Setting extreme weather thresholds
With the exception of the index relating to drought, thresholds were set
based on deviation from the mean value of the whole dataset for each
grid point. Maximum daily temperature or rainfall above the
95th percentile and minimum daily temperatures below
the 5th percentile were considered extreme (World
Meteorological Organization, 2017a[49].
Temperature and rainfall conditions are spatially variable across the UK
and utilising percentiles as the threshold instead of a fixed value
allows for regional variation in normal conditions. What is considered
an extreme temperature or rainfall amount in one location may be
relatively normal for another and it is likely that the largest impact
on ecosystem function occurs when conditions are outside the norm rather
than at a fixed value (Zhang et al., 2011).
Using this approach, the following thresholds were proposed as an event
metric for extreme heat, cold and rainfall based on recommendations
provided in the draft guidelines on the definition and monitoring of
extreme weather and climate events produced by the World Meteorological
Organization (WMO) (World Meteorological Organization, 2018a).
Heat: The number of times each year where the 5-day rolling mean
of the maximum temperature exceeds the 95th percentile
of the whole dataset for 3 or more days.
Cold: The number of times each year where the 5-day rolling mean
of the minimum temperature is below the 5th percentile
of the whole dataset for 3 or more days.
Extreme rainfall: The number of times each year where the 5-day
rolling mean of the daily rainfall total is above the
95th percentile for 3 or more days.
Low rainfall: The number of times each year where the 5-day
rolling mean of total daily precipitation was below 1 mm for 14 days or
more, based on a historical definition of agricultural drought used in
Britain of rainfall below 1 mm for more than 15 days (Heim et al.,
2002).
For this study, extreme rainfall was used as a proxy for flood risk.
While it is recognised that flood generation encompasses many complex
variables, including the hydrology and topography of the landscape, we
focus on rainfall totals as an indicator of the change in risk
potential. Daily rainfall totals in the preceding 0 to 3 days was shown
to be the best predictor of river flood events across the Swiss Alps
(Froidevaux et al., 2015). In the UK the total rainfall over 3 days was
linked to 40 year maximum peak river discharge and recorded flood events
in 3 out of 4 studied river catchments (Lavers et al., 2011). In China,
persistent extreme precipitation events, considered to indicate high
damage potential were defined as daily precipitation total above 50 mm
for 3 or more days (Chen and Zhai, 2013). Similarly to the flood index,
we used rainfall as a proxy indicator for drought risk. However,
prolonged dry spells, rather than a deviation from the minimum rainfall
long-term average are likely to be more significant in reducing soil
moisture content and increasing risk of drought.