Bedload transport, hydrology and river hydraulics in the Ngarradj Creek catchment, Jabiluka, Northern Territory, Australia

Supervising Scientist Report 199
Erskine WD, Saynor MJ, Evans KG & Moliere D
Department of Sustainability, Environment, Water, Population and Communities, 2011

ISSN 1325-1554
ISBN-13: 978-1-921069-17-8


Executive summary

Rainfall, discharge and bedload were measured at three gauging stations in the Ngarradj Creek catchment at Jabiluka, Northern Territory. These gauging stations were East Tributary, Upper Swift Creek and Swift Creek, and all had catchment areas less than 45 km2. Hand-held pressure-difference Helley-Smith bedload samplers were used to measure bedload fluxes for the 1998/1999, 1999/2000, 2000/2001 and 2001/2002 wet seasons. The bedload sampling procedure involved the completion of two traverses of the channel with at least four measurement points on each traverse at East Tributary, five at Upper Swift Creek and six at Swift Creek. Minimum sample collection time was 120 seconds and the maximum was 660 seconds. These variations were determined by bedload flux so that no more than 40% of the sample bag was filled at a time.

Rainfall is strongly seasonal over the Ngarradj Creek catchment, being concentrated in the wet season between November and April. Mean annual point rainfall between 1998 and 2007 for the water year (September to August inclusive) varied over the Ngarradj Creek catchment from 1731 ± 98 mm (SE) at Upper Swift Creek gauge to 1737 ± 105 mm at Swift Creek to 1754 ± 116 mm at East Tributary. Using the Thiessen polygon method, mean annual catchment (areal) rainfall for the same time period was 1735 ± 100 mm based on four stations in the Ngarradj Creek catchment. The recent time period has been characterised by above average rainfall with annual catchment rainfall being much greater than the mean for six of the nine years. CUSUM analysis of the long-term nearby Oenpelli rainfall record (1910–2010) found that there are alternating wet and dry periods that usually persist for at least a decade. The wet periods have a mean annual rainfall between 1537 and 1648 mm whereas the dry periods have a mean between 1267 and 1287 mm. Rainfall oscillates between these two different states, with one exception, 1955–1972, when rainfall was essentially constant at an intermediate value of 1436 mm. These different rainfall states are statistically significantly different from each other and all bedload measurements were completed during the last wet period. Between 190 and 440 mm of rainfall are required before streamflow commences in December in most years. Streamflow persists until at least April. Mean annual runoff, as a percentage of mean annual rainfall, decreases slightly with catchment area from 48 ± 8% at East Tributary to 46 ± 11% at Upper Swift Creek to 39 ± 9% at Swift Creek. Bankfull discharge usually occurs at least once during each wet season.

At-a-station hydraulic geometry equations were calculated for the velocity-area gauging data at each station. At East Tributary, the exponents exhibited the trend m>f>b whereas at Upper Swift Creek it was f>m>b and at Swift Creek, f>b>m. East Tributary is a type 4 river whereas both Upper Swift Creek and Swift Creek are type 10 rivers according to the Rhodes classification scheme. For type 4 rivers, width-depth ratio and velocity-area ratio decrease while Froude Number and slope-roughness ratio increase with increasing discharge. For type 10 rivers, all the above morphologic and hydrodynamic parameters decrease with increasing discharge. This indicates that the East Tributary gauge is characterised by higher stream powers than the other two stations and that all three stations respond to increasing discharge differently.

A total of 52 double traverses at East Tributary, 57 at Upper Swift Creek and 60 at Swift Creek were completed over the four wet seasons. Bedload ratings were calculated for four data sets, namely the whole data set at each gauge, the above threshold data set at East Tributary, the reliable data set at each gauge and the censored data set at Upper Swift Creek and Swift Creek. The ‘whole data sets’ comprised every mean bedload flux for each paired bedload transect at each gauge. The ‘above threshold data set’ at East Tributary only included the bedload fluxes for discharges greater than 0.223 m3/s because at lower discharges the fluxes clustered around zero flux. The ‘reliable data sets’ comprised all mean bedload fluxes where paired transect values differed by less than 4 times and where the gauge height change during the paired transect gaugings was ≤ 0.02 m at each gauge. The ‘censored data sets’ involved those bedload fluxes measured during equilibrium conditions when there was no pronounced scour or fill at Upper Swift and Swift Creek gauges.

Significant bedload ratings were defined as those that were not only statistically significant (ρ ≤ 0.05) but also explained a ‘meaningful’ amount of the variance in bedload flux. At least 0.60 of the variance in bedload flux had to be explained for a bedload rating to be accepted as reliable. For the three stations, twenty-three bedload ratings complied with the above criteria. Sixteen equations were accepted for East Tributary, thirteen for the ‘whole data set’, two for the ‘above threshold data set’ and one for the ‘reliable data set’. For Upper Swift Creek, four bedload ratings were accepted for the ‘censored data set’ and for Swift Creek, three bedload ratings were accepted for the ‘censored data set’. Significant bedload ratings were established between bedload flux and discharge, unit bedload flux and discharge, transport rate of unsuspended bedload by immersed weight per unit width and time and both unit and excess unit stream power, and adjusted submersed bedload weight and both unit and excess unit stream power for raw and log10-transformed data.

Bedload yields were calculated by thirty-nine methods at East Tributary, nine methods at Upper Swift Creek and eleven methods at Swift Creek. These methods involved combining the above bedload rating curves with either the hourly or daily hydrographs or the flow duration curves for the period 1 September 1998 to 31 August 2005. Ferguson’s (1986) and Duan’s (1983) corrections for bias were used with all methods based on log10-transformed ratings. Mean annual bedload yields varied by three orders of magnitude at East Tributary and by two orders of magnitude at Upper Swift Creek and Swift Creek. Hourly discharges usually produced higher yields than daily discharges. The bedload rating-flow duration curve technique overestimates yields and bias correction methods always produce even higher yields. Ratings using both immersed bedload weight and adjusted immersed bedload weight always underpredict yields because they contain an implicit threshold of motion condition that is at least four times greater than that predicted by Bagnold (1980). Such a result questions the applicability of Bagnold’s (1980) threshold to the Ngarradj Creek catchment. The best estimates of mean annual bedload yield at East Tributary, Upper Swift Creek and Swift Creek are 575 ± 65 (SE), 1000 ± 120 and 1625 ± 180 t/yr respectively.

Bedload sediments are similar at all sites. At East Tributary, bedload is a moderately sorted, coarse skewed, leptokurtic, coarse sand. At Upper Swift Creek, bedload is a moderately sorted, coarse skewed, mesokurtic, medium sand. At Swift Creek, bedload is a moderately sorted, coarse skewed, leptokurtic, coarse sand. There is little difference in grain size statistics between wet season bedload and dry season bed material. The differences that were significant suggest that most of the bed material is transported as bedload during the wet season. There may be some size selective transport at all three gauging stations with bedload being better sorted. At East Tributary, bedload samples are also less coarse skewed than the bed material. All these differences in grain size statistics indicate that bedload may be a slightly finer fraction of the total bed material but the differences are mainly in the extreme coarse fraction which may be mobile only under extreme events.