Ten methods for estimating the loads of constituents in a river were testedusing data from the River Don in North-East Scotland. By treating loads derivedfrom flow and concentration data collected every 2 days as a truth to bepredicted, the ten methods were assessed for use when concentration data arecollected fortnightly or monthly by sub-sampling from the original data.Estimates of coefficients of variation, bias and mean squared errors of themethods were compared; no method consistently outperformed all others anddifferent methods were appropriate for different constituents. The widely usedinterpolation methods can be improved upon substantially by modelling therelationship of concentration with flow or seasonality but only if theserelationships are strong enough.
展开▼
