Effects of sampling frequency and calculation methods on estimation of annual nutrient loads: A case study of Manawatu River, New Zealand

Abstract

<abstract> <bold>Abstract.</bold> The estimation of nutrient loads is of great importance in the assessment and management of water quality. Streamflow in most rivers across New Zealand is measured continuously at high frequency (minutes or hours) but nutrient concentrations are measured less frequently (weekly or monthly) because of the high cost of sampling and analysis. A number of methods have been developed to estimate annual nutrient loads in streams and rivers using the available limited nutrient concentration data. Infrequent sampling and the differences between these methods are the main causes of uncertainty in estimates of annual nutrient loads. The objective of this study was to evaluate the uncertainty in nutrient load estimates caused by combinations of different sampling frequency and load estimation methods. We used daily measurements of three nutrient parameters (nitrate-nitrogen "NO₃-N", soluble inorganic nitrogen "SIN", and dissolved reactive phosphorus "DRP") in the Manawatu River at Palmerston North for the period from May 2010 to April 2011. Three sampling frequencies (weekly, fortnightly, and monthly) were considered in combination with five calculation methods (Global Mean, Rating Curve, Ratio Estimator, Flow-Stratified, and Flow-weighted) to estimate annual nutrient loads in the river. The estimated annual nutrient loads were compared with the annual nutrient loads calculated using the daily measured streamflow and nutrient concentrations data in the river. When compared with the annual nutrient loads calculated using the daily data, the bias increased with the decrease of sampling frequency (from weekly to fortnightly) for all methods except the Global Mean (GM) and the Rating Curve (RC). However, from fortnightly to monthly sampling frequency the bias values decreased for all methods except GM and RC. The root-mean square error (RMSE) increased with the decrease of sampling frequency in all methods except GM and RC. The GM method consistently underestimated while the RC method overestimated the annual nutrient loads. The Flow-Stratified (FS) performed better for weekly sampling whereas the Flow weighted (FW) and the Ratio Estimator (RE) performed better for fortnightly and monthly sampling frequencies. A combination of weekly sampling and the use of FW, RE and FS to estimate annual loads of NO₃-N, SIN and DRP yielded the lowest bias and RMSE in the estimated annual nutrient loads.