A comparison of discrete and intensive sampling for measuring the loads of nitrogen and phosphorus in the river main, County Antrim

Abstract The River Main, County Antrim, was intensively monitored during November 1974 to May 1975 to obtain accurate measurements of the loads of nitrogen and phosphorus to compare with loads predicted by a statistical model from discrete sampling and continuous flow data. During the survey period 2-hourly samples were collected on 102 days. Flow data were available from continuous recording of river level. Depending on the prevailing flow conditions 2-hourly or composite 8-hourly samples were analysed for soluble reactive P, soluble unreactive P, total soluble P, particulate P, total P, nitrate N, Kjeldahl N and total N. For the discrete sampling programme grab samples were collected at 8-day intervals and analysed for soluble reactive P, total P and nitrate N. River flow data showed the “flashy” nature of the discharge and flow variation by two orders of magnitude. During intensive monitoring particulate P concentration varied by two orders of magnitude while all other fractions varied by about one order of magnitude. Flow was therefore more important in determining loads for all fractions except particulate P. The concentration of all fractions except soluble unreactive P were significantly related to flow. Loads and flows were generally best related by log load-log flow equations. The slopes of these equations indicated that soluble reactive P and nitrate N sources were diluted by low concentration water at high flows whereas particulate P sources increased at high flows due to erosion and surface run-off. Using log load-log flow equations developed from the discrete 8-day sampling data, loads were predicted and compared with loads for the same days as were intensively sampled. The predictive method based on discrete sampling overestimated soluble reactive P and nitrate N loads by 12 and 18% respectively but underestimated particulate P load by 43%. The error in predicted loads was due to discrete sampling frequencies being biased towards low flow conditions which prevail for the greatest proportion of the time. Since soluble reactive P and nitrate N concentrations decreased with increasing flow, predicted loads were therefore too high. The converse was true for particulate P whose concentration increased markedly with increased flow. Instead of designing a sampling programme based on a rigid time period between samples, it should be possible using river flow history for the months of the year to select sampling frequencies which would more closely approximate to flow-proportional sampling. For example, fixed-period 8-day sampling of the major rivers entering Lough Neagh might be replaced by monthly sampling in June, July and August, fortnightly sampling in April, May and September, and sampling every 5 days for the other 6 months.