Step 9 First step in the optimization
5.8 Discussion and conclusions
This study aimed at improving (1) the detection and (2) the understanding of groundwater quality changes in time in large areas with homogeneous land use, soil types and
geohydrological situation. In the presented approach, a distinction is made in trend detection and trend understanding (Figure 5.2). The trend detection uses measured data of time series and time-averaged concentration-depth profiles. The interpretation of trends was aided by prognoses of conservative and reactive transport using tritium age dating and information on the historical loads of solutes in recharging groundwater. The two aspects are discussed separately, and implications for groundwater quality monitoring strategies are indicated briefly.
134 de pt h (m ) 0 20 60 80 K (mg/l) 0 5 10 15 20 25 30 1998 1992 1986 increasing 1992-1998 40
Figure 5.14 - Model reconstruction of the downward movement of the reactive front of potassium between 1986 and 1998 based on the reactive prognosis. Depth interval with increasing trends over the period 1992-1998 is indicated.
Trend detection
The complementary use of time-series information and time-averaged concentration-depth profiles yields additional value compared to the individual methods. Time series analysis alone might fail to detect groundwater quality changes reliably because non-conservative behaviour limits the breakthrough of the solutes. Concentration-depth profiles reveal this non-
conservative behavior and indicate the position of contamination fronts. The detection of groundwater quality changes is supported by comparing significant trends in time series at the monitoring depths with the vertical concentration profiles.
Especially, the use of conditionally conservative indicators helps to identify groundwater quality changes, because their pollution front moves through the aquifer with groundwater velocity. The use of conditionally conservative indicators is advantageous because trends are detected earlier, directly after the advective travel time has passed. This was demonstrated for the two homogeneous areas, where many trends in the time series of OXC and SUMCAT were found, which indicate the downward movement of the agricultural pollution front. No temporal trends in nitrate have been detected, because nitrate disappears with depth and the pollution front of nitrate probably moves too slowly to be detected using the regular monitoring depths.
Understanding of observed trends
The understanding of quality changes is strongly supported by comparison of the position of the pollution fronts and the temporal trends from time series with prognoses of conservative and reactive transport. These prognoses require information on the historical inputs of the contaminants and an indication of the groundwater age.
An unknown uncertainty is inherent to the calculated historical inputs. The uncertainty mainly arises from the estimates of crop uptake, the distribution of the produced animal manure over the various land uses and spatial variations in the use of manure. However, the uncertainty is reduced by the fact that we used regionally averaged data on manure production, fertiliser use and atmospheric deposition for the whole province of Noord-Brabant. This results in the averaging out of local deviations in the mineral surplus. No large uncertainties in the order of magnitude and the period of increasing and decreasing inputs are expected for the investigated components. These two aspects are the most important for understanding the depths and magnitudes of temporal trends. The uncertainty in the groundwater age distribution was evaluated using the median, 10 and 90 percentiles of ages derived from the tritium measurements. This part of the uncertainty of the prognoses was visualised and used for the data analysis.
For the Noord-Brabant case, the prognoses for conditionally conservative indicators correspond much better with the measured concentrations than the prognoses for individual targeted contaminants, such as nitrate, sulphate and potassium. In the Noord-Brabant homogeneous areas, the downward movement of the agricultural pollution front was demonstrated by the increasing trends between 18 and 25 m depth for OXC and SUMCAT. The 1985 peak of OXC and SUMCAT has not yet arrived at the deeper screens and further increase of concentrations is expected. No trends for these indicators were found at 7-12 m depth, which is plausible because the contaminant peak was predicted to be within this interval, with decreasing concentrations above it and increasing concentrations below it.
The large misfit between the predicted conservative position of the nitrate and potassium front and the actual position of the fronts indicate retarding and/or attenuating processes in the aquifer. For potassium the retarded front was modelled using a simple 1D geochemical model of cation exchange between potassium and calcium and magnesium. The model indicates that the increasing concentrations of potassium between 7 and 12 m depth are well
explained by the movement of a retarded potassium front. This indicates that concentration changes of the retarded species are expected and demonstrated at shallower depth than concentration changes of the conservatively transported indicators.
No significant quality changes have been demonstrated for nitrate, due to nitrate reduction by oxidation of pyrite and organic matter. The significant changes of OXC in the deep groundwater that lacks nitrate are due to increasing sulphate concentrations, which is a reaction product of the pyrite oxidation. A future increase in sulphate concentrations above the drinking water standard of 150 mg l-1is anticipated, given the loads of nitrogen and sulphate
that are still underway.
Implications for groundwater quality monitoring
Often, the design of groundwater quality monitoring networks implicitly assumes conservative transport, and sampling depths and monitoring frequency are adapted to advective transport velocities. In the Netherlands’ monitoring networks, regular monitoring depths of
approximately 10 and 25 m depth, and an annual frequency were chosen. Using data from these networks, we were able to demonstrate increasing concentrations at both depths for chemical indicators that behave conservatively. However, large concentration gradients with depth are present for many targeted contaminants, which indicate retarding or attenuating processes within the first 20 m of the subsurface. Groundwater quality changes of these components are expected to occur at more shallow depth. For potassium, this was indeed demonstrated in the study.
The concentration-depth profiles of the two homogeneous areas show a lack of data between 12 and 20 m depth, because the monitoring screens at this depth are not sampled frequently. This results in larger uncertainty of the LOWESS smooth at this depth. The concentration- depth profiles would clearly benefit from sampling these intermediate screens, because this would better define the vertical position of the contamination fronts (section 5.8). Since many concentration changes are expected between the two regular monitoring depths, one could consider to sample all available screens in the wells annually to obtain time series information. Alternatively, and cheaper, we advise the sampling of the complete concentration-depth profile every 4 years, including the shallow screens and the screens between 12 and 20 m depth. This would allow for a four yearly evaluation of the vertical propagation of the conservative and reactive contamination fronts.
General conclusions
The combined use of time series information and time-averaged concentration-depth profiles helps detect temporal changes in groundwater, especially when both potentially reactive and conditionally conservative indicators are used in the data analysis. Interpretation of the trend analysis results is aided by comparing the depths of the demonstrated temporal trends with prognoses of conservative and reactive transport, using information on the historical inputs of the contaminants and age dating using tritium.
6
Evaluating monitoring strategies for groundwater quality at phreatic well fields: a 3D travel time approach6.1 Introduction
Phreatic well fields used for public water supply are vulnerable to groundwater contamination originating from diffuse sources, such as agricultural pesticides and nutrients. Usually protection zones are established to regulate the use of pesticides, fertiliser and manure and to protect the groundwater resource from calamities of other origin. The effectiveness of the groundwater protection and the threats of contaminated groundwater from outside the protection zone can be assessed using an adequate groundwater quality monitoring network. However, groundwater quality monitoring is expensive because it requires the frequent sampling and analysis of many samples and installation costs of wells are high. An effective monitoring strategy should therefore yield valuable information at justifiable costs.
Groundwater quality monitoring by water supply companies is done for five distinct objectives (after Baggelaar 1996):
1. to fulfil the legal monitoring necessity 2. to reassure customers
3. to signal unexpected or new threats to the quality of extracted groundwater 4. to support operational decisions by the prediction of future quality changes, and 5. to evaluate protection measures in the protection zone.
In the Netherlands the first two monitoring objectives are covered by the three-monthly analysis of the macro chemistry of the extracted groundwater at the pumping well and the monthly analysis on Coli-bacteria (Baggelaar, 1996). Especially the last three monitoring objectives require a specific observation network, apart from the pumping wells. In the following the monitoring objectives 3, 4 and 5 will be referred to as early warning,
prediction and protection, respectively. The design of an adequate monitoring network requires
the evaluation of the hydrogeology of the area, the expected threats from agricultural or atmospheric origin and the possible hydrogeochemical and microbial processes that attenuate or worsen the groundwater quality evolution (Baggelaar 1992, Stuyfzand 1996).
Baggelaar (1992, 1996) designed an overall monitoring strategy for the Dutch water supply companies. For phreatic well fields the recommended monitoring configurations were: 1. the monitoring of the extracted groundwater at the pumping well
2. the monitoring of groundwater in the pumped aquifer that is 10 to 15 years travel time from the well, and
3. the monitoring of shallow groundwater in the protection zone and in a control area. Several adaptations and improvements to this strategy have been proposed during the last five years (Baggelaar 1996, Zhang 1996, Foppen & Kremers 1997, Hetterschijt & Foppen 1998). None of these strategies have evaluated the 3D travel time distribution in the aquifer in order to design an appropriate network.
The objective of the present study was to judge the effectiveness of different network configurations, using a 3D travel time approach, evaluating advective and simple reactive transport. The selected monitoring configurations are relevant for typical Dutch conditions, which include shallow groundwater tables and unconsolidated aquifers in a flat landscape. The study concentrated on partially penetrating wells in an unconfined situation, since most phreatic well fields conform to this situation. Since the concern was the monitoring of the quality of the groundwater, the water flow and degradation processes in the unsaturated zone
were neglected, although processes in the unsaturated zone may be important for the break- through of specific solutes such as pesticides at the pumping well (Beltman et al. 1996).
Numerical simulations of groundwater flow and travel times were made to evaluate seven monitoring configurations for pollution scenarios with advective transport, first-order
degradation and linear sorption, using a block input. Existing concepts and analytical solutions for travel time and solute breakthrough of fully and partially penetrating wells are introduced first.
6.2 Analytical solutions for the travel time distribution and solute breakthrough