• No se han encontrado resultados

Propuesta: Un modelo de CVAT para apoyar la gestión del turismo

Conclusiones y propuesta

5.2. Propuesta: Un modelo de CVAT para apoyar la gestión del turismo

In the evaluation methodology for the joint Dutch provincial monitoring networks, the monitoring depths and screen lengths were considered fixed. Changing these would require a completely new network set-up and high installation costs. However, more effective

monitoring and cost reduction were achieved using an area-specific differentiation of monitoring frequency at the different monitoring depths.

The existing monitoring frequency, monitoring depths and screen lengths of the network are tuned to the average vertical flow velocities of recharging groundwater. The Dutch networks consist of screens of 2 m length at about 9 and 24 m depth that are sampled annually (for details see Chapter 3). Using a downward velocity of 1 m per year and a screen length of 2 m and assuming strictly horizontal inflow during sampling, the sample contains the water of approximately two years of recharge (Meinardi 1994). Since flow velocity decreases with depth (Chapter 2) and concentric flow will occur during sampling, one expects a sample to be a mix of about three or four years of recharge. Mixing 3 or 4 years recharge water averages out short term fluctuations in groundwater quality. Consequently, a yearly sampling frequency is sensible and gradual changes in the sampled groundwater are anticipated (Meinardi 1994, Baggelaar & van Beek 1997).

Following the monitoring information goals listed in Table 3.4 (Chapter 3), the detection of temporal changes in groundwater quality was considered relevant only for high and moderate- risk areas. Corresponding statistical information goals for high-risk, moderate-risk and low-risk areas are given in Table 4.5. For the moderate-risk areas, the signalling of trends was sufficient, whereas higher ambitions are pursued for high-risk areas. Therefore, a differentiation of monitoring frequency according to monitoring ambitions is quite possible.

The information about typical concentrations and proportions of contaminated ground- water is required in reports on the state of the environment in order to provide an overview of groundwater chemical status. These reports have a typical publication frequency of about once every 4 years. Therefore, using a monitoring frequency of once every four years was considered sufficient to meet the monitoring information goals A to D of Table 3.4. The detection of changes of groundwater quality with time (statistical information goals E, F and G of Table 4.5) requires a higher, annual monitoring frequency (see below).

Criterion E

Criterion E refers to the probability of detection of temporal trends in individual wells. Here, a trend was defined as a change in groundwater quality over a specific period in time which is

Monitoring information goal High-risk Moderate-risk Low-risk

areas areas areas

E Probability of detection of temporal trends x

F Precision of median temporal trend x

G Precision of trend in proportion of contaminated groundwater x

related to land use or water quality management (Loftis 1996). The probability of detecting such a trend depends on a large number of unknowns: the type of trend (monotonic, step trend), the magnitude of the trend, the natural temporal variations of the observations, the monitoring period and the type of statistical trend test used (Loftis 1996, Baggelaar & Van Beek 1997, Burn & Hag Elnur 2002, Yue et al. 2002).

In general, trend detection is more difficult if the trend is small relative to the natural temporal variations and the number of independent observations is small. In trend analysis of groundwater quality data a monotonic trend type is often assumed (for example: Pebesma 1996, Frapporti 1993, Reijnders et al. 1998, Baggelaar & Van Beek 1997). Because the probability distribution of the concentrations is unknown, Baggelaar and Van Beek (1997) recommended the use of non-parametric tests for the trend analysis. However, they used parametric methods to evaluate the probability of detection using examples from the Dutch regional networks. Using linear regression and the assumption of log-normal distributed data in a time series, they showed that the probability of detection decreases strongly when the number of observations decreases from 9 to 5 in a ten-year monitoring period. This corresponds with the experience that trend detection in individual wells using non-parametric methods has never shown significant trends for annual time series with less than 7 years (Broers 1996, see also Chapter 5). Therefore, time series of at least 10 observations were considered the minimum for significant trend detection, which requires 10 years of monitoring in the present set-up. Accordingly, decreasing the monitoring frequency from once every year to once every two years is no option if the aim is to detect trends over a 10 year monitoring period.

Increasing the monitoring frequency was also not considered sensible, because serial correlation between the observations will increase (Loftis 1996, Baggelaar & Van Beek 1997). Already with a monitoring frequency of once every year, substantial serial correlation is expected because of the slow downward velocity of groundwater relative to the length of the well screens (maximum 1 m per year versus 2 m screen length).

Criteria F and G

The criteria F and G of Table 4.5 refer to temporal trends in homogeneous areas. The trend definition of Loftis (1996) was adopted, which defines a temporal trend as a change in ground- water quality over a specific period in time, over a given region, which is related to land use or water quality management. Note that this trend definition is different from that commonly used in geostatistics. The probability of detecting trends in areas is dependent on the five factors mentioned under criterion E, plus the spatial variability in the area under consideration.

The statistical information goals F and G were considered relevant only for the high-risk areas for which the aim is to determine the median trend and the trend in the proportion of

contaminated groundwater in the homogeneous areas. These information goals strongly depend on the precision of the estimates of the median concentrations and the proportions of

contaminated groundwater in the homogeneous areas. These were evaluated using criteria A and

94

Risk assignment High Moderate Low

Ambition level High Moderate Low

Frequency shallow screens 1 yr-1 1 yr-1 1 (4 yr)-1

Frequency deep screens 1 yr-1 1 (4 yr)-1 1 (4 yr)-1

D, which were considered required first steps in the evaluation of the networks. In Chapter 5, examples of trend detection are given for two homogeneous areas in Noord-Brabant.

Step 7 Evaluating monitoring frequency, monitoring depths and sets of chemical