1. Introduction
Recently in the UK, and elsewhere, there is increased interest in ‘natural flood management’ (NFM) or ‘nature-based solutions’ for flood peak mitigation (Dadson et al ., 2017; Environment Agency, 2018; Jongman, Winsemius, Fraser, Muis, & Ward, 2018; Lane, 2017; Wingfield, et al ., 2019; World Bank, 2017). Tree planting may be one intervention that has the potential for flood peak reduction through: 1. increased soil infiltration capacity; 2. enhanced soil drying resulting from transpiration; 3. increased ground-surface roughness and 4. enhanced wet-canopy evaporation (Ewc ). However, some studies suggest that the positive effects of tree cover on flood peaks declines as event magnitude increases, such that it is likely to be insignificant for large and extreme flood events (e.g. Bathurstet al ., 2018; Dadson et al ., 2017; Robinson & Newson, 1986; Stratford et al ., 2017). These results suggest, implicitly, that Ewc is insignificant during large and extreme events.
For paired grassland and forest catchments on the Plynlimon massif (UK), Kirby, Newson, & Gillman (1991, p60) observed, using flood frequency analysis, that mature conifer cover had little or no effect on the magnitude of peak flows. They showed, using chronological pairingof flood peaks, that very small hydrograph peaks were consistently greater from the grassland catchment compared to the forested catchment and that moderately sized event hydrographs showed no significant difference. At another paired forest and grassland study at Coalburn, Northern England, Bathurst et al . (2018) reported that forests can reduce flood peaks for small to moderate events but that hydrograph responses tend to converge at extreme events. Bathurst et al.(2011) explicitly tested the hypothesis: as the size of the hydrological event increases, the effect of forest cover becomes less important ; they concluded, for a number of study sites across Latin America, that forests do not eliminate floods and are unlikely to reduce significantly peak flows generated by extreme rainfall. Bathurstet al. (2011) however acknowledged that their analyses were based on relatively short periods with few extreme events such that conclusive support for the test hypothesis is still lacking.
Recent NFM-related literature reviews of forest effects on flood peaks support the idea of a diminishing effect with event magnitude. Stratfordet al. (2017) carried out a systematic review of studies to answer the question: Do trees in UK-relevant river catchments influence fluvial flood peaks? Their review focussed directly on the magnitudes of flood peaks rather than on individual hydrological processes and they concluded that the evidence is uncertain for the impact of increasing tree cover on large floods but it is consistent in showing increasing tree cover reduces small floods. Dadson et al . (2018) also reviewed evidence of the effects of forest cover and reported the findings of a number of studies; they recognised that forest management practices complicate determination of forest effects but that under sustained winter rainfall, soil saturation will occur and little mitigation of high flood flows would be expected.
From a process point of view, the benefits of increased infiltration rates and drier antecedent soil moisture conditions are likely to diminish with increasing event magnitude (Calder & Aylward, 2006; Lull and Reinhart 1972; Pereiea, 1989); it is also likely that that boundary layer vapour pressure deficits, which exert a strong control onEwc , are likely to decrease during large and extreme rainfall events but the extent to which they decrease across large areas is not well known. The studies cited above did not explicitly included evidence from forest plot studies which estimate Ewc in a more direct way using a canopy water balance (described below), perhaps because only very few studies report Ewc for large or extreme events; they primarily considered the detection of hydrograph change from catchment studies globally.
Worldwide, catchment studies taken as a whole provide conflicting results regarding effects on large flood peaks; compare for example Jones & Grant (1996), Thomas and Megahan (1998) and Beschta et al ., 2000. Many studies have found that the magnitudes or frequencies of large flood peaks are changed significantly by afforestation or forest harvesting (e.g. Alila et al ., 2009; Belmar et al ., 2018; Fahey and Payne, 2017; Guillemette, Plamondon, Prévost & Lévesque, 2005; Jones & Grant, 1996; López-Moreno et al., 2006); many do, however, show decreasing effects on flood peak as event magnitude increases or no significant change (Beschta et al ., 2000, Birkinshaw et al ., 2014; Robinson & Newson, 1986; Thomas and Megahan, 1998; Whitehead and Robinson, 1993; Newson & Calder, 1989). Uncertainties associated with catchment studies of hydrograph change, particularly for extreme events, can be very large (Bathurst et al ., 2018; Beschta et al ., 2000; Carrick et al ., 2018; Dadson et al ., 2017). Underlying signals of change associated with specific processes (e.g. evaporation or infiltration) can also be obscured by effects resulting from forestry practices, such as road construction, drainage or harvesting method (Bathurst et al ., 2018; Beschta et al ., 2000; Jones & Grant, 1996; Guillemetteet al ., 2005; Robinson & Newson, 1986; Thomas and Megahan, 1998). These factors combined with limited observations for extreme events (Lewis, Reid & Thomas, 2010) mean that simple conclusions regarding forest effects on large or extreme flood peaks cannot be made (Andreassian, 2004; Carrick et al . 2017).