Spatial and temporal changes in the structure of groundwater nitrate concentration time series (1935–1999) as demonstrated by autoregressive modelling

Abstract Autoregressive modelling is used to investigate the internal structure of long-term (1935–1999) records of nitrate concentration for five karst springs in the Mendip Hills. There is a significant short term (1–2 months) positive autocorrelation at three of the five springs due to the availability of sufficient nitrate within the soil store to maintain concentrations in winter recharge for several months. The absence of short term (1–2 months) positive autocorrelation in the other two springs is due to the marked contrast in land use between the limestone and swallet parts of the catchment, rapid concentrated recharge from the latter causing short term switching in the dominant water source at the spring and thus fluctuating nitrate concentrations. Significant negative autocorrelation is evident at lags varying from 4 to 7 months through to 14–22 months for individual springs, with positive autocorrelation at 19–20 months at one site. This variable timing is explained by moderation of the exhaustion effect in the soil by groundwater storage, which gives longer residence times in large catchments and those with a dominance of diffuse flow. The lags derived from autoregressive modelling may therefore provide an indication of average groundwater residence times. Significant differences in the structure of the autocorrelation function for successive 10-year periods are evident at Cheddar Spring, and are explained by the effect the ploughing up of grasslands during the Second World War and increased fertiliser usage on available nitrogen in the soil store. This effect is moderated by the influence of summer temperatures on rates of mineralization, and of both summer and winter rainfall on the timing and magnitude of nitrate leaching. The pattern of nitrate leaching also appears to have been perturbed by the 1976 drought.