sampling variability can affect the quality of the index flood
model. As an illustration, µi(D = 0) was calculated for vhm148 considering 15-year moving
windows (Fig. 10). The effect of sampling variability on the estimation of µi(D = 0) appears
very clearly. The observed AMF series at the eight gauged sites are of different lengths and do
not always overlap in time, which adds uncertainty to µi(D
/media/vedurstofan/utgafa/skyrslur/2015/VI_2015_007.pdf
the preceding time period during which WT-
occurrences influence streamflow, was identified by a correlation analysis between the daily
RDAI series and the total frequency of potentially drought supporting WTs over different
preceding time windows. Potentially drought supporting WTs were identified based on
composite maps of average precipitation amounts for each of the 29 WTs. dreg was defined
/media/ces/ces-oslo2010_proceedings.pdf
based on accumulated
precipitation over running 24-hour windows. Both maps include important details that the earlier
version could not encompass. Higher return levels are found on the Snæfellsness and Tröllaskagi
peninsulas, the Bláfjöll mountainous region, as well as in the East- and Westfjords. Those new
results have several potential uses, including thresholds for extreme precipitation
/media/vedurstofan-utgafa-2020/VI_2020_008.pdf
and other characteristics.
Because of time limitation, the analogy domain and temporal windows were not strictly speaking
optimized. The following tests were performed:
- Analogy domain 1 (AD1): 60–70 N, 35–5 W.
- Analogy domain 2 (AD2): 55–75 N, 40–0 W.
- Temporal window 1 (TW1): 45-day window centered on the predicted calendar day.
- Temporal window 2 (TW2): 30-day window centered
/media/vedurstofan/utgafa/skyrslur/2014/VI_2014_006.pdf