STATISTICAL BEHAVIOUR OF LOAD ESTIMATORS BASED ON ROUTINE MONTHLY DATA SERIES

摘要

Irrigation contributes to the pollution of water bodies through the pollutant loads in the irrigation return flows. Establishing the relationship between changing irrigation and agricultural practices and pollutant loads over long periods may help toidentify the irrigation-related factors that most affect water quality.This paper aims to ascertain the statistical performance of 5 salt and nitrate load estimators based on the long-term monthly records of the surface water quality monitoring network (SWQ) of the Ebro Basin Authority (CHE). These estimates were compared with daily estimates in the Arba River monitoring station at Tauste (taken as reference loads), included in the newer irrigation return flows network (ReCor-Ebro; R-E) during April 2004 to September 2010.Three estimation methods used grab-samples monthly 7DS,- and N03l from the SWQ network (multiplied by instant, Q/, mean daily, Qd, or monthly, Qm, flows), whilst the other two were the product of the regression estimates of TDS and N03 from Qd by Qd or Qm. The instant concentration-based models were also tested with daily data from the R-E network, with more complete records. The regression estimators performed better than the models based on instant samples for salt loads. But for nitrogen loads, the estimators based on N03i and Qd or Qm also performed well when drawing data from the more complete R-E data series. Although the biases for the 5 methods were not significant; only these estimators presented errors low enough to allowtheir use in generating reliable load time series.