The influence of contrasting suspended particulate matter transport regimes on the bias and precision of flux estimates

Abstract

A large database (507 station-years) of daily suspended particulate matter (SPM) concentration and discharge data from 36 stations on river basins ranging from 600 km2 to 600,000 km2 in size (USA and Europe) was collected to assess the effects of SPM transport regime on bias and imprecision of flux estimates when using infrequent surveys and the discharge-weighted mean concentration method. By extracting individual SPM concentrations and corresponding discharge values from the database, sampling frequencies from 12 to 200 per year were simulated using Monte Carlo techniques. The resulting estimates of yearly SPM fluxes were compared to reference fluxes derived from the complete database. For each station and given frequency, bias was measured by the median of relative errors between estimated and reference fluxes, and imprecision by the difference between the upper and lower deciles of relative errors. Results show that the SPM transport regime of rivers affects the bias and imprecision of fluxes estimated by the discharge-weighted mean concentration method for given sampling frequencies (e.g. weekly, bimonthly, monthly). The percentage of annual SPM flux discharged in 2% of time (Ms2) is a robust indicator of SPM transport regime directly related to bias and imprecision. These errors are linked to the Ms2 indicator for various sampling frequencies within a specific nomograph. For instance, based on a deviation of simulated flux estimates from reference fluxes lower than ±20% and a bias lower than 1% or 2%, the required sampling intervals are less than 3 days for rivers with Ms2 greater than 40% (basin size<10,000 km2), between 3 and 5 days for rivers with Ms2 between 30 and 40% (basin size between 10,000 and 50,000 km2), between 5 and 12 days for Ms2 from 20% to 30% (basin size between 50,000 and 200,000 km2), 12–20 days for Ms2 in the 15–20% range (basin size between 200,000 and 500,000 km2).