1. Testing of GW-quality data from subsequent surveillance monitoring for a significant increase
Proposal developed by Umweltbundesamt and quo data (subcontractor) for the
Federal Ministry for Agriculture, Forestry, Environment and Water Management (Austria)

2. Background information (1)
Trend assessment - Proposal of WG 2.8
Minimum length of time series for the detection of an upward trend
Annual data: Length >= 8 years and >= 8 values
Half-yearly data: Length >= 5 years and >=10 values
Quarterly data: Length >= 5 years and >=15 values
this is practical for operational monitoring
in case of surveillance monitoring: 42 years would be needed

3. Background information (2)
WFD, Art. 4: ... to prevent the deterioration of the status of all bodies of groundwater ...
in case of surveillance monitoring: criteria for deriving activities at an earlier stage than 42 years would be needed

4. Proposal (1)
Testing of GW-quality data from subsequent surveillance monitoring activities for a significant increase at the GW-body level
Starting point:
There is random variability in the annual mean (not larger than 30%, say)
Conclusion:
Trends <30% cannot be assessed by subsequent surveillance monitoring!

5. Proposal (2)
Idea:
No action is required if it can be proven statistically that the actual increase is below 30 %. It may be concluded that there is random variability only - no action required
if it cannot be proven statistically - action is required
application of precautionary principle: burden of proof is with the monitoring manager
Formal test hypotheses:
according to null hypothesis it is assumed that the actual increase is at least 30%; i.e. that there is a systematic trend
the null hypothesis will be rejected only if the increase is significantly below 30 % (or below another appropriate limit)

6. Proposal (3) Proposed methods Analysis of site-wise level differences
in case of temporal correlation within the sites, the hypotheses may be tested using the relative site-wise level differences
Comparison of the spatial mean
The hypotheses can be tested with the test statistic AM1/AM0, where AM1 and AM0 denote the spatial mean in the sampling periods t=0 and t=1, respectively.

7. Proposal (4)
Both methods consider number of sampling sites and variability within the GW-body
actual increase > 30% decision for action (with both methods)
actual increase % decision depends on number of sampling sites and variability within the GW-body
e.g. for poor network > 10 % decision for action might be possible (depends on parameters)
for good network > 20 % decision for action might be possible (depends on parameters)

8. Proposal (5) Proposed procedure:
test of subsequent surveillance monitoring data
in case that the increase is not significantly below the set limits => action is required - operational monitoring to be introduced
the proposed test does not replace a trend assessment

9. Current state Currently both methods are tested
with Austrian data and data from CIS WG 2.8
for various GW-body and parameters
results and formal description of methods can be provided before the next EAF-GW
it seems that for some parameters e.g. pesticides due to high variability 40 or 50 % seem to be adequate