Figure 8 shows the scatter plot of the turbulent deposition velocity determined from each probe for cases 1 and 2 .sx The regression line was formula .sx The standard error of the coefficient was 0.218 , while the correlation coefficient was 0.80. graphs&captions .sx ( b ) Turbulence data from the sonic anemometer .sx To compare deposition velocities from a variety of sites and canopies with different wind speeds , v t was normalised with v m as suggested by Slinn ( 1982) .sx For the calculations presented here we have used the in-situ measurements of u * and wind speed from the sonic anemometer .sx These were compared with u * values obtained from the profile mast which allowed values of the zero-plane displacement and roughness length to be determined .sx This was done to provide an estimate of the reliability of point estimates of fluxes of such canopies .sx The profile information was also used to show whether the turbulence was in local equilibrium with respect to the canopy .sx This is important to determine if such measurements are to be compared with model results of cloud deposition to hills in complex terrain as discussed by Gallagher et al. ( 1990) .sx As discussed later , good agreement was found between the data from the sonic and profile mast , where a value of d of approximately 2h/3 , where d is the zero-plane displacement and h is the canopy height , was used for computing v m from the profile data .sx The relationship between d and h is consistent with a general consensus from a large number of studies , for example Jarvis et al. ( 1976 ) ; however , during one case study , when a kink developed in the wind profile , the value of d was uncertain and agreement between sonic and profile mast data required a larger value of d to be used .sx During case 1 the mean friction speed was 0.37 unch 0.05 m s -1 , with a mean wind speed at the measurement height ( 9.03m ) of 2.73 unch 0.36 m s -1 .sx This yielded a value for the canopy roughness length , z 0 , of 0.30 unch 0.09 m ( z 0 ) .sx For case 2 the observed values were u * =0.70 unch 0.19 m s -1 with a wind speed of 6.12 unch 0.35 m s -1 .sx This yielded a somewhat smaller value for z 0 of 0.17 unch 0.04 m ( z 0 ) .sx This was probably due to a slight change in wind direction .sx ( c ) Wind profiles above the canopy and comparison with the sonic anemometer data .sx Wind profiles measured during case 1 ( Fig. 9 ) showed the presence of a large kink at 5 m above the zero-plane displacement with a significant reduction in wind speed at this level .sx Since the profile is not logarithmic , the measurements here must be considered as point flux measurements whose representativeness is open to debate .sx However , the profile data obtained during case 2 ( Fig. 10 ) showed much closer agreement with a logarithmic profile indicating that the fetch during this period was more ideal , with the turbulence in equilibrium with the underlying canopy ( at least up to the level of the USAT) .sx The wind speeds during case 2 were very much higher than during case 1 , with no sign of a kink at the 5 m level above the zero-plane displacement .sx The profile anemometers did not overlap the USAT and a detailed comparison between flux profile and eddy-correlation-derived u * is difficult for case 1 where it would appear that the USAT was sampling a different turbulent boundary layer from most of the profile anemometers .sx The shape of the profile from case 1 is characteristic of flow from rough to smooth terrain , with a sharp transition between the equilibrium boundary layer and the outer boundary layer .sx If a similar process has caused the kink at 5 m above the zero-plane displacement , then the air in the inner boundary layer would accelerate slightly , resulting in horizontal divergence and subsidence of the air above the inner layer .sx If one assumes that the upper two points of the profile are representative of rougher terrain upwind then a roughness length quite close to that measured with the USAT can be calculated .sx In contrast the inner-layer data produced a roughness length of 0.13 m , less than half that derived from the USAT in the outer layer .sx This is consistent with the decrease in canopy height up the hill .sx The fluxes derived at the measurement site are thus more characteristic of the upwind terrain .sx graphs&captions .sx Since the canopy height changes significantly over the hill it is difficult to derive a value for the zero-plane displacement for the upwind terrain .sx The variables quoted above have all been derived assuming a value of d=2 .sx 81 m which corresponds to the local canopy height .sx The profile measurements indicate a local zero-plane displacement between 3 and 3.5 m corresponding to a d/h ratio of between 0.71 and 0.83. The profiles from case 2 showed close agreement with a logarithmic wind profile , with a correlation coefficient of 0.98. The zero-plane displacement was again estimated to be between 3.0 and 3.5 m with a roughness length of 0.08 to 0.14 m. Applying the profile-derived d values to the USAT measurements yielded a roughness length of 0.15-0.17 m. Because of the uncertainty in obtaining good estimates of d , absolute comparison between the USAT and profile anemometers is limited .sx For case 2 , using d=3 .sx 5 m , good agreement between profile-derived and eddy correlation u * was obtained with the ratio formula with a correlation coefficient of 0.98. Data obtained during other periods of thin cloud during case 2 produced values of z 0 varying from 0.25 unch 0.10 to 0.29 unch 0.10 m ( z 0 /h=0 .sx 06 to ) .sx The changes in roughness length probably resulted from small changes in wind direction which could considerably alter the fetch at this measuring site .sx The observed roughness lengths are generally smaller than those quoted in the literature .sx Unsworth ( 1984 ) presented data from a forest of Pinus sylvestris and Pinus nigra var .sx maritima with a canopy height of 15.5 m ; d/h was 0.76 and z 0 /h was 0.060 ( z 0 =0.93 m) .sx Jaeger ( 1985 ) measured wind profiles above a growing Scots Pine forest ( Pinus sylvestris ) over a period of ten years and obtained linear regression estimates of d and z 0 as a function of stand height h. Using his relationships we obtain d=2 .sx 42 m and z 0 =0.96 m. Both the profile and sonic-derived z o values are significantly smaller than those predicted by Jaeger , as is also the case with the data from Unsworth .sx The scatter in Jaeger's data was , however , considerable ( ) .sx Our z 0 /h values are within the scatter of results reported elsewhere in the literature , for example Brutsaert ( 1982) .sx It is debatable whether the concept of zero-plane displacement is at all applicable in the context of this work owing to the uncertainties posed by use of the gradient technique when used over non-ideal terrain .sx In particular , recent work over forest canopies .sx Duyzer et al. ( 1990 ) , tends to suggest that the normal flux-gradient relationships for momentum and moisture vapour may need to be modified close to the canopy , otherwise fluxes may be significantly over-estimated by as much as 40% .sx For the experimental site described here , the uncertainties in the zero-plane displacement would lead to an over-estimate of fluxes , in some wind directions by , at worst , 30% .sx ( d ) The turbulence coefficients and atmospheric stability .sx Typical values of turbulence coefficients A , B , formula were 4.23 unch 1.06 , 2.69 unch 1.05 and 1.22 unch 0.38 respectively .sx This shows that the vertical component of the turbulence field was locally homogeneous and compares well to flat idealized terrain results ( see , for example , Smith 1975 ) , whereas the longitudinal component may have been subjected to slope distortion and roughness change effects ( see , for example , Panofsky et al. 1982) .sx These values are consistent with turbulence measurements reported by Hogstrom et al. ( 1984 ) over a coastal forest .sx This behaviour is to be expected as the measurements are all made well within the inner layer of the hill defined by Hunt et al. ( 1988 ) , which is about 30 m in this case .sx Sensible-heat-flux measurements made with the USAT enabled the stability parameter z/L to be calculated , where L is the Monin-Obukhov length .sx This parameter varied from +0 .sx 005 to +0 .sx 018. The correction to the flux measurements was thus of the order of 5 to 13% justifying the assumption of near-neutral stability .sx It is not clear how reliable the measurements of heat flux from sonic devices are in cloud .sx Schotanus et al. ( 1983 ) show that under non-neutral conditions out-of-cloud contributions due to vapour fluxes may be as large as 20% .sx ( e ) The normalized deposition velocities for liquid water content .sx For case 1 the mean ratio formula at the measurement height was 0.22 unch 0.09 for the FSSP-100 and 0.26 unch 0.12 for the PVM-100 .sx ( These do not include periods where the deposition velocity was negative) .sx The deposition velocity for momentum was 0.052 unch 0.012 m s -1 averaged over the data set .sx For case 2 the normalized deposition velocities were slightly higher , with the period indicated in Fig. 2 producing values of formula of 0.28 unch 0.23 for the PVM-100 and 0.38 unch 0.29 for the FSSP-100 .sx Most of the variation in these results occurred during the first hour after the onset of cloud .sx If these data are ignored the ratios for the rest of the period were 0.24 unch 0.18 ( PVM ) and 0.37 unch 0.14 ( FSSP) .sx The deposition velocity of rmomentum was 0.069 unch 0.028 m s -1 averaged over the data set .sx The reasons for the deposition velocities for liquid water being substantially less than for momentum in these two case studies are discussed in detail below .sx Qualitatively , however , for case 1 the wind speeds were very low , resulting in a very low collection efficiency for the droplets .sx For case 2 the wind speed was much higher , but the mean volume-weighted drop radius was only 4.4 unch 0.5 mu m , Fig. 3 , and so the collection efficiency was relatively small .sx These case studies serve to underline the importance of a knowledge of the drop size distribution undergoing deposition to a canopy and show that simple momentum considerations employed in some models can lead to over - estimates of cloud water fluxes .sx ( f ) Variation in deposition velocity with droplet size .sx Chamberlain ( 1967 ) measured , in a wind tunnel , the deposition velocity of particles to grass as a function of size .sx Figure 11 shows his results for different values of u * .sx Deposition velocity increases with particle size below 10 mu m radius and this is particularly marked for drops between 1 and 5 mu m. Figure 11 also shows the results of Gallagher et al. ( 1988 ) obtained over moorland .sx The behaviour is very similar between 2 and 10 mu m although beyond 12 mu m a decrease in deposition velocity was observed .sx Figures 12 and 13 show formula as a function of drop radius for case 1 and case 2 with error bars calculated as described earlier .sx The solid line shows formula for droplets deposited at a rate equal to momentum .sx For case 1 , as expected , all but the very largest droplets fall below the momentum value .sx Despite the appreciable liquid water content , turbulent deposition rates in low wind speeds were small .sx Figure 13 shows formula obtained for the first period of case 2 as cloud formed over the canopy , again showing the error bars .sx The deposition velocities are all much higher than for case 1 , owing to the higher wind speeds , and for larger droplets exceed that for momentum .sx Figure 14 combines the results from the two case studies from this work and compares these results with those of Gallagher et al. ( 1988 ) , normalized in the same way .sx The general conclusions from this figure are that for droplet sizes of up to about 12 mu m radius the results from case study 2 are very similar to those reported by Gallagher et al. ( 1988 ) , measured over moorland by a very different technique .sx Above about 12 mu m radius the results diverge , but the uncertainties were large in this range for both studies .sx The light winds of case study 1 produced significantly different results .sx diagram&caption .sx 5 .sx MODEL COMPARISONS .sx The assumption that cloud water deposition velocity approximates to that for momentum is not valid under all conditions .sx The variation of collection efficiency with drop size must be taken into account , using a knowledge of the droplet size spectrum .sx To provide descriptions for the deposition of cloud water that are more accurate , two model studies were performed .sx