annual rainfall ranging from 271 mm at Giles to 2431 mm at Koombooloomba (Bureau of Meteorology 1989). Pluviograph data at 6-min intervals were extracted from Bureau of Meteorology archives for these 43 sites. R-factor and its monthly distribution were calculated using the RECS program (Yu and Rosewell 1998). Recommendations for calculating R-factor using pluviograph data from the RUSLE manual were strictly followed (Renard et al. 1997). Dry periods of 6 h or longer were used to separate storm events; monthly erosivity was the sum of E[I.sub.30] values of all storm events in the month; and the energy equation of Brown and Foster (1987) was used to determine total storm energy. Details about these 43 sites are presented in Table 1.

Model and method of analysis

The model to estimate the sum of E[I.sub.30] values for the month j, [E.sub.j], using daily rainfall amounts can be written in the form (Yu and Rosewell 1996a):

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [R.sub.d] is the daily rainfall amount, [R.sub.0] is the threshold rainfall amount to generate runoff, and N is the number of days with rainfall amount in excess [R.sub.0] in the month, and [alpha], [beta], [eta], and [omega] are model parameters. The sinusoidal function with a fundamental frequency f = 1/12 is used to describe the seasonal variation of the coefficient. It is used to describe the seasonal variation of rainfall erosivity for a given amount of daily rainfall.

Equation 1 differs from previous models in 2 important aspects. In previous models using daily or event rainfall amounts, E[I.sub.30] was estimated for individual events and model parameters were determined using log-linear or non-linear regression techniques (Richardson et al. 1983; Elsenbeer et al. 1993; Posch and Rekolainen 1993). Since monthly erosivity is much less variable than event E[I.sub.30] and only monthly values are needed to compute the R-factor and its monthly distribution, Eqn 1 contains more relevant parameters. Parameters of Eqn 1 are optimised on a monthly basis to ensure minimum bias. Secondly, the sinusoidal term was introduced to take into account the possibility of having different storm types in different seasons. This term allows erosivity for a given amount of rain to vary seasonally.

This model has a maximum of 5 parameters: [alpha], [beta], [eta], [omega], and [R.sub.0]. The parameter [omega] is set at [pi]/6, implying that for a given amount of daily rainfall the corresponding rainfall intensity is the highest in January, when the temperature is the highest for most parts of the continent. Two different values of rainfall threshold [R.sub.0] (12.7 mm and 0 mm) were used for the 43 sites in this study. In the USLE, Wischmeier and Smith (1978) suggested using 12.7 mm as the threshold rainfall [R.sub.0]. When the isoerodent map was prepared for the eastern part of the USA, a rainfall threshold of 12.7 mm was used (Wischmeier and Smith 1978). Most of the previous R-factor values presented in Fig. 2 were calculated using [R.sub.0] = 12.7 mm. The RUSLE manual has recommended that all storms be included in R-factor calculations (Renard et al. 1997). Yu (1999) found that the discrepancy in the calculated R-factor due to different rainfall thresholds increases as mean annual rainfall decreases because of the high relative contribution of small storm events to the R-factor. Two rainfall thresholds were considered in this paper to examine the effects of rainfall threshold on annual R-factor and its seasonal distribution at large space scale. To be consistent, the same 2 thresholds were used to calculate the R-factor and its seasonal distribution for the 43 sites both using pluviograph data and the daily rainfall erosivity model. Regional relationships were derived using 79 stations located in New South Wales, South Australia, and the tropics for parameters [alpha], [beta], [eta]. For the case of [R.sub.0] = 12.7 mm, the following sets of equations are used (Yu 1998):

(2) [alpha] = 0.395 [1 + 0.098 exp(3.26 [PSI]/[M.sub.R])]

(3) [beta] = 1.49

(4) [eta] = 0.29

where [M.sub.R] is the mean annual rainfall and [PSI] is the mean summer rainfall (November to April; Bureau of Meteorology 1989). For the case of [R.sub.0] = 0 mm, we use:

(5) [alpha] = 0.369 [1 + 0.098 exp(3.26 [PSI]/[M.sub.R])]

while values of [beta] and [eta] as same as Eqns 3 and 4.

[FIGURE 2 OMITTED]

Two measures were used to quantify the model performance. Firstly, the predictive capacity of R-factor is measured by the coefficient of efficiency, [E.sub.c] (Nash and Sutcliffe 1970). It is the fraction of total variation in the original data that can be explained by the model:

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [E.sub.i] and [E.sub.i] are the annual R-factor calculated using pluviograph data and the daily rainfall erosivity model for site i, respectively, [bar]E is average value of the R-factor calculated for all sites considered using pluviograph data. Essentially, [E.sub.c] is an indicator of how close the scatters of predicted versus actual values are to the 1:1 line. It is equivalent to the coefficient of determination ([r.sup.2]) for linear regression models and can be considered as a measure of model efficiency for any other types of models. [E.sub.c] is commonly used to assess model performance in hydrology (Loague and Freeze 1985) and soil science (Risse et al. 1993; King et al. 1996). Secondly, the accuracy of estimated seasonal distribution of rainfall erosivity is assessed by a discrepancy measure, [delta]. It is defined as the mean absolute difference between actual and estimated seasonal distribution of rainfall erosivity. Let [p.sub.j] and [p.sub.j] be the percentage contribution of the month j to the R-factor calculated by the model using pluviograph data and the daily rainfall erosivity model, respectively, then:

(7) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

In this study, the daily erosivity model is applied to predict the mean annual R-factor (averaged annual E[I.sub.30]), and mean monthly E[I.sub.30] values using 20 years daily rainfall data from 1980 to 1999. The SI unit of MJ mm/(ha.h.year) is used for the R-factor throughout this paper.

Results and discussion

The R-factor predicted using the daily model was compared with that calculated by several previous researchers for 132 sites (Rosenthal and White 1980; McFarlane et al. 1986; Yu and Rosewell 1996a, 1996b; Yu 1998). Figure 2 shows the comparison between the predicted and calculated R-factor using pluviograph data. The coefficient of efficiency [E.sub.c] = 0.81 with root mean squared error (rmse) of 1832 MJ.mm/(ha.h.year), or 48% of the mean and [r.sup.2] = 0.82. The average value of predicted R-factor for the 132 sites is 3987 MJ.mm/(ha.h.year) compared with 3854 MJ.mm/(ha.h.year) calculated using pluviograph data. No noticeable bias of the model is observed. Figure 3 shows the similar R-factor comparison using 2 different values of rainfall threshold [R.sub.0] for the 43 sites where long-term pluviograph data were available. With [R.sub.0] = 0 mm, the coefficient of efficiency [E.sub.c] = 0.94 with rmse of 908 MJ.mm/(ha.h.year), or 29% of the mean and [r.sup.2] = 0.95. When [R.sub.0] = 12.7 mm, the coefficient of efficiency [E.sub.c] = 0.93 with rmse of 946 MJ.mm/(ha.h.year), or 31% of the mean and [r.sup.2] = 0.95. Lowering the rainfall threshold from 12.7 mm to 0 mm increases the R-factor. This is true for both the R-factor calculated from pluviograph data and that predicted from daily rainfall. The amount of increase is smaller for areas with large R-factor values [<1% on average when R >1000 MJ mm/(ha.h.year)] than areas with a relatively smaller R-factor [over 10% on average when R [less than or equal to] 1000 MJ mm/(ha.h.year)]. For R-factor <1000 MJ mm/(ha.h.year), the amount of increase is slightly larger for the daily rainfall model compared with that based on pluviograph data. In general, R-factor predicted by the daily rainfall model compares well with various R-factors calculated using pluviograph data.

[FIGURE 3 OMITTED]

Table 1 summarises of the R-factor calculated based on pluviograph data and daily rainfall erosivity model together with the discrepancy measure [delta] for all 43 sites. Overall the agreement is better for the sites with higher R-factor values. The average discrepancy measure [delta] is 2.3% when [R.sub.0] = 0 mm and 2.5% when [R.sub.0] is set to 12.7 mm. The discrepancy measure [delta] ranges from 1.2% at Mount Gambier to 5% at Woomera when [R.sub.0] = 0, and from 1.1% at Koombooloomba to 6.4% at Woomera when [R.sub.0] is set to 12.7 mm. The predicted R-factor and its monthly distribution are both slightly improved by using threshold [R.sub.0] = 0 mm. It was also found that the daily rainfall model works almost equally well for the winter rainfall area, e.g. Perth, Adelaide, and Albany, where modelling erosivity from the rain total is challenging because the seasonal distributions of rainfall and rainfall erosivity could be out of phase. Six sites, representing different climatic regimes, were selected to illustrate the model predictive capacity of seasonal distribution of rainfall erosivity. The 6 sites are: Canberra, temperate climate with a uniform rainfall throughout the year; Perth, dominant rainfall in winter; Brisbane, subtropical climate; Darwin, tropical climate with a distinct wet season in summer. Koombooloomba has the highest mean annual rainfall among the 43 sites, while Giles is the driest site. Figure 4 shows the calculated and predicted monthly rainfall erosivity for these 6 sites. Except ,for Giles, the estimated seasonal patterns of rainfall erosivity for the other 5 sites match closely those based on long-term 6-min pluviograph data with the discrepancy measure d ranging from 1.1% to 2.8%. The larger discrepancy at Giles (4.2%) is due to a lack of storms in this arid environment and partially due to larger interpolation error in the grided rainfall data. The first problem could be relatively easy to fix by using longer periods of record. Fixing the second problem is more difficult. In the arid areas, the rain gauge density is sparse. This makes the interpolation of daily rainfall data across the 0.05[degrees] grid fundamentally difficult and likely to produce larger errors. Sites similar to Giles where rainfall is low also include Woomera (16001), Oodnadatta (17043), and Alice Springs (15590). The sites in dry areas tend to have above average discrepancy in the seasonal distribution of rainfall erosivity.

[FIGURE 4 OMITTED]

The predicted spatial patterns of the R-factor and the monthly distributions across the continent with rainfall threshold [R.sub.0] = 0 mm are shown in Figs 5 and 6. For the northern part of the continent, the monthly distributions of R-factor estimated using Eqn 3 generally show peaks in the summer period from December to February. Approximately 80% of the annual rainfall erosivity occurs between December and March. A negligible fraction occurs from April to October in northern Australia. This is consistent with the common rainfall pattern in the Australia's tropics of intense storms during summer and little rainfall during winter (Rosenthal and White 1980; McIvor et al. 1995). For the south-eastern part of the continent, predicted monthly R-factor distributions change gradually from summer dominance to uniform when moving from north to south, which is comparable with continent rainfall intensity distribution (Bureau of Meteorology 1989; Yu and Rosewell 1996a, 1996b; Yu 1998). Rainfall erosivity dominates in winter in the coastal area of southwest of Western Australia. The pattern then changes to a summer dominance inland within 100 km from the coast (Fig. 5). This is also comparable with the distributions of the R-factor estimated using pluviograph data for the region.(McFarlane et al. 1986).

[FIGURES 5-6 OMITTED]

The predicted R-factor and its seasonal distribution have been used to assess rill and sheet erosion rate at the continental scale (Lu et al. 2001). A digital version of the annual R-factor and its monthly distribution using a rainfall threshold [R.sub.0] = 12.7 mm can be obtained from the web site of the National Land and Water Resources Audit at: http:// audit.ea.gov.au/ANRA/atlas/.

Conclusions

This study of spatial and seasonal distribution of rainfall erosivity in Australia and previous investigations (Yu and Rosewell 1996a, 1996b; Yu 1998) have shown conclusively that the daily rainfall erosivity model can be used to accurately predict the R-factor and its seasonal distribution. Despite the uncertainty of previous R-factor calculations using pluviograph data from different periods, the minimum value of coefficient of efficiency is 0.81 for 132 sites across Australia. The coefficient of efficiency was increased to 0.93-0.94 for the 43 sites where the long-term pluviograph data were used. The average discrepancy between calculated and predicted seasonal distribution was no more than 3%. Changing rainfall threshold from 12.7 mm to 0 mm increases the R-factor by no more than 5% on average. The discrepancy in the R-factor due to different rainfall thresholds increases as mean annual rainfall decreases. Based on the recommendations for the RUSLE and the results from this study, we would recommend the use of 0 mm as the threshold for areas with a mean annual rainfall of <400 mm. Both thresholds are suitable for other areas. The erosivity model can reproduce the effect of using different thresholds on predicted R-factor. In general, the predictive accuracy of the annual R-factor and its seasonal distributions decreases from the tropics and subtropics, through temperate regions and winter rainfall areas, to the arid regions. Two factors contribute the low accuracy in the arid inland. One is a lack of sufficient storm events to obtain reliable long-term mean value of the R-factor. Another is the much coarser true spatial resolution in those areas where the number of rain gauge stations is small (Jeffrey et al. 2001). The high-resolution digital maps of the R-factor and its monthly distribution produced in this study can be used for assessing erosion hazard and determining the timing of erosion control strategies. The maps could readily be updated and their quality improved as longer-term daily rainfall data become available from the Bureau of Meteorology or the Queensland Department of Natural Resources and Mines.

Acknowledgments

This work is in part funded by the National Land and Water Resources Audit (NLWRA), Australia. It contains part of the results from Theme 5.4b, sediment transport and delivery project of NLWRA led by Chris Moran and Ian Prosser at CSIRO Land and Water. We thank the Queensland Department of Natural Resources for permission to use of its interpolated daily rainfall data. Graeme Priestley is acknowledged for his GIS support in producing Figs 5 and 6.

References

Arnoldus JMJ (1977) Methodology used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. FAO Soils Bulletin 34, 39-51.

Ateshian JKH (1974) Estimation of rainfall erosivity index. Journal of the Irrigation and Drainage Division. Proceedings of the American Society of Civil Engineers 110(IR3), 293-307.

Brown LC, Foster GR (1987) Storm erosivity using idealized intensity distributions. T ASAE 30, 379-386. Bureau of Meteorology (1989) `Climate of Australia.' (AGPS Press: Canberra)

Elsenbeer HD, Cassel K, Tinner W (1993) A daily rainfall erosivity model for Western Amazonia. Journal of Soil Water Conservation 48, 439-444.

Garvin RJ, Knight MR, Richmond TJ (1979) Guidelines for minimising soil erosion and sedimentation from construction sites in Victoria. Soil Conservation Authority, Melbourne, Vic.

Jeffrey SJ, Carter JO, Moodie KB, Beswick AR (2001) Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environmental Modelling and Software 16, 309-330.

King KW, Richardson CW, Williams JR (1996) Simulation of sediment and nitrate loss on a vertisol with conservation tillage practices. T ASAE 39, 2139-2145.

Loague KM, Freeze RA (1985) A comparison of rainfall-runoff modeling techniques on small upland catchments. Water Resources Research 21, 229-248.

Lu H, Gallant J, Prosser IP, Moran C, Priestley G (2001) Prediction of sheet and rill erosion over the Australian continent: Incorporating monthly soil loss distribution. Technical Report 13/01, CSIRO Land and Water, Canberra.

McFarlane DJ, Davies RJ, Westcott T (1986) Rainfall erosivity in Western Australia. Hydrology and Water Resources Symposium, Institution of Engineers, Australia, National Conference Publication No. 86/13, pp. 350-354.

McIvor JG, Williams J, Gardener CJ (1995) Pasture management influences runoff and soil movement in the semi-arid tropics. Australian Journal of Experimental Agriculture 35, 55-65.

Nash JE, Sutcliffe JV (1970) River flow forecasting through conceptual models. Part I--A discussion of principles. Journal of Hydrology 10, 282-290.

Pilgrim DH (1987) `Australian rainfall and runoff--A guide to flood estimation.' (The Institution of Engineers, Australia: Barton, ACT)

Posch M, Rekolainen S (1993) Erosivity factor in the Universal Soil Loss Equation estimated from Finnish rainfall data. Agricultural Sciences in Finland 2, 271-279.

Renard KG, Foster GA, Weesies GA, McCool DK (1997) Predicting soil erosion by water: A guide to conservation planning with the Revised Universal Soil Loss Equation (RUSLE). Agriculture Handbook No. 703. USDA, Washington, DC.

Renard KG, Freimund JR (1994) Using monthly precipitation data to estimate the R factor in the Revised USLE. Journal of Hydrology 157, 287-306.

Richardson CW, Foster GR, Wright DA (1983) Estimation of erosion index from daily rainfall amount. T ASAE 26, 153-157, 160.

Risse LMM, Nearing A, Nicks AD, Laflen JM (1993) Error assessment in the Universal Soil Loss Equation. Soil Science Society of America Journal 57, 825-833.

Roose E (1977) Erosion et Ruissellement en Afrique de l'ouest--2vingt annees de mesures en petites parcelles experimentales. Travaux et Documents de l'ORSTOM No. 78, OSTROM, Paris.

Rosenthal KM, White BJ (1980) Distribution of a rainfall erosion index in Queensland. Division of Land Utilisation Report 80/8, Queensland Department of Primary Industries, Brisbane, Qld.

Rosewell CJ (1993a) Estimation of rainfall erosivity. Erosion Research News Letter No. 11.

Rosewell CJ (1993b) `SOILOSS--A program to assist in the selection of management practices to reduce erosion.' 2nd edn. Technical Handbook No. 11. (Department of Conservation and Land Management: Sydney)

Rosewell CJ (1997) `Potential source of sediments and nutrients: Sheet and rill erosion and phosphorus sources.' Australia: State of the Environment Technical Paper Series (Inland Waters). (Department of the Environment, Sport and Territories: Canberra)

Rosewell CJ, Turner JB (1992) Rainfall erosivity in New South Wales. CaLM Technical Report No. 20, Department of Conservation and Land Management, Sydney.

Stocking MA, Elwell HA (1976) Erosivity determinations for seven rainfall stations in Rhodesia. Research Bulletin No. 21, Department of Conservation and Extension, Salisbury.

Wischmeier WH (1974) New developments in estimating water erosion. In `Proceedings of the 29th Annual Meeting of the Soil Society of America'. Madison, pp. 179-186.

Wischmeier WH, Smith DD (1958) Rainfall energy and its relationship to soil loss. Transactions of American Geophysical Union 39, 285-291.

Wischmeier WH, Smith DD (1978) `Predicting rainfall erosion losses: A guide to conservation planning.' Agriculture Handbook No. 537. (USDA: Washington, DC)

Yu B (1998) Rainfall erosivity and its estimation for Australia's tropics. Australian Journal of Soil Research 36, 143-165.

Yu B (1999) A comparison of the R-factor in the Universal Soil Loss Equation and Revised Universal Soil Loss Equation. T ASAE 42, 1615-1620.

Yu B, Rosewell CJ (1996a) An assessment of a daily rainfall erosivity model for New South Wales. Australian Journal of Soil Research 34, 139-152.

Yu B, Rosewell CJ (1996b) Rainfall erosivity estimation using daily rainfall amounts for South Australia. Australian Journal of Soil Research 34, 721-733.

Yu B, Rosewell CJ (1996c) A robust estimator of the R-factor for the Universal Soil Loss Equation. Transactions of the American Society of Agricultural Engineers 39, 559-561.

Yu B, Rosewell CJ (1998) RECS: A program to calculate the R-factor for the USLE/RUSLE using BOM/ AWS Pluviograph data. ENS Working Paper No. 8/98.

Manuscript received 19 December 2001, accepted 28 March 2002

Hua Lu (A) and Bofu Yu (B)

(A) CSIRO Land and Water, Canberra Laboratory, GPO Box 1666, Canberra, ACT 2601, Australia.

(B) Faculty of Environmental Sciences, Griffith University, Nathan, Qld 4109, Australia.

Erosion risk assessment is required for land management and conservation planning. The most commonly used method for predicting the average soil loss rate at large-scale remains the universal soil loss equation (USLE, Wischmeier and Smith 1978) and its recent modification, the Revised USLE (RUSLE, Renard et al. 1997). To use the USLE/RUSLE for soil loss prediction or to determine soil erodibility for the USLE/RUSLE at a particular site, the numerical value of a rainfall and runoff factor, known as the R-factor, is needed. The R-factor is a measure of rainfall erosivity and is defined as the mean annual sum of individual storm erosion index values, E[I.sub.30], where E is the total storm kinetic energy and [I.sub.30] is the maximum 30-min rainfall intensity. When factors other than rainfall are held constant, soil losses are directly proportional to the magnitude of rainfall erosivity (Wischmeier and Smith 1958, 1978). The R-factor represents the climatic influence on water-related soil erosion, and therefore can be used to quantify broad-scale, climate-driven, soil erosion potential. Monthly distribution of the rainfall erosivity is needed to determine a weighted cover factor for the RUSLE.

To compute storm E[I.sub.30] values, continuous rainfall intensity data at time intervals of less than 30 min are needed. Wischmeier and Smith (1978) recommended that at least 20 years of rainfall intensity data at short time intervals be used so that the natural climatic variations can be accommodated. Rainfall intensity data at short time intervals are available either in digitised pluviographs (commonly known as break-point data) or discrete rainfall rates associated with tipping bucket rain gauge. For simplicity, we call, henceforth, rainfall intensity data at short time intervals (<30-min) pluviograph data, as distinct from daily rainfall data. Spatial and temporal coverage of pluviograph data is usually limited. When available, pluviograph data are often incomplete and the recorded period is mostly short. Daily rainfall data, in contrast, are more widely available and for longer periods. It is therefore desirable to be able to estimate the R-factor and its monthly distribution, needed to apply the USLE/RUSLE for soil erosion prediction, from daily rainfall amounts.

In areas where long-term pluviograph data are not available, the R-factor may be estimated using mean annual rainfall or the Modified Fournier Index (e.g. Stocking and Elwell 1976; Arnoldus 1977; Roose 1977; Renard and Freimund 1994; Yu and Rosewell 1996c). Alternatively, 2-year, 6-h rainfall intensity probability values may be used to estimate the R-factor (Ateshian 1974; Wischmeier 1974; Wischmeier and Smith 1978; Rosewell 1993a). These approaches, however, do not allow determination of the seasonal distribution of rainfall erosivity. Information on seasonal distribution is needed to calculate average annual cover and management factors in the USLE/RUSLE (Renard et al. 1997). Seasonal distribution of rainfall erosivity is important for assessing erosion hazards. When peak rainfall erosivity coincides with exposure of bare soils through, for example, bare fallow, forest harvesting, or land clearing at construction sites, soil erosion risk is increased considerably. To adequately represent the erosive potential of rainfall for each temporally distinctive period, it is recommended that the USLE/RUSLE cover and management factor needs to be calculated on 15-day or monthly basis (Renard et al. 1997). At the continental scale, problematic to the annual-based application of the RUSLE is the pronounced wet-dry precipitation regime in the tropics and in regions with a Mediterranean climate.

Some work on the estimation of R-factor and its monthly distributions has been done in the past for most States of Australia. Isoerodent maps showing lines of equal exosivity were published for Victoria (Garvin et al. 1979), Queensland (Rosenthal and White 1980), Western Australia (McFarlane et al. 1986), New South Wales (Rosewell and Turner 1992), and South Australia (Yu and Rosewell 1996b). In addition, a relationship between the R-factor and 2-year, 6-h rainfall intensity was developed for a number of sites in Australia (Rosewell and Turner 1992; Rosewell 1993a). Such a relationship has been used to estimate rainfall erosivity for all States and territories based on rainfall intensity data published in Australian Rainfall and Runoff (Pilgrim 1987; Rosewell 1993b, 1997). The quality of the estimated R-factor depends on the quality of the limited pluviograph data and it is also known that rainfall intensity data cannot be used to estimate the seasonal distribution of rainfall erosivity. To overcome the problem with limited pluviograph data and the inability to estimate the seasonal distribution of rainfall erosivity, a model using daily rainfall data has been tested for both temperate and tropical regions of Australia (Yu and Rosewell 1996a, 1996b; Yu 1998). Regional relationships of model parameters were also developed to allow prediction of the R-factor and its monthly distribution from daily rainfall anywhere in Australia (Yu 1998). However, despite the increasing demand for environmental management, an up-to-date, high spatial resolution and consistent national digital map of R-factor and its seasonal distribution is currently lacking.

There are 2 major objectives of this study. The first is to assess the accuracy of estimating the R-factor and its monthly distribution from daily rainfall amounts. The second is to provide high resolution, up-to-date maps of R-factor and its monthly distribution across Australia that can be readily used by soil conservationists and environmental managers. In this paper, the daily erosivity model of Yu and Rosewell (1996a, 1996b) is tested for 132 sites around Australia by comparing the modelled R-factor with that compiled from literature based on pluviograph data. R-factor and its monthly distribution are further evaluated at 43 sites where long-term pluviograph data are used. The 43 sites cover a wide range of climates across Australia. No calibration of model parameters was attempted so that model errors could be independently assessed. An isoerodent map and maps showing monthly variation in rainfall erosivity are then produced using high-resolution daily rainfall data at 0.05[degrees] for Australia.

Methods

Data

Grid daily rainfall data

The model was applied to the Australian continent using 0.05[degrees] resolution daily rainfall data interpolated by Queensland Department of Natural Resources and Mines. The description of daily rainfall interpolation can be found in Jeffrey et al. (2001). Ordinary kriging was first used to spatially interpolate monthly rainfall values. Then, for each grid cell, the daily distribution of rainfall in the month was calculated by accessing the rainfall record from the station nearest to the point of interest, and partitioning the interpolated monthly rainfall onto individual days according to the historical record of daily rainfall at the nearest station. In this study, 20 years of grided daily rainfall data from 1 January 1980 to 31 December 1999 were used. Mean monthly rainfall, needed by the erosivity model to estimate model parameters, was calculated using the same daily rainfall data.

Pluviograph data

Two types of pluviograph sites were used in this study. The first consisted of 132 sites around Australia (Fig. 1). These R-factor values were compiled from previous studies (Rosenthal and White 1980; McFarlane et al. 1986; Yu and Rosewell 1996a, 1996b; Yu 1998). The R-factor was calculated by different researchers using pluviograph data available at the time. Different storm energy equations and algorithms were used to calculate the R-factor. Multiple R-factor values at the same sites were used as either different energy equations or period of rainfall data were used by different researchers. In some cases, the pluviograph data were incomplete or the record length was particularly short (Rosenthal and White 1980; McFarlane et al. 1986). In total, 153 R-factor values were compiled. All these R-factor values were used in this study to compare with the R-factor predicted using the 20-year grid daily rainfall data to assess conservatively the magnitude of model errors. No attempt was made to compare seasonal R-factor distribution for these sites as insufficient information about seasonal distribution was available from previous studies.

[FIGURE 1 OMITTED]

The second group contained 43 pluviograph sites distributed throughout Australia (Fig. 1). They were selected not only for their spatial coverage across Australia but also for their record length (at least 20 years) and their period of operation (mostly covering the period from 1980 to 1999). These sites cover all the major climate zones in Australia with the mean annual rainfall ranging from 271 mm at Giles to 2431 mm at Koombooloomba (Bureau of Meteorology 1989). Pluviograph data at 6-min intervals were extracted from Bureau of Meteorology archives for these 43 sites. R-factor and its monthly distribution were calculated using the RECS program (Yu and Rosewell 1998). Recommendations for calculating R-factor using pluviograph data from the RUSLE manual were strictly followed (Renard et al. 1997). Dry periods of 6 h or longer were used to separate storm events; monthly erosivity was the sum of E[I.sub.30] values of all storm events in the month; and the energy equation of Brown and Foster (1987) was used to determine total storm energy. Details about these 43 sites are presented in Table 1.

Model and method of analysis

The model to estimate the sum of E[I.sub.30] values for the month j, [E.sub.j], using daily rainfall amounts can be written in the form (Yu and Rosewell 1996a):

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [R.sub.d] is the daily rainfall amount, [R.sub.0] is the threshold rainfall amount to generate runoff, and N is the number of days with rainfall amount in excess [R.sub.0] in the month, and [alpha], [beta], [eta], and [omega] are model parameters. The sinusoidal function with a fundamental frequency f = 1/12 is used to describe the seasonal variation of the coefficient. It is used to describe the seasonal variation of rainfall erosivity for a given amount of daily rainfall.

Equation 1 differs from previous models in 2 important aspects. In previous models using daily or event rainfall amounts, E[I.sub.30] was estimated for individual events and model parameters were determined using log-linear or non-linear regression techniques (Richardson et al. 1983; Elsenbeer et al. 1993; Posch and Rekolainen 1993). Since monthly erosivity is much less variable than event E[I.sub.30] and only monthly values are needed to compute the R-factor and its monthly distribution, Eqn 1 contains more relevant parameters. Parameters of Eqn 1 are optimised on a monthly basis to ensure minimum bias. Secondly, the sinusoidal term was introduced to take into account the possibility of having different storm types in different seasons. This term allows erosivity for a given amount of rain to vary seasonally.

This model has a maximum of 5 parameters: [alpha], [beta], [eta], [omega], and [R.sub.0]. The parameter [omega] is set at [pi]/6, implying that for a given amount of daily rainfall the corresponding rainfall intensity is the highest in January, when the temperature is the highest for most parts of the continent. Two different values of rainfall threshold [R.sub.0] (12.7 mm and 0 mm) were used for the 43 sites in this study. In the USLE, Wischmeier and Smith (1978) suggested using 12.7 mm as the threshold rainfall [R.sub.0]. When the isoerodent map was prepared for the eastern part of the USA, a rainfall threshold of 12.7 mm was used (Wischmeier and Smith 1978). Most of the previous R-factor values presented in Fig. 2 were calculated using [R.sub.0] = 12.7 mm. The RUSLE manual has recommended that all storms be included in R-factor calculations (Renard et al. 1997). Yu (1999) found that the discrepancy in the calculated R-factor due to different rainfall thresholds increases as mean annual rainfall decreases because of the high relative contribution of small storm events to the R-factor. Two rainfall thresholds were considered in this paper to examine the effects of rainfall threshold on annual R-factor and its seasonal distribution at large space scale. To be consistent, the same 2 thresholds were used to calculate the R-factor and its seasonal distribution for the 43 sites both using pluviograph data and the daily rainfall erosivity model. Regional relationships were derived using 79 stations located in New South Wales, South Australia, and the tropics for parameters [alpha], [beta], [eta]. For the case of [R.sub.0] = 12.7 mm, the following sets of equations are used (Yu 1998):

(2) [alpha] = 0.395 [1 + 0.098 exp(3.26 [PSI]/[M.sub.R])]

(3) [beta] = 1.49

(4) [eta] = 0.29

where [M.sub.R] is the mean annual rainfall and [PSI] is the mean summer rainfall (November to April; Bureau of Meteorology 1989). For the case of [R.sub.0] = 0 mm, we use:

(5) [alpha] = 0.369 [1 + 0.098 exp(3.26 [PSI]/[M.sub.R])]

while values of [beta] and [eta] as same as Eqns 3 and 4.

[FIGURE 2 OMITTED]

Two measures were used to quantify the model performance. Firstly, the predictive capacity of R-factor is measured by the coefficient of efficiency, [E.sub.c] (Nash and Sutcliffe 1970). It is the fraction of total variation in the original data that can be explained by the model:

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [E.sub.i] and [E.sub.i] are the annual R-factor calculated using pluviograph data and the daily rainfall erosivity model for site i, respectively, [bar]E is average value of the R-factor calculated for all sites considered using pluviograph data. Essentially, [E.sub.c] is an indicator of how close the scatters of predicted versus actual values are to the 1:1 line. It is equivalent to the coefficient of determination ([r.sup.2]) for linear regression models and can be considered as a measure of model efficiency for any other types of models. [E.sub.c] is commonly used to assess model performance in hydrology (Loague and Freeze 1985) and soil science (Risse et al. 1993; King et al. 1996). Secondly, the accuracy of estimated seasonal distribution of rainfall erosivity is assessed by a discrepancy measure, [delta]. It is defined as the mean absolute difference between actual and estimated seasonal distribution of rainfall erosivity. Let [p.sub.j] and [p.sub.j] be the percentage contribution of the month j to the R-factor calculated by the model using pluviograph data and the daily rainfall erosivity model, respectively, then:

(7) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

In this study, the daily erosivity model is applied to predict the mean annual R-factor (averaged annual E[I.sub.30]), and mean monthly E[I.sub.30] values using 20 years daily rainfall data from 1980 to 1999. The SI unit of MJ mm/(ha.h.year) is used for the R-factor throughout this paper.

Results and discussion

The R-factor predicted using the daily model was compared with that calculated by several previous researchers for 132 sites (Rosenthal and White 1980; McFarlane et al. 1986; Yu and Rosewell 1996a, 1996b; Yu 1998). Figure 2 shows the comparison between the predicted and calculated R-factor using pluviograph data. The coefficient of efficiency [E.sub.c] = 0.81 with root mean squared error (rmse) of 1832 MJ.mm/(ha.h.year), or 48% of the mean and [r.sup.2] = 0.82. The average value of predicted R-factor for the 132 sites is 3987 MJ.mm/(ha.h.year) compared with 3854 MJ.mm/(ha.h.year) calculated using pluviograph data. No noticeable bias of the model is observed. Figure 3 shows the similar R-factor comparison using 2 different values of rainfall threshold [R.sub.0] for the 43 sites where long-term pluviograph data were available. With [R.sub.0] = 0 mm, the coefficient of efficiency [E.sub.c] = 0.94 with rmse of 908 MJ.mm/(ha.h.year), or 29% of the mean and [r.sup.2] = 0.95. When [R.sub.0] = 12.7 mm, the coefficient of efficiency [E.sub.c] = 0.93 with rmse of 946 MJ.mm/(ha.h.year), or 31% of the mean and [r.sup.2] = 0.95. Lowering the rainfall threshold from 12.7 mm to 0 mm increases the R-factor. This is true for both the R-factor calculated from pluviograph data and that predicted from daily rainfall. The amount of increase is smaller for areas with large R-factor values [<1% on average when R >1000 MJ mm/(ha.h.year)] than areas with a relatively smaller R-factor [over 10% on average when R [less than or equal to] 1000 MJ mm/(ha.h.year)]. For R-factor <1000 MJ mm/(ha.h.year), the amount of increase is slightly larger for the daily rainfall model compared with that based on pluviograph data. In general, R-factor predicted by the daily rainfall model compares well with various R-factors calculated using pluviograph data.

[FIGURE 3 OMITTED]

Table 1 summarises of the R-factor calculated based on pluviograph data and daily rainfall erosivity model together with the discrepancy measure [delta] for all 43 sites. Overall the agreement is better for the sites with higher R-factor values. The average discrepancy measure [delta] is 2.3% when [R.sub.0] = 0 mm and 2.5% when [R.sub.0] is set to 12.7 mm. The discrepancy measure [delta] ranges from 1.2% at Mount Gambier to 5% at Woomera when [R.sub.0] = 0, and from 1.1% at Koombooloomba to 6.4% at Woomera when [R.sub.0] is set to 12.7 mm. The predicted R-factor and its monthly distribution are both slightly improved by using threshold [R.sub.0] = 0 mm. It was also found that the daily rainfall model works almost equally well for the winter rainfall area, e.g. Perth, Adelaide, and Albany, where modelling erosivity from the rain total is challenging because the seasonal distributions of rainfall and rainfall erosivity could be out of phase. Six sites, representing different climatic regimes, were selected to illustrate the model predictive capacity of seasonal distribution of rainfall erosivity. The 6 sites are: Canberra, temperate climate with a uniform rainfall throughout the year; Perth, dominant rainfall in winter; Brisbane, subtropical climate; Darwin, tropical climate with a distinct wet season in summer. Koombooloomba has the highest mean annual rainfall among the 43 sites, while Giles is the driest site. Figure 4 shows the calculated and predicted monthly rainfall erosivity for these 6 sites. Except ,for Giles, the estimated seasonal patterns of rainfall erosivity for the other 5 sites match closely those based on long-term 6-min pluviograph data with the discrepancy measure d ranging from 1.1% to 2.8%. The larger discrepancy at Giles (4.2%) is due to a lack of storms in this arid environment and partially due to larger interpolation error in the grided rainfall data. The first problem could be relatively easy to fix by using longer periods of record. Fixing the second problem is more difficult. In the arid areas, the rain gauge density is sparse. This makes the interpolation of daily rainfall data across the 0.05[degrees] grid fundamentally difficult and likely to produce larger errors. Sites similar to Giles where rainfall is low also include Woomera (16001), Oodnadatta (17043), and Alice Springs (15590). The sites in dry areas tend to have above average discrepancy in the seasonal distribution of rainfall erosivity.

[FIGURE 4 OMITTED]

The predicted spatial patterns of the R-factor and the monthly distributions across the continent with rainfall threshold [R.sub.0] = 0 mm are shown in Figs 5 and 6. For the northern part of the continent, the monthly distributions of R-factor estimated using Eqn 3 generally show peaks in the summer period from December to February. Approximately 80% of the annual rainfall erosivity occurs between December and March. A negligible fraction occurs from April to October in northern Australia. This is consistent with the common rainfall pattern in the Australia's tropics of intense storms during summer and little rainfall during winter (Rosenthal and White 1980; McIvor et al. 1995). For the south-eastern part of the continent, predicted monthly R-factor distributions change gradually from summer dominance to uniform when moving from north to south, which is comparable with continent rainfall intensity distribution (Bureau of Meteorology 1989; Yu and Rosewell 1996a, 1996b; Yu 1998). Rainfall erosivity dominates in winter in the coastal area of southwest of Western Australia. The pattern then changes to a summer dominance inland within 100 km from the coast (Fig. 5). This is also comparable with the distributions of the R-factor estimated using pluviograph data for the region.(McFarlane et al. 1986).

[FIGURES 5-6 OMITTED]

The predicted R-factor and its seasonal distribution have been used to assess rill and sheet erosion rate at the continental scale (Lu et al. 2001). A digital version of the annual R-factor and its monthly distribution using a rainfall threshold [R.sub.0] = 12.7 mm can be obtained from the web site of the National Land and Water Resources Audit at: http:// audit.ea.gov.au/ANRA/atlas/.

Conclusions

This study of spatial and seasonal distribution of rainfall erosivity in Australia and previous investigations (Yu and Rosewell 1996a, 1996b; Yu 1998) have shown conclusively that the daily rainfall erosivity model can be used to accurately predict the R-factor and its seasonal distribution. Despite the uncertainty of previous R-factor calculations using pluviograph data from different periods, the minimum value of coefficient of efficiency is 0.81 for 132 sites across Australia. The coefficient of efficiency was increased to 0.93-0.94 for the 43 sites where the long-term pluviograph data were used. The average discrepancy between calculated and predicted seasonal distribution was no more than 3%. Changing rainfall threshold from 12.7 mm to 0 mm increases the R-factor by no more than 5% on average. The discrepancy in the R-factor due to different rainfall thresholds increases as mean annual rainfall decreases. Based on the recommendations for the RUSLE and the results from this study, we would recommend the use of 0 mm as the threshold for areas with a mean annual rainfall of <400 mm. Both thresholds are suitable for other areas. The erosivity model can reproduce the effect of using different thresholds on predicted R-factor. In general, the predictive accuracy of the annual R-factor and its seasonal distributions decreases from the tropics and subtropics, through temperate regions and winter rainfall areas, to the arid regions. Two factors contribute the low accuracy in the arid inland. One is a lack of sufficient storm events to obtain reliable long-term mean value of the R-factor. Another is the much coarser true spatial resolution in those areas where the number of rain gauge stations is small (Jeffrey et al. 2001). The high-resolution digital maps of the R-factor and its monthly distribution produced in this study can be used for assessing erosion hazard and determining the timing of erosion control strategies. The maps could readily be updated and their quality improved as longer-term daily rainfall data become available from the Bureau of Meteorology or the Queensland Department of Natural Resources and Mines.

Table 1. Station number and names, longitudes, latitudes, pluviograph data availability, mean annual rainfall (MAR) for both the pluviograph data and 20-year daily rainfall, rainfall erosivity calculated using pluviograph data and daily rainfall model with two different rainfall thresholds [R.sub.0] (mm), and discrepancy measures [delta] for 43 selected sites in Australia Station Station name Long. Lat. No. (east) (south) 2012 Halls Creek Airport 127[degrees]40' 18[degrees]14' 3003 Broome Airport 122[degrees]14' 17[degrees]57' 4032 Port Hedland Airport 118[degrees]37' 20[degrees]22' 6011 Carnarvon Airport 113[degrees]40' 24[degrees]53' 7045 Meekatharra Airport 118[degrees]33' 26[degrees]37' 8051 Geraldton Airport 114[degrees]42' 28[degrees]48' 9021 Perth Airport 115[degrees]58' 31[degrees]56' 9741 Albany Airport 117[degrees]48' 34[degrees]57' 9789 Esperance 121[degrees]54' 33[degrees]50' 12038 Kalgoorlie Boulder Airport 121[degrees]28, 30[degrees]47' 13017 Giles Meteorological 128[degrees]18 25[degrees]02' Office 14015 Darwin Airport 130[degrees]52' 12[degrees]25' 14508 Gove Airport 136[degrees]49' 12[degrees]17' 15135 Tennant Creek Airport 134[degrees]11' 19[degrees]38' 15590 Alice Springs Airport 133[degrees]54' 23[degrees]49' 16001 Woomera Aerodrome 136[degrees]49' 31[degrees]09' 17043 Oodnadatta Airport 135[degrees]27' 27[degrees]34' 18012 Ceduna Amo 133[degrees]43' 32[degrees]08' 23034 Adelaide Airport 138[degrees]32' 34[degrees]57' 26021 Mount Gambier Aero 140[degrees]47' 37[degrees]45' 27006 Coen Airport 143[degrees]07' 13[degrees]46' 27022 Thursday Island Mo 142[degrees]13' 10[degrees]35' 31083 Koombooloomba Dam 145[degrees]36' 17[degrees]50' 32040 Townsville Aero 146[degrees]46' 19[degrees]15' 33119 Mackay Mo 149[degrees]13' 21[degrees]07' 36031 Longreach Aero 144[degrees]17' 23[degrees]26' 39083 Rockhampton Aero 150[degrees]29' 23[degrees]23' 40223 Brisbane Aero 153[degrees]07' 27[degrees]23' 44021 Charleville Aero 146[degrees]16' 26[degrees]25' 48027 Cobar Mo 145[degrees]50' 31[degrees]29' 55024 Gunnadah 150[degrees]16' 31[degrees]02' 59040 Coffs Harbour Mo 153[degrees]07' 30[degrees]19' 66037 Sydney Airport Amo 151[degrees]10' 33[degrees]56' 70014 Canberra Airport 149[degrees]12' 35[degrees]19' 74114 Wagga Wagga Amo 147[degrees]18' 35[degrees]08' 76031 Mildura Airport 142[degrees]05' 34[degrees]14' 85072 East Sale Airport 147[degrees]09' 38[degrees]06' 86282 Melbourne Airport 144[degrees]51' 37[degrees]41' 91104 Launceston Airport 147[degrees]13' 41[degrees]33' 94008 Hobert Airport 147[degrees]30' 42[degrees]50' MAR (mm) Station Station name Period Pluvio. 1980- No. (year) 1999 2012 Halls Creek Airport 1955-2000 601 613 3003 Broome Airport 1948-2000 635 602 4032 Port Hedland Airport 1953-1998 361 317 6011 Carnarvon Airport 1956-1998 268 220 7045 Meekatharra Airport 1953-1998 253 253 8051 Geraldton Airport 1953-2000 474 451 9021 Perth Airport 1961-1998 779 783 9741 Albany Airport 1965-1998 787 880 9789 Esperance 1971-1998 589 601 12038 Kalgoorlie Boulder Airport 1939-1999 289 291 13017 Giles Meteorological 1956-1998 293 271 Office 14015 Darwin Airport 1953-2000 1718 1726 14508 Gove Airport 1966-1998 1318 1359 15135 Tennant Creek Airport 1969-1996 462 363 15590 Alice Springs Airport 1951-1999 329 240 16001 Woomera Aerodrome 1955-1999 201 166 17043 Oodnadatta Airport 1961-1985 230 161 18012 Ceduna Amo 1954-2000 294 277 23034 Adelaide Airport 1967-2000 441 434 26021 Mount Gambier Aero 1942-2000 694 727 27006 Coen Airport 1967-1997 1204 1223 27022 Thursday Island Mo 1961-1993 1793 1682 31083 Koombooloomba Dam 1960-1997 2592 2431 32040 Townsville Aero 1953-1999 1083 946 33119 Mackay Mo 1959-1997 1599 1451 36031 Longreach Aero 1966-1997 454 395 39083 Rockhampton Aero 1939-1999 821 709 40223 Brisbane Aero 1949-1998 1166 1151 44021 Charleville Aero 1953-1999 506 444 48027 Cobar Mo 1962-2000 414 366 55024 Gunnadah 1946-1997 638 638 59040 Coffs Harbour Mo 1960-2000 1671 1687 66037 Sydney Airport Amo 1962-2000 1124 1084 70014 Canberra Airport 1937-2000 613 663 74114 Wagga Wagga Amo 1947-1996 586 575 76031 Mildura Airport 1953-1998 295 283 85072 East Sale Airport 1953-2000 592 588 86282 Melbourne Airport 1970-2000 535 529 91104 Launceston Airport 1938-2000 665 565 94008 Hobert Airport 1960-2000 530 483 R-factor (MJ mm/ha.h.year) Pluviograph data Station Station name [R.sub.0] [R.sub.0] No. = 12.7 = 0.0 2012 Halls Creek Airport 2744 2908 3003 Broome Airport 4352 4451 4032 Port Hedland Airport 1265 1322 6011 Carnarvon Airport 632 682 7045 Meekatharra Airport 451 517 8051 Geraldton Airport 820 953 9021 Perth Airport 1172 1312 9741 Albany Airport 557 723 9789 Esperance 498 645 12038 Kalgoorlie Boulder Airport 402 474 13017 Giles Meteorological 653 722 Office 14015 Darwin Airport 13 556 13 856 14508 Gove Airport 7677 7884 15135 Tennant Creek Airport 1915 2026 15590 Alice Springs Airport 951 1012 16001 Woomera Aerodrome 264 315 17043 Oodnadatta Airport 430 479 18012 Ceduna Amo 264 333 23034 Adelaide Airport 285 419 26021 Mount Gambier Aero 406 549 27006 Coen Airport 6109 6308 27022 Thursday Island Mo 12 705 12 965 31083 Koombooloomba Dam 8060 8210 32040 Townsville Aero 6375 6532 33119 Mackay Mo 9475 9707 36031 Longreach Aero 1773 1873 39083 Rockhampton Aero 2877 3003 40223 Brisbane Aero 4706 4901 44021 Charleville Aero 1214 1315 48027 Cobar Mo 1030 1124 55024 Gunnadah 1433 1574 59040 Coffs Harbour Mo 7101 7348 66037 Sydney Airport Amo 3548 3746 70014 Canberra Airport 766 901 74114 Wagga Wagga Amo 834 980 76031 Mildura Airport 378 452 85072 East Sale Airport 512 654 86282 Melbourne Airport 644 786 91104 Launceston Airport 440 595 94008 Hobert Airport 719 830 R-factor (MJ mm/ha.h.year) Daily rainfall model Station Station name [R.sub.0] [R.sub.0] No. = 12.7 = 0.0 2012 Halls Creek Airport 4342 4056 3003 Broome Airport 5443 5085 4032 Port Hedland Airport 1815 1696 6011 Carnarvon Airport 304 284 7045 Meekatharra Airport 473 442 8051 Geraldton Airport 467 436 9021 Perth Airport 897 838 9741 Albany Airport 745 696 9789 Esperance 549 513 12038 Kalgoorlie Boulder Airport 432 404 13017 Giles Meteorological 828 773 Office 14015 Darwin Airport 15 093 14 100 14508 Gove Airport 9471 8848 15135 Tennant Creek Airport 2181 2037 15590 Alice Springs Airport 712 665 16001 Woomera Aerodrome 206 192 17043 Oodnadatta Airport 451 421 18012 Ceduna Amo 215 201 23034 Adelaide Airport 318 297 26021 Mount Gambier Aero 577 539 27006 Coen Airport 10 038 9377 27022 Thursday Island Mo 14 282 13 342 31083 Koombooloomba Dam 8100 7567 32040 Townsville Aero 7108 6640 33119 Mackay Mo 9130 8529 36031 Longreach Aero 1637 1529 39083 Rockhampton Aero 2773 2590 40223 Brisbane Aero 4320 4036 44021 Charleville Aero 1205 1126 48027 Cobar Mo 746 697 55024 Gunnadah 1524 1424 59040 Coffs Harbour Mo 6708 6266 66037 Sydney Airport Amo 2976 2780 70014 Canberra Airport 1418 1325 74114 Wagga Wagga Amo 822 768 76031 Mildura Airport 309 289 85072 East Sale Airport 857 801 86282 Melbourne Airport 684 639 91104 Launceston Airport 675 631 94008 Hobert Airport 604 564 [delta] (%) Station Station name [R.sub.0] [R.sub.0] No. = 12.7 = 0.0 2012 Halls Creek Airport 2.0 1.9 3003 Broome Airport 1.7 1.5 4032 Port Hedland Airport 2.9 2.8 6011 Carnarvon Airport 2.6 2.5 7045 Meekatharra Airport 2.9 2.7 8051 Geraldton Airport 2.2 2.1 9021 Perth Airport 1.4 1.3 9741 Albany Airport 1.9 1.6 9789 Esperance 2.7 1.9 12038 Kalgoorlie Boulder Airport 3.1 2.7 13017 Giles Meteorological 4.7 4.2 Office 14015 Darwin Airport 2.0 1.9 14508 Gove Airport 2.8 2.6 15135 Tennant Creek Airport 2.7 2.5 15590 Alice Springs Airport 3.6 3.1 16001 Woomera Aerodrome 6.4 5.0 17043 Oodnadatta Airport 4.3 4.0 18012 Ceduna Amo 2.3 2.2 23034 Adelaide Airport 3.1 2.0 26021 Mount Gambier Aero 1.8 1.2 27006 Coen Airport 1.4 1.3 27022 Thursday Island Mo 1.8 1.7 31083 Koombooloomba Dam 1.1 2.0 32040 Townsville Aero 1.5 1.5 33119 Mackay Mo 1.6 1.6 36031 Longreach Aero 1.7 1.4 39083 Rockhampton Aero 2.7 1.5 40223 Brisbane Aero 2.2 1.6 44021 Charleville Aero 1.5 1.7 48027 Cobar Mo 3.4 2.5 55024 Gunnadah 2.8 2.1 59040 Coffs Harbour Mo 1.5 1.6 66037 Sydney Airport Amo 2.7 3.0 70014 Canberra Airport 2.8 2.6 74114 Wagga Wagga Amo 2.9 1.4 76031 Mildura Airport 2.9 2.4 85072 East Sale Airport 1.7 2.4 86282 Melbourne Airport 2.3 2.8 91104 Launceston Airport 2.6 3.2 94008 Hobert Airport 5.7 1.5

Acknowledgments

This work is in part funded by the National Land and Water Resources Audit (NLWRA), Australia. It contains part of the results from Theme 5.4b, sediment transport and delivery project of NLWRA led by Chris Moran and Ian Prosser at CSIRO Land and Water. We thank the Queensland Department of Natural Resources for permission to use of its interpolated daily rainfall data. Graeme Priestley is acknowledged for his GIS support in producing Figs 5 and 6.

References

Arnoldus JMJ (1977) Methodology used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. FAO Soils Bulletin 34, 39-51.

Ateshian JKH (1974) Estimation of rainfall erosivity index. Journal of the Irrigation and Drainage Division. Proceedings of the American Society of Civil Engineers 110(IR3), 293-307.

Brown LC, Foster GR (1987) Storm erosivity using idealized intensity distributions. T ASAE 30, 379-386. Bureau of Meteorology (1989) `Climate of Australia.' (AGPS Press: Canberra)

Elsenbeer HD, Cassel K, Tinner W (1993) A daily rainfall erosivity model for Western Amazonia. Journal of Soil Water Conservation 48, 439-444.

Garvin RJ, Knight MR, Richmond TJ (1979) Guidelines for minimising soil erosion and sedimentation from construction sites in Victoria. Soil Conservation Authority, Melbourne, Vic.

Jeffrey SJ, Carter JO, Moodie KB, Beswick AR (2001) Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environmental Modelling and Software 16, 309-330.

King KW, Richardson CW, Williams JR (1996) Simulation of sediment and nitrate loss on a vertisol with conservation tillage practices. T ASAE 39, 2139-2145.

Loague KM, Freeze RA (1985) A comparison of rainfall-runoff modeling techniques on small upland catchments. Water Resources Research 21, 229-248.

Lu H, Gallant J, Prosser IP, Moran C, Priestley G (2001) Prediction of sheet and rill erosion over the Australian continent: Incorporating monthly soil loss distribution. Technical Report 13/01, CSIRO Land and Water, Canberra.

McFarlane DJ, Davies RJ, Westcott T (1986) Rainfall erosivity in Western Australia. Hydrology and Water Resources Symposium, Institution of Engineers, Australia, National Conference Publication No. 86/13, pp. 350-354.

McIvor JG, Williams J, Gardener CJ (1995) Pasture management influences runoff and soil movement in the semi-arid tropics. Australian Journal of Experimental Agriculture 35, 55-65.

Nash JE, Sutcliffe JV (1970) River flow forecasting through conceptual models. Part I--A discussion of principles. Journal of Hydrology 10, 282-290.

Pilgrim DH (1987) `Australian rainfall and runoff--A guide to flood estimation.' (The Institution of Engineers, Australia: Barton, ACT)

Posch M, Rekolainen S (1993) Erosivity factor in the Universal Soil Loss Equation estimated from Finnish rainfall data. Agricultural Sciences in Finland 2, 271-279.

Renard KG, Foster GA, Weesies GA, McCool DK (1997) Predicting soil erosion by water: A guide to conservation planning with the Revised Universal Soil Loss Equation (RUSLE). Agriculture Handbook No. 703. USDA, Washington, DC.

Renard KG, Freimund JR (1994) Using monthly precipitation data to estimate the R factor in the Revised USLE. Journal of Hydrology 157, 287-306.

Richardson CW, Foster GR, Wright DA (1983) Estimation of erosion index from daily rainfall amount. T ASAE 26, 153-157, 160.

Risse LMM, Nearing A, Nicks AD, Laflen JM (1993) Error assessment in the Universal Soil Loss Equation. Soil Science Society of America Journal 57, 825-833.

Roose E (1977) Erosion et Ruissellement en Afrique de l'ouest--2vingt annees de mesures en petites parcelles experimentales. Travaux et Documents de l'ORSTOM No. 78, OSTROM, Paris.

Rosenthal KM, White BJ (1980) Distribution of a rainfall erosion index in Queensland. Division of Land Utilisation Report 80/8, Queensland Department of Primary Industries, Brisbane, Qld.

Rosewell CJ (1993a) Estimation of rainfall erosivity. Erosion Research News Letter No. 11.

Rosewell CJ (1993b) `SOILOSS--A program to assist in the selection of management practices to reduce erosion.' 2nd edn. Technical Handbook No. 11. (Department of Conservation and Land Management: Sydney)

Rosewell CJ (1997) `Potential source of sediments and nutrients: Sheet and rill erosion and phosphorus sources.' Australia: State of the Environment Technical Paper Series (Inland Waters). (Department of the Environment, Sport and Territories: Canberra)

Rosewell CJ, Turner JB (1992) Rainfall erosivity in New South Wales. CaLM Technical Report No. 20, Department of Conservation and Land Management, Sydney.

Stocking MA, Elwell HA (1976) Erosivity determinations for seven rainfall stations in Rhodesia. Research Bulletin No. 21, Department of Conservation and Extension, Salisbury.

Wischmeier WH (1974) New developments in estimating water erosion. In `Proceedings of the 29th Annual Meeting of the Soil Society of America'. Madison, pp. 179-186.

Wischmeier WH, Smith DD (1958) Rainfall energy and its relationship to soil loss. Transactions of American Geophysical Union 39, 285-291.

Wischmeier WH, Smith DD (1978) `Predicting rainfall erosion losses: A guide to conservation planning.' Agriculture Handbook No. 537. (USDA: Washington, DC)

Yu B (1998) Rainfall erosivity and its estimation for Australia's tropics. Australian Journal of Soil Research 36, 143-165.

Yu B (1999) A comparison of the R-factor in the Universal Soil Loss Equation and Revised Universal Soil Loss Equation. T ASAE 42, 1615-1620.

Yu B, Rosewell CJ (1996a) An assessment of a daily rainfall erosivity model for New South Wales. Australian Journal of Soil Research 34, 139-152.

Yu B, Rosewell CJ (1996b) Rainfall erosivity estimation using daily rainfall amounts for South Australia. Australian Journal of Soil Research 34, 721-733.

Yu B, Rosewell CJ (1996c) A robust estimator of the R-factor for the Universal Soil Loss Equation. Transactions of the American Society of Agricultural Engineers 39, 559-561.

Yu B, Rosewell CJ (1998) RECS: A program to calculate the R-factor for the USLE/RUSLE using BOM/ AWS Pluviograph data. ENS Working Paper No. 8/98.

Manuscript received 19 December 2001, accepted 28 March 2002

Hua Lu (A) and Bofu Yu (B)

(A) CSIRO Land and Water, Canberra Laboratory, GPO Box 1666, Canberra, ACT 2601, Australia.

(B) Faculty of Environmental Sciences, Griffith University, Nathan, Qld 4109, Australia.

Printer friendly Cite/link Email Feedback | |

Author: | Lu, Hua; Yu, Bofu |
---|---|

Publication: | Australian Journal of Soil Research |

Geographic Code: | 8AUST |

Date: | Nov 1, 2002 |

Words: | 6226 |

Previous Article: | Soil properties of a widely spaced, planted poplar (Populus deltoides)--pasture system in a hill environment. |

Next Article: | Interpretation of morphological features in a salt-affected duplex soil toposequence with an altered soil water regime in western Victoria. |

Topics: |