A novel experimental design to investigate the wind-induced undercatch

Transcription

1 A novel experimental design to investigate the wind-induced undercatch M. Pollock (3,4), M. Colli (1,2), G. O Donnell (3) M. Stagnaro (1,2), L.G. Lanza (1,2), P. Quinn (3), M. Dutton (4), M. Wilkinson (5), A. Black (6),C. Kilsby (3) and P. E. O Connell (3,4) (1) University of Genova, Dep. of Civil, Chemical and Environmental Engineering, Via Montallegro 1, Genoa, Italy, (2) WMO/CIMO Lead Centre B. Castelli on Precipitation Intensity, Italy (3) School of Civil Engineering and Geosciences, Newcastle University, UK. m.d.pollock@ncl.ac.uk (4) Environmental Measurements Ltd, UK. (5) James Hutton Institute, Aberdeen, UK (6) University of Dundee, Dundee, UK Abstract: Data from rain gauges are critical for flood forecasting and flood risk management, water resource management and hydrological modelling. They are often considered to provide the most accurate practicable measure of rainfall at a point in space and time, but they remain subject to environmental errors. Inaccuracies in measurements are compounded in modelling applications resulting in potentially misleading or incorrect results; it is therefore of utmost importance to understand uncertainty in rainfall observations. Wind-induced undercatch is widely considered the most significant source of uncertainty, the extent of this phenomenon is affected by a number of factors including the shape of the collecting orifice. Recent Computational Fluid Dynamics (CFD) simulations confirm this (Colli et al., 2016). One-minute resolution hydro-meteorological observations are recorded at four UK field sites. The undercatch is investigated by comparing measurements made by gauges mounted in different configurations; in a pit with the orifice at ground level which is the recommended international reference for rainfall, and directly on the ground surface as is the convention in the UK (eg. by SEPA, Met Office, EA). Results from these sites are analysed. To investigate the impact of gauge shape on wind speed, for the magnitude of velocity and turbulence, CFD simulations are compared for aerodynamic versus conventional rain gauges. 1. Introduction and review: From Jevons (1861) to Strangeways (2004), much has been written about the rainfall measurement problem, and a lot of the focus gravitated towards the environmental impact of the wind s influence on rainfall catch. Scientific consensus dictates that the physical presence of a rain gauge causes an acceleration of the airflow over the orifice, distorting the trajectories of rainfall particles and leading to an undercatching effect. Additionally, the shape of the rain gauge has an influence on the catch performance. Despite numerous intercomparison studies, the rainfall undercatch problem remains unresolved. Studies such as those by Sevruk and Hamon (1984) and Goodison, Louie and Yang (1997) significantly advanced knowledge and succeeded in raising further research questions, but the complexity involved ensures the solution to this problem remains elusive. The WMO laboratory (2005) and field ( ) intercomparisons on rainfall intensity took the necessary step of linking the meteorological and metrological communities, moving the science in the direction of operational standardisation. These studies have contributed to improvements in measurement over the last decade, particularly with regards to tipping bucket rain gauge (TBR) counting errors. Now, the wind issue is coming back into focus once again, for example Wolff et al. (2015) and Colli et al. (2015).

2 2. Methodology: Two sites are selected for investigation, at Nafferton Farm and Talla Water in the UK shown in Figures 1a and 1b respectively. The former is situated in the North East of England and can broadly be classified as a lowland east coast site with good exposure to the wind (elevation of 110m). The latter is categorised as an upland western UK site with an elevation of 430m. Both sites were chosen for their good exposure to the wind and both possess a mixture of aerodynamic and straight-sided cylindrical rain gauges, the specifics of which are detailed below. All rain gauges in this analysis operate using the tipping bucket principle. Figure 1a (left) & 1b (right). 1a - Nafferton Farm research site, Northumberland, UK. 1b - Station at Talla Water, Scottish Borders, UK Table 1 and Figure 2 below correspond to the data and the equipment used in this analysis from the Nafferton Farm station. Table 1: Description of the rain gauge, wind speed and air temperature measurements at Nafferton Farm used in this analysis Name of data series: Description of instrument: Units: P2_SBS1_Tot SBS500 pit gauge (orifice height = 0m) mm (1min totals) P5_CAS2_Tot CASELLA pit gauge (orifice height = 0m) mm (1min totals) G3_SBS3_Tot SBS500 ground-mounted gauge (orifice height = ~0.45m) mm (1min totals) G4_CAS1_Tot CASELLA ground-mounted gauge (orifice height = ~0.35m) mm (1min totals) WindSpd_Gnd_Max Max. 1-min wind speed (measurement height = ~0.35m) m/s AirTemp_Gnd_Avg Average 1-min air temp. (measurement height = ~0.35m) degc Figure 2a (left) and 2b (right). 2a photograph and 3D model of the aerodynamic aluminium SBS500 tipping bucket rain gauge. 2b photograph and 3D model of the conventional cylindrical CASELLA tipping bucket rain gauge.

3 Table 2 and Figure 3 relate to the research site at Talla Water. Both sites are equipped with wind speed measurement and air temperature measurements which are captured at the height of a ground mounted gauge (around 0.35 metres above the ground). Table 2: Description of the rain gauge, wind speed and air temperature measurements at Talla Water used in this analysis Name of data series: Description of the instrument: Units: P1_ARG1 ARG100 pit gauge (orifice height = 0m) mm (1min totals) G2_ARG4 ARG100 ground-mounted gauge (orifice height = ~0.35m) mm (1min totals) AG1_ARG3 ARG100 pedestal-mounted gauge (orifice height = ~1.5m) mm (1min totals) G3_CAS1 CASELLA ground-mounted gauge (orifice height = ~0.35m) mm (1min totals) WindSpd_Gnd_Max Max. 1-min wind speed (measurement height = ~0.35m) m/s AirTemp_Gnd_Avg Average 1-min air temp. (measurement height = ~0.35m) degc Figure 3: Figure 3a (left) and 3b (right). 3a photograph and 3D model of the aerodynamic plastic ARG100 tipping bucket rain gauge. 3b photograph and 3D model of the conventional cylindrical CASELLA tipping bucket rain gauge. As a means of trying to view the wind-induced undercatch in a slightly different way, two primary focal points are outlined as follows. The mounting height of a gauge is predicted to affect the undercatch due to the hypothesis that there are increasing wind speeds as the distance above the ground increases, within the boundary layer. The greater the height, the greater the undercatch. Therefore three different mounting heights are selected. As the rain gauge shape also affects the undercatch by causing different acceleration and turbulence structures, aerodynamic shapes are compared against conventional cylindrical Visualisations of a number of winter storms are produced for each of the rain gauges listed in Tables 1 and 2, and displayed in Figures 2 and 3. Results for Nafferton Farm are as a graph for one storm, and summarised in a table. Results for Talla Water are displayed for 5 storms and also summarised in a tale. The focus is the degree of undercatch for each gauge in each storm. The graphs produced in the results section depict as a snapshot in time the evolution of rain events. In this analysis, a rain event is defined arbitrarily as a period of rain prior to which and after which there are four hours of no rainfall. The first rain and the last rain are detected using the reference pit rain gauge. This 4 hour period prior to the first and after the last rainfall provides a buffer zone, aiding visualisation. There are many methods for defining events (Dunkerley, 2008), for the purposes of exploratory analysis this method was deemed to be a fitting solution.

4 3(a). Results and discussion: Nafferton Farm Figure 4 displays the evolution of the storm event Eva, named by the Met Office, taking place in December This was the largest 2015 winter storm observed at Nafferton Farm. The legend in the top left corner shows what each instrument represents. A detailed description of this is in Table 1 above. The primary y axis displays the rainfall (mm) cumulating over the event, as depicted by the blue, green, red and light blue lines. Plotted on the secondary y axis are the wind speed and the air temperature. The wind speed can be clearly seen as the highly dynamic time series. Figure 4: Time series Storm Eva at Nafferton Farm. P2_SBS1 (aerodynamic) and P5_CAS2 (cylindrical) are both pit gauges. G3_SBS3 (aerodynamic) and G4_CAS1 (cylindrical) are gauges mounted on the ground surface with orifice height around 0.45m) Figure 4 demonstrates that there is a 1% difference in the measurement between the aerodynamic and the cylindrical pit-mounted gauges. Both gauges agree well, which is what is expected for two gauges which are equally sheltered from the wind i.e. within a pit the shape of the gauge should theoretically be irrelevant. The ground mounted aerodynamic gauge performs less well than the two pit gauges, exhibiting a 4% undercatch. The ground mounted conventional straight-sided gauge is subject to greater undercatch, 9.2%. Again, this fulfils the expectation that the shape of the rain gauge is significant, with the aerodynamic ground-mounted rain gauge collecting more rainfall. Table 3 provides a description of undercaatch across 3 other 2015 UK winter storms for the 2 ground gauges along with the average of the maximum 1-minute wind speed. Table 3: Description of the undercatch occurring for 4 storms at Nafferton Farm occurring during November December Storm event undercatch at Nafferton Farm (%): Colour coded data series: Barney Clodagh Desmond Eva Frank (no rain storm) P2_SBS1 (reference: mm) 27.4mm 9.2mm 26.6mm 41.4mm - G3_SBS3 (aerodynamic) 2.2% 6.5% 9.8% 4.8% - G4_CAS1 (cylindrical) 8.0% 13.0% 22.6% 9.2% - Avg:(WindSpd_Gnd_Max) 2.8 m/s 5.5 m/s 8.3 m/s 2.5 m/s -

5 In Table 3 it is clear that the ground-mounted aerodynamic gauge is subject to an undercatch when compared to the pit gauge. From Barney to Eva respectively, this is 2.2%, 6.5%, 9.8% and 4.8%. It can also be seen that the ground-mounted cylindrical gauge consistently has a greater undercatch across all storms than its aerodynamic ground-mounted counterpart. Again, this is to be expected due to the less aerodynamic shape offered by the conventional straight-sided gauge. Figure 5 shows the percentage undercatch plotted against average rainfall event wind speed for two series; the ground mounted aerodynamic gauge (G3_SBS3) and the ground-mounted cylindrical gauge (G4_CAS1). Figure 5: Linear regression of the ground-mounted aerodynamic and cylindrical gauges plotted against the percentage undercatch, for 4 storms at Nafferton Farm. Simple linear regression models were plotted to the data in Figure 5; these provide a favourable R 2 value which demonstrates that there is a good fit between the data and the fitted regression line. However, this is clearly not a robust statistical analysis for a number of reasons. The most significant criticism would be the lack of events plotted, however there are other limitations which are outlined in the final chapter of this paper, and indicate the direction in which the author intends to go with future analyses. Despite the limitations it remains interesting to visualise these relationships in this way at this stage.

6 3(b). Results and discussion: Talla Water Figure 6 shows Storm Barney as it hits the rainfall research site at Talla Water on November 14 th Like the plot of Storm Eva at Nafferton Farm, there is also a marked difference between the reference pit gauge (blue line) and the other rain gauges. However, it appears that the distances between each line for this event are larger than at Nafferton Farm, which would signify a greater undercatch. Figure 6: Storm Barney at Talla Water, Scottish Borders. November 14 16, Splitting the event into sections affords a closer look in the wind-induced undercatch. A basic way of doing this is now demonstrated. The first part of the event may be characterised as a low wind section, whereas the second half is characterised as high wind. Figure 7 shows the undercatch plotted for each gauge during the low wind (left) and high wind (right) period. The average wind over each section is marked as an orange line and plotted on the secondary y axis. Undercatch is greater across all gauges for the section of the event with higher average wind speed (2.6 m/s) than for the section with lower average wind speed, 1.5 m/s. The magnitude of the undercatch for each gauge follows the same pattern established in the Nafferton Farm analysis above. The total rainfall in each section of the event is comparable. Figure 7: Gauge undercatch rations for low wind section of Storm Barney (left) and high wind section of Storm Barney (right)

7 Figure 8 presents the event evolution of storms Clodagh (top left), Desmond (top right), Eva (bottom left) and Frank (bottom right). The blue line is the reference pit ARG100 gauge, the green line is the ground-mounted ARG100, the red line is the pedestal-mounted ARG100 and the light blue line is the ground-mounted CASELLA. Therefore, the first three gauges listed are aerodynamic and the final gauge is cylindrical in shape. The plots indicate that the pit gauge consistently collects more than all other gauges. There is a possibility that some of the precipitation in Storm Clodagh (top left) may have been snowfall, due to the low temperatures. Figure 8: Storms Clodagh (top left), Desmond (top right), Eva (bottom left) and Frank (bottom right) at Talla Water, Winter Table 4 shows in blue the absolute value in millimetres recorded by the pit gauge across 5 storms. The following 3 rows show the undercatch for the 3 gauges, compared to the pit reference measurement. The final two rows display the average event wind speed and temperature, respectively. It may be seen that the total values of the rainfall events far exceeds those recorded at Nafferton Farm for the same events. This is expected for the type of storms which come from the west and therefore hit the Scottish west coast and are forced up into the mountains where the Talla Water research site is. Table 4: Description of the undercatch occurring for 5 storms at Talla Water occurring during November December Storm event undercatch at Talla Water (%) to nearest 0.5% Colour coded data series: Barney Clodagh Desmond Eva Frank P1_ARG1 (reference: mm) 66.6mm 50.8mm 102.8mm 56.8mm 122.6mm G2_ARG4 15% 16% 14.5% 12.5% 10.5% AG1_ARG3 24% 20.5% 17% 13.5% 13% G3_CAS1 21.5% 17.5% 17.5% 16.5% 14% Avg:(WindSpd_Gnd_Max) 3.1 m/s 5.0 m/s 5.3 m/s 3.5 m/s 3.78 m/s Avg:(AirTemp_Gnd_Avg) 8.1 C 1.7 C 8.0 C 2.6 C 7.0 C

8 3(c). Results and discussion: CFD analysis Computational Fluid Dynamic (CFD) analyses were carried out on two rain gauge shapes to assess their aerodynamic capabilities with reference to their suitability for measuring rainfall accurately. The two types of rain gauge used in the field study at Talla Water are modelled here, and visualisations of their aerodynamic performance are presented in Figure 9. Figure 9: CFD visualisations of the stream-wise vertical view of different parameters, for wind speeds of 7m/s. The figures are displayed as pairs, with the aerodynamic profile of the ARG100 on the top row (1a, 2a and 3a), and the conventional straight-sided gauge shown on the bottom row (1b, 2b, and 3b). 1a and 1b show the air velocity magnitude colour Um (m/s) plot and vector plot of the airflow. 2a and 2b show the air vertical velocity component Uz (m/s) plot and vector plot of the airflow. 3a and 3b show the turbulent kinetic energy k (m 2 s -2 ) plot and vector plot of the airflow. The plots are presented as vertical pairs to allow a direct vertical comparison between the different shapes. The ARG100 aerodynamic rain gauge is presented along the top row (1a 3a), and a conventional straight-sided gauge is displayed along the bottom row of the figure (1b 3b). The magnitude of velocity Um represented on a stream-wise vertical plane for the two gauges (1a-1b) confirms the presence of a shear layer (white band). This separates the region characterised by strong airflow regimes above the collector ( Uw < Um, red colour), from the recirculating airflow inside the gauge ( Uw > Um, blue colour). In the case of the ARG100 this shear layer spans over the orifice and touches the downwind rim of the collector, whereas in the case of the straight-sided gauge it develops far beyond the downwind edge of the collector and reaches higher vertical levels. This behaviour is partially explained by the stronger downdraft occurring inside the ARG100 collector (Figure 2a-2b). The straight-sided gauge shape also causes higher turbulent kinetic energy values just above the collector (Figure 3a-3b), demonstrating an improved aerodynamic behaviour of the ARG100. The actual implications of this evidence on the rainfall trajectories should be evaluated by coupling the CFD airflows with Lagrangian particle tracking models. However, the results appear to confirm the superior aerodynamic performance of the ARG100 aerodynamic gauge compared to the CASELLA cylindrical gauge. Therefore, the CFD simulations support the measurement results observed in the field at Talla Water and presented in section 3(b).

9 4. Conclusions and recommendations: The regressions carried out in Figure 5 show that there is a relationship between undercatch and wind speed. However, clearly this is only for a small number of storms. When the regressions are carried out for 34 storms the R 2 value reduces to , depending on whether the linear regression is forced through the origin. However, the p-value is extremely low indicating that the relationship is highly significant. This could be due to other codependent variables which also have an impact on the undercatch, such as the drop size distribution. At present there are no drop size measurements available at site it may be possible to infer the drop size from an assessment of rainfall intensity. This could be enough to categorise in a qualitative fashion the size of the drops, thus forming another term in the correction model. To sum up, further exploration of co-dependent variables is required through multi-variate analysis to improve the correction model. Additionally, the event definition used in this analysis is imperfect and gives rise to an unfair comparison because all events are of different duration. The average of the maximum 1- minute wind speed will not provide a truly representative appraisal to adequately characterise the wind throughout an event. In the particular case of this paper it was deemed, by viewing the event plots, that the storms were sufficiently similar in nature to allow a relatively fair comparison. Therefore, there is an assumption made that the winter storms that whip up off the Atlantic and hit the UK are similar in terms of their characteristics, such as rainfall intensity / drop size and distribution. It is possible therefore that this contributes to the reason that the regression relationship in Figure 5 is so strong. Ongoing analyses look at binning wind speed by various statistical grouping methods. Furthermore, as there is more than a full year of data at both Talla Water and Nafferton Farm it is possible to carry out more multi-event statistics and investigate the relationships across a range of events, instead of concentrating on only several. Additionally, short-duration field campaigns are capturing data at 1 second resolution. This provides greater temporal detail of the rainfall intensities and of the wind speeds. A new prototype has been developed to capture these data. Analyses will be presented in future works. References: Colli, M., R. Rasmussen, J. M. Theriault, L. G. Lanza, C. B. Baker & J. Kochendorfer (2015) An Improved Trajectory Model to Evaluate the Collection Performance of Snow Gauges. Journal of Applied Meteorology and Climatology, 54, Dunkerley, D. (2008), Identifying individual rain events from pluviograph records: a review with analysis of data from an Australian dryland site. Hydrol. Process., 22: doi: /hyp.7122 Goodison, B., P. Louie & D. Yang (1997) The WMO solid precipitation measurement intercomparison. World Meteorological Organization-Publications-WMO TD, Jevons, W. S On the deficiency of rain in an elevated raingauge as caused by wind. In London, Edinburgh and Dublin Phil. Mag, Pollock, M., M. Dutton, P. Quinn, P. E. O Connell, M. Wilkinson & M. Colli (2014) Accurate Rainfall Measurement: The Neglected Achilles Heel of Hydro-Meteorology Sevruk, B. & W. Hamon International comparison of national precipitation gauges with a reference pit gauge. Secretariat of the World Meteorological Organization. Strangeways, I. (2004) Improving precipitation measurement. International Journal of Climatology, 24, Wilkinson, M. E A multiscale nested experiment for understanding and prediction of high rainfall and flood response spatial behaviour in the Eden Catchment, Cumbria, UK. Newcastle University. Wolff, M. A., K. Isaksen, A. Petersen-Overleir, K. Odemark, T. Reitan & R. Braekkan (2015) Derivation of a new continuous adjustment function for correcting wind-induced loss of solid precipitation: results of a Norwegian field study. Hydrology and Earth System Sciences, 19,