3.4 Objective 4: implications for estimating the magnitude ofEwc across large catchments in complex terrain
The results from objectives 1 to 3 show that Ewc losses up to approximately 40 mm d-1 have been observed at temperate sites around the world and that meteorological conditions that have the potential to give rise to such large losses can exist in mountainous regions of the UK. However, these findings must be treated with caution because concurrent meteorological observations are rarely reported with CWB Ewc data, particularly during extreme events, and Penman-Monteith estimates are extremely sensitive toestimated aerodynamic exchange and small changes in RH at the higher windspeeds that often prevail during the large rainfall events considered here. The analysis has also shown that both wind speed and RH varies significantly with spatial location. Given that Ewc estimates are required for hydrological simulation of large catchments, a representation of this spatially variable control of Ewcmagnitude is required; it is not appropriate to sample a statistical distribution of Ewc loss generated from the worldwide observations ofEwc data (for a given gross rainfall total) as the autocorrelation of Ewc , controlled by autocorrelated meteorological variables, through sequences of real events is needed. With respect to this requirement, even a spatially sparse time series of meteorological observations contains important information describing temporal patterns of some of the primary controls onEwc and this information must be retained. Spatial interpolation and extrapolation from these sparse meteorological observations will inevitably be inherently uncertain but is an important prerequisite for appropriate estimation of Ewc losses.
Although simple empirical models can be used to estimate Ewc where there is a scarcity of adequate meteorological data and knowledge of appropriate parameter values for more complex models (e.g. see Lu, McNulty & Amatya, 1995), their use is limited as they may not explicitly include important meteorological controls. Consequently, the Penman-Monteith equation is still used to simulate evaporation from wetted surfaces in the majority of Ewc models (Muzylo et al ., 2009). Thus, Penman-Monteith equation remains a useful method to determine the potential for Ewc loss but the magnitude of any estimates made will be highly uncertain without meaningful calibration of critical and sensitive parameters such as \(r_{a}\_s\). However, as there are so few Ewc data associated with concurrent meteorological observations, particularly large rainfall events, it is rarely possible to calibrate the parameters of the Penman-Monteith equation and any calibration would need to include the joint-calibration of parameters of an (e.g. Rutter-type) effective canopy store model (e.g. see Calder, 1977).
Our theoretical analyses show that it is possible to get a very wide range of Ewc estimates depending upon, in particular, the way that \(r_{a}\_s\) is estimated. These analyses used 3 scenarios of\(r_{a}\_s\) which were based upon a range of published values derived both directly from micrometeorological observations and via model calibration. Ratios of \(z0\_s\)/\(z0\_m\) have been reported to be: of the order 0.1-0.2 (Klingaman, Levia, & Frost, 2007; Lankreijer, Hendriks, & Klaassen, 1993); approximately 0.3-0.5 (Brutstaert, 1982, p114; Stewart & Thom, 1973) and around 1 in some cases (Bosvelt, 1999; Gash, Valente, & David, 1999; Moors, 2012). Significant uncertainties exist when estimating \(r_{a}\_m\) and the relative magnitude of\(r_{a}\_s\) compared to \(r_{a}\_m\). When only momentum is considered, representing the degree of exchange is not simple as it has been shown to vary, and to be enhanced compared to theoretical estimates, in complex terrain and over tall canopies (Cellier & Brunet,1992; Holwerda et al ., 2012); \(r_{a}\_m\) also varies with canopy roughness and canopy density (Brutstaert, 1982, Fig. 5.1; Cellier & Brunet,1992; Holwerda et al ., 2012) as well as atmospheric stability and wind speed (Bosvelt, 1999; Cellier & Brunet,1992; Szeicz, & Endrödi, 1969). The ratio \(z0\_s\)/\(z0\_m\)also varies widely and with the same factors as \(r_{a}\_m\) and current understanding of scalar exchange for tall canopies in complex terrain remains rudimentary (Belcher, Harman & Finnigan, 2012). There are, however, a relatively large number of published studies which report \(r_{a}\_m\) and \(r_{a}\_s\) for various vegetation of differing roughness which may help elucidate the relevant range of\(r_{a}\_s\) for use in Ewc estimation for a given application: a review of these studies is, however, beyond the scope of this paper.
Given the need for interpolation and extrapolation from sparse meteorological data to estimate meteorological controls on Ewcspatially, uncertainties will be very large such that a scenario-based approach may be most appropriate. Any defined scenario will beconditional on the evidence base used in its development and any additional modelling assumptions. The conditionality of each scenario must be made explicit and each scenario can be associated with a confidence-weighting which can be propagated to simulation results. This will be the subject of future publications.