3.1 Objective 1: canopy water balance observations of wet-canopy evaporation during large and extreme events
The data search provided observations from 18 study sites that have CWB observations of Ewc for large storms associated with either daily or normalised event 𝑃𝑔 (see Table Supp. 3): none of the events were extreme events based upon the classification employed (>150 mm d-1). From these sites 1387 𝑃𝑔-Ewc pairs were obtained with a maximum 𝑃𝑔 of approximately 118 mm. Only 35 of these pairs were associated with 𝑃𝑔 observations over 50 mm d-1 (Table Supp. 3). In absolute terms, the Ewcdata include some high magnitudes; they include a maximum Ewcloss of 39.4 mm d-1 with 40 events where Ewcloss was over 10 mm d-1. When Ewc is expressed as a percentage of 𝑃𝑔 (%Ewc ), there is a decreasing trend in %Ewc as 𝑃𝑔 increases (Figure 2). This pattern of declining relative loss has been shown previously in many studies (e.g. Iroume & Huber, 2002; Bulcock & Jewitt, 2012), including the seminal review by Horton (1919). From these 1387 data pairs, 58 gave a negative %Ewc and 25 were over 100 %Ewc ; these data lie predominantly at low 𝑃𝑔 magnitudes and are not presented in Figure 2 and can be caused by error, sampling truncation of events and fog-drip and will be incorporated into uncertainty analysis in future work. The subset of the data from UK catchments are overlain by green filled-circles in Figure 2; the individual numbered events where 𝑃𝑔 > 50 mm are specified in Table Supp. 3. These UK data tend to span the higher rates of %Ewc for higher 𝑃𝑔 magnitudes and include absolute losses up to 26.3 mm d-1. The few data which are associated with plots under entirely deciduous species are highlighted as orange-filled circles in Figure 2: only one data pair was associated with 𝑃𝑔 greater than 50 mm per day of which approximately 11% was lost to Ewc . The degree to which %Ewc continues to decrease with increasing 𝑃𝑔, or whether it has reached a stable range is unclear from Figure 2 given the few data available at higher 𝑃𝑔 magnitudes. This is important as the highest 𝑃𝑔 from these data is significantly lower than extreme daily rainfall totals recorded in the Cumbrian Mountains which has been observed to be as high as 341 mm in a 24-hour period at Honister Pass, Cumbria (Met. Office 2018) which led to widespread severe flooding.
Observations associated with studies where no event duration data were available provided 1144 𝑃𝑔-Ewc pairs including some high %Ewc losses (ranging from approximately 7 to 35 %) for very large 𝑃𝑔 values (up to 435 mm: see Table Supp. 2). These data are plotted as black filled-circles in Figure 3 with plots under entirely deciduous species highlighted as orange-filled circles. Although these data are difficult to compare to the daily or normalised daily data, the high losses observed are significant given that these are potentially extreme events, likely to be of maximum 3 or 4 days in duration and hence serve as a useful reference. An exception to this rule are the data from Deguchi, Hattori, & Park (2006) where it is possible that observations were made over a period of up to 2 weeks; however, the largest event from their study (which was identified as taking place on September 11-12th, 2000) deposited 347 mm 𝑃𝑔, with 14% of this being lost to Ewc . The magnitude of the losses from these non-normalised events are qualitatively consistent with the normalised events presented in Figure 2.
The data for large and extreme events presented in Figures 2 and 3 show a large range of Ewc loss: approximately 2-38 % of 𝑃𝑔. These losses are apparently significant in the context of flood mitigation with absolute losses of up to approximately 40 mm d-1. Unfortunately, concurrent meteorological observations were generally not reported for events greater than 50 mm d-1: concurrent observations were only available for 4 events at one site (Dolydd, mid-Wales; events 7-10, Table Supp. 3). This lack of meteorological data means that it is, in general, not possible to link the observedEwc losses with the magnitude of important meteorological variables which would allow some form of model calibration. Model calibration of this kind is problematic and also needs to include canopy storage limitation of Ewc (see Calder, 1977).