CHOICE OF SAMPLING STRATEGY AND ESTIMATION METHOD FOR CALCULATING NITROGEN AND PHOSPHORUS TRANSPORT IN SMALL LOWLAND STREAMS

As reliable estimates of stream nutrient transport are required for many purposes including trend analysis, mass balances and model development, the impact of sampling strategy and estimation method on the bias and precision of stream nitrogen (N) and phosphorus (P) transport calculations was evaluated. The study was undertaken in two catchments in eastern Denmark. Selection of the most accurate sampling strategy and estimation method, i.e. with the lowest root mean square error (RMSE) was based on random (Monte Carlo) runs for generating replicate data sets from an essentially complete record of the concentration of total N (TN), total P (TP), particulate P (PP) and dissolved P (DP) during a two-year period (June 1987 to June 1989). The evaluation comprised 13 different estimation methods and seven discrete sampling strategies involving three categories (regular, stratified and strata sampling).  The regular sampling strategies were more accurate (lower RMSE) during high-flow periods than stratified sampling. The greatest improvement in RMSE for TN, TP, PP and DP transport was obtained when increasing the sampling frequency from 12 each year (monthly) to 18 (monthly in summer and fortnightly in winter) and 26 each year (fortnightly). The increase in accuracy (RMSE) was less when increasing the sampling frequency to 52 (weekly) or 104 (biweekly).  Nearly all the methods evaluated underestimated the annual transport of TP and PP, whereas TN and DP were both under- and overestimated. The best method of estimating N and P transport when utilizing discrete sampling was both site- and time-dependent. The overall best and most reproducible (stream to stream, year to year) method for estimating annual transport of TN, TP, PP and DP was a linear interpolation method. When this method was used to derive estimates of annual TN and TP transport based on fortnightly sampling, the RMSE was 1.4–5.4 and 20.2–38.5%, respectively, in the Gelbaek stream and 1.1–4.9 and 10.5–15.0%, respectively, in the Gjern A stream. Subdividing the hydrograph into two strata (low-flow and high-flow periods) and sampling these strata separately for calculating TP transport was superior to discrete sampling for the smaller of the two catchments. A combination of regular sampling (monthly) and pooled high-flow sampling (eight events out of a total of 43) reduced the RMSE of the annual TP load to 10.4%.