• Open Access

A comparison of three Met Office wind observing systems



Results of a side-by-side comparison of wind speed, as measured by three Met Office observing systems, are presented. Mean wind speeds as measured by each system differed by < 1.0 knots (0.5 m s−1) at wind speeds of up to 30 knots (15.4 m s−1). Although differences increased slightly with increasing wind speed, they remained within user-specified accuracy limits over the whole range of wind speeds experienced. The trial set-up involved the mounting of anemometers on a single mast. Although a slight sheltering of anemometers occurred at given wind directions, the effect on the overall wind speed differences was negligible. © 2012 British Crown copyright, the Met Office. Published by John Wiley & Sons Ltd.

1. Introduction

During 2008 and 2009, the Met Office implemented a new system for data collection and coding throughout its network of surface stations. The new system, the Meteorological Monitoring System (MMS), replaces a number of different automatic weather station (AWS) types previously employed within the network (Green, 2010). As part of the MMS project, inter-comparison of various Met Office observing systems was conducted at Camborne, UK. The full set of inter-comparison results is presented by Clark et al. (2010). Here, we present the results of a side-by-side comparison of wind speed as measured by three different Met Office wind systems. In this paper, ‘wind system’ is defined as the whole observing system for wind speed and direction, i.e. the sensors (comprising a three-cup anemometer and a vane), and the system used to process and log the data.

The inter-comparison of old and new systems is a vital aspect of replacement projects. Such comparisons allow identification and quantification of any differences in the measurements from each system, thereby ensuring good continuity in the data record (World Meteorological Organization (WMO) (2008)). Although long-period records (i.e. several decades or longer) of wind speed are rare in the UK (and elsewhere), those that do exist show inter-decadal variations of the order of 1 knot (0.5 m s−1) (Palutikof et al., 1997). Rockel and Woth (2007), using simulated wind series constructed from regional climate model data, predicted increases in the 99th percentile of daily mean wind speeds of around 1–5% by 2071–2100, compared with 1961–1990 values. Typical measurement uncertainties associated with instrument exposure and instrument type are at least as large as these values, suggesting that accurate documentation of the effects of any changes to wind observing systems is required if the wind record is to be useful in future climate change studies.

The remainder of our paper is structured as follows. A brief history of recent changes to Met Office wind systems is given in Section 2, together with a description of each of the wind systems compared. The inter-comparison set-up is described in Section 3. Results are presented and discussed in Section 4. Conclusions are presented in Section 5.

2. Overview of current and recent Met Office wind systems

During the 1950s, the Met Office introduced Munro three-cup anemometers and vanes for wind speed and direction measurements at surface stations. These sensors, together with the recording and display infrastructure, collectively became known as the ‘Mark 4’ (hereafter Mk 4) wind system. During the period in which the Munro anemometer and vane were in use, a number of changes to the data logging and telemetry infrastructure were made as a result of developments in communication and obsolescence of component parts. Prior to the advent of digital recording, mean wind speeds were displayed on analogue dials and gust speeds were recorded on an anemograph, from which values were manually read. During the 1980s, the Semi-Automatic Meteorological Observing System (SAMOS) was introduced, which automated much of the data processing and downstream communication. In particular, mean and gust speeds were calculated by using an intelligent sensor unit (ISU) which was mounted on site, near the wind mast. The wind speed was sampled at a frequency of 4 Hz, from which rolling 3 s mean values were calculated. The maximum 3 s wind speed was used as the gust speed, as recommended by the WMO (2008), and 2- and 10 min mean speeds were also calculated. The observations were then encoded and sent back to the Met Office headquarters at a frequency generally not exceeding one per hour. Extensive comparisons were conducted in order to quantify the differences between the analogue and digital systems; in particular, much work was undertaken to compare the chart recorder and ISU-derived gusts speeds (Sparks, 1997).

From 1997, the Met Office began a gradual replacement of the Mk 4 wind system. The new wind system, the ‘Mark 6’ (hereafter Mk 6), comprises a Vector instruments A100 three-cup anemometer and W200P vane. Although the Mk 6 wind system had been installed at most sites prior to the roll-out of MMS, a few Mk 4 systems remained. These were replaced during MMS roll-out, so that the Mk 6 system is now used throughout the network. In the Mk 6 system, the wind speed is also sampled at a frequency of 4 Hz, from which rolling 3 s averages are calculated. A Campbell scientific CR-series logger performs various calculations, in place of the SAMOS system ISU. Mean and gust wind speeds are calculated by the logger at 1 min intervals. The minute-resolution data are sent to the Met Office headquarters, where they are stored in a rolling archive for a period of 1 year. Conversion of the raw data into coded messages, such as the surface synoptic observation (SYNOP) message, and calculation of longer-period averages (e.g. 10 min mean wind speed), are performed centrally at the Met Office headquarters. However, at manned stations, longer-period averages are also calculated on site in order to facilitate the local display of wind data.

3. Trial set-up

The introduction of MMS provides a timely opportunity for the inter-comparison of several currently- and recently-used Met Office wind systems. In order to investigate changes associated with both the introduction of MMS and the relatively recent upgrade from the Mk4 to the Mk 6 sensors, three wind systems were compared: (1) SAMOS with Mk 4 anemometer, (2) SAMOS with Mk 6 anemometer, and (3) MMS with Mk 6 anemometer. Throughout the duration of the trial, the SAMOS Mk 6 system remained the operational wind system at Camborne. The operational sensors were mounted on a mast at 10 m above local ground level, the standard height for surface wind measurement as stipulated by WMO (WMO, 2008). The additional wind systems, the MMS Mk 6 and SAMOS Mk 4, were also mounted on the operational wind mast, using a cross-bar set-up (Figure 1). WMO does not provide any recommendations for the set-up of sensors in anemometer inter-comparisons. However, an earlier pilot trial at the Camborne site, using anemometers located on separate masts, revealed large minute-to-minute variability in mean wind speeds when compared to measurements made on the same mast. For this reason, the decision was made to mount all sensors on the same mast. A further advantage of this set-up is that the operational wind mast ensures the best possible exposure of sensors afforded by the Camborne site.

Figure 1.

Left panel: Camborne wind mast and location of anemometers. Right panel: Plan view diagram showing location of trial and operational anemometers and separation distances. Black circles denote the locations of vertical connecting bars. Grey circles and black triangles indicate the locations of cups and vanes, respectively

In each system, 2 min mean wind speed was calculated from the raw samples, logged and archived at a frequency of once per minute. The 2 min mean wind speed was chosen for the analysis because it is the shortest averaging period for which data are routinely available in the SAMOS systems. For the MMS Mk 6 system, 2 min means were calculated from the archived 1 min mean values. Data were collected for a period of 5 months, ending 30 November 2009. This period comprises over 220 000 individual 2 min mean wind speed measurements for each system.

WMO guidelines stipulate a required accuracy for wind speed measurements of ± 1 knot (0.5 m s−1) at wind speeds of < 9.7 knots (5 m s−1) and ± 10% at higher speeds. The Met Office applies stricter accuracy limits, as stipulated by various external users of the wind data, of ± 1 knots (0.5 m s−1) at wind speeds of < 20 knots (10.3 m s−1), and ± 5% at higher speeds. In this comparison, the latter will be used as the acceptance criteria for differences in wind speed as measured by each of the systems. The anemometers were all subject to routine operational calibration and maintenance procedures. Anemometers are calibrated in a wind tunnel, at simulated wind speeds of up to 80 knots (41.2 m s−1), typically at 10 knot (5.1 m s−1) intervals, and the threshold (start-up) speed is also documented. For the Mk 4 anemometer, the tolerances for wind tunnel testing are ± 1.0 knots (0.5 m s−1) for wind speed < 40 knots (20.6 m s−1), and ± 2.0 knots (1.0 m s−1) at higher speeds. A smaller tolerance of ± 0.085 knots for wind speed < 10 knots (5.1 m s−1), and ± 0.85% at greater speeds, is used for the Mk 6 anemometer. In addition to the initial calibration, on-site conformance checks are performed approximately every 6 months, to detect and rectify any significant drift in the sensor output, or to check the performance of the ISU. For Mk 4 wind systems, the anemometer head is driven at a set speed and the frequency output is logged. In the Mk 6 system, a ‘single-point’ check of the ISU is provided by a test unit, which simulates a wind speed of 30 knots (15.4 m s−1). Wind systems failing the conformance checks are returned to the Quality Assurance Laboratories in Exeter for further testing and possible replacement. Where anemometer drift is identified, replacement of anemometer moving parts, such as the internal bearings, is carried out.

4. Results

4.1. SAMOS Mk 6—MMS Mk 6

Over the whole trial period, the mean (SAMOS Mk 6—MMS Mk 6) difference was − 0.17 knots, with a standard deviation of 0.56 knots. This is well within user-required accuracy limits (the overall mean wind speed was 10.1 knots over the trial period). The maximum possible difference between two newly calibrated Mk 6 units, given the calibration tolerance of 0.085 knots at wind speeds of 10 knots, is ± 0.17 knots. Considering the additional potential sources of error associated with a field trial, including slight exposure differences, and the possible drift in anemometer output over time, the observed mean difference of − 0.17 knots is encouragingly small.

Figure 2 shows the magnitude of the (SAMOS Mk 6—MMS Mk 6) difference as a function of mean wind speed. The difference varies little with wind speed up to 25 knots, remaining between 0 and − 0.4 knots. Thereafter, the magnitude of the mean difference increases slowly, to approximately − 1.5 knots as the mean speed approaches 40 knots. The sample size becomes small for mean wind speed > 35 knots, which likely explains the increasing variability of the mean difference, with increasing speed, at the higher end of the range of mean speeds observed. The mean difference lies within the user requirement accuracy limits through the whole range of wind speeds sampled, although it falls towards the low end of the acceptable range at mean speeds > 25 knots. Although mean wind speeds above 40 knots are occasionally observed at lowland, inland UK stations, they are climatologically rare. The results therefore suggest that mean wind speed differences between SAMOS Mk 6 and MMS Mk 6 will be within user-required accuracy limits throughout the normal range of wind speeds experienced at UK sites.

Figure 2.

Mean wind speed difference (SAMOS Mk 6—MMS Mk 6) (bold line) and mean ± 2 standard deviations (dashed lines) of the differences, as a function of MMS 2 min mean wind speed. Shaded area shows accuracy limits of user requirement

The standard deviation of the difference also increases slowly as mean speed increases (Figure 2). This is as expected, given that the magnitude of high frequency fluctuations in the wind speed, and hence minute-to-minute differences in wind speed as measured by each system, naturally increase with increasing mean wind speed. For individual minutes within the trial period, differences in the 2 min mean wind speed exceeded ± 3.0 knots on 0.03% of occasions. Differences exceeding ± 2.0 and ± 1.0 knots occurred on 0.42 and 8.20% of occasions, respectively.

An alternative explanation for some of the instances of large wind speed differences within individual minutes is preferential sheltering of one of the anemometers, for example, by the cross-bar apparatus or neighbouring anemometers, at specific wind directions. To investigate this possibility, data were split into 120 wind direction bins, each of width 3°. A wind rose showing the distribution of wind direction throughout the trial period (Figure 3; left panel), reveals a predominance of winds from the sector 180°–270°, which tends towards the climatological mean for Camborne and most UK sites. Although the percentage of the trial period for which wind direction was within certain bins between 0° and 90° was as small as 0.25%, this still comprised a sample of over 560 min; consequently, meaningful statistics can still be derived even at these directions. Figure 3 (right panel) shows that the mean wind speed was relatively uniform over most wind directions, at between 8 and 12 knots. However, in the sector 30°–80°, mean speeds were rather lower, at approximately 5–7 knots.

Figure 3.

Left panel: percentage of time within the trial period for which MMS wind direction was within each 3° bin. Right panel: MMS Mk 6 mean wind speed (knots) over the whole trial period as a function of wind direction

Figure 4 shows the mean (SAMOS Mk 6—MMS Mk 6) two-min mean wind speed difference as a function of wind direction. The difference between the mean speed as measured by each system is less than ± 0.4 knots at the large majority of wind directions. However, peak differences of ± 0.5 to ± 0.7 knots occur over three narrow sectors (highlighted red in Figure 4):

  • In the 0°–30° sector, the magnitude of the difference exceeds − 0.4 knots, reaching a maximum of about − 0.65 knots at 10° (i.e. SAMOS Mk 6 lower). Considering the trial set-up (Figure 1), the SAMOS Mk 6 anemometer may be expected to suffer from a degree of sheltering in this sector, owing to the location of the cross bar and connecting upright section, located to the north of the sensor.

  • In the sector 80°–110°, differences exceed 0.4 knots, reaching a maximum of ∼0.7 knots at 100° (MMS Mk 6 lower). Considering the trial set-up (Figure 1), it is clear that this is also likely an effect of sheltering. In this case the sheltering is caused by the SAMOS Mk 4 anemometer, which is located due east of the MMS Mk 6 anemometer.

  • A difference of − 0.6 knots is evident at 170°. This is harder to explain considering the trial set-up, since the SAMOS Mk 6 anemometer appears, if anything, more exposed than the MMS Mk 6 anemometer at this direction. An alternative explanation is provided by Figure 3 (right panel); mean speeds were higher (∼13 knots) in this sector, which may have acted to accentuate the magnitude of the differences.

In order to investigate the effect of these sheltering problems on the overall difference in mean wind speed as measured by each system, and its standard deviation, the dataset was filtered to remove all data for which the wind direction was within the sectors 0°–30° and 80°–110°. Analysis of the filtered dataset revealed negligible change in the overall (SAMOS Mk 6—MMS Mk 6) mean difference compared to the un-filtered dataset; however, the standard deviation and inter-quartile range of the difference were very slightly lower in the filtered dataset (0.01 and 0.07 knots, respectively). The small magnitude of the differences between the filtered and un-filtered datasets likely reflects the fact that the wind direction only fell within the filtered segments for ∼3% of the trial period. Furthermore, since the mean wind speed differences are of opposite sign in each removed sector, the effect of removal of both would tend to cancel each other out as far as the overall mean difference is concerned. In summary, therefore, comparison of the filtered and un-filtered datasets shows that the anemometer shading had negligible impact on the results of the comparison.

Figure 4.

Mean wind speed difference (SAMOS Mk6—MMS Mk6), expressed as a function of wind direction in 3° bins. Red sections denote mean differences exceeding approximately ± 0.4 knots

4.2. SAMOS Mk 6—SAMOS Mk 4

Over the whole trial period, the mean difference (SAMOS Mk 6—SAMOS Mk 4) was 0.78 knots and the standard deviation 0.72 knots. The slightly larger standard deviation, compared to that found for the comparison of the two Mk 6 systems, is likely to due to the different response times of Mk 4 and Mk 6 anemometers, leading to larger minute-to-minute differences in rapidly fluctuating winds. Figure 5 shows the mean difference and (mean ± 2 standard deviations) as a function of mean wind speed. The peak in the difference at wind speeds of between 2 and 5 knots is due to the difference in threshold (start-up) and stall speeds of the Mk 4 and Mk 6 anemometers; in the Mk 4 anemometer, it is between 5 and 8 knots, depending on the instrument. By comparison, the threshold speed of the Mk 6 anemometer is generally < 0.5 knots. Consequently, data for mean speeds of less than 5–8 knots are not strictly comparable. Between 5 and 35 knots, the mean difference remains nearly constant with increasing mean speed, at 0.6–1.0 knots. The overall mean difference is considerably larger than found for (SAMOS Mk 6—MMS Mk 6) over the same range of mean wind speed. However, the difference still lies within user-required accuracy limits.

Figure 5.

Mean wind speed difference (SAMOS Mk6—SAMOS Mk4) (bold line) and mean ± 2 standard deviations (dashed lines) of the differences, as a function of mean wind speed. Shaded area shows accuracy limits of user requirement

Figure 6 shows a radial plot of the mean difference (SAMOS Mk 6—SAMOS Mk 4) as a function of wind direction. The plot was constructed using a filtered dataset, following the removal of all instances of mean wind speed < 5 knots (in order to remove the effects of the aforementioned differences in threshold speeds of the Mk 4 and Mk 6 anemometers). The mean difference varies between about 0.4 and 1.2 knots over the large majority of wind directions (i.e. a variation of ± 0.4 knots from the overall mean difference). Larger deviations from the overall mean difference can be seen, for example, in the 20°–40° sector. This would be consistent with sheltering of the SAMOS Mk 6 anemometer by the SAMOS Mk 4 anemometer. A larger deviation from the overall mean difference also occurs in the sector 160°–200°. This is difficult to explain by consideration of the trial set-up, as both anemometers appear equally exposed to winds from these directions. However, the higher overall mean wind speeds in this sector (Figure 4) may partially account for the anomaly.

Figure 6.

Radial plot of (SAMOS Mk6—SAMOS Mk4) wind speed difference (knots) as a function of wind direction, in 3° bins

5. Conclusions

Mean winds as measured by SAMOS Mk 4, SAMOS Mk 6, and MMS Mk 6 wind systems compared well; overall mean differences between systems were within ± 1.0 knots. Differences generally increased with increasing mean wind speeds; however, the differences remained within user-defined accuracy requirements (i.e. ± 1.0 knots up to 20 knots and ± 5% at higher mean speeds) over the full range of wind speeds experienced. In general, the differences attributable to the different anemometer design (Mk 4 vs Mk 6) were larger than those attributable to the different measuring systems (SAMOS vs MMS). These results are comparable with those of Wauben (2007), in which reported differences in mean wind speed as measured by three anemometers located on the same mast were of order ± 0.2 knots (over all wind directions) with mean differences of ± 0.6 to ± 0.8 knots in certain wind direction sectors.

The set-up of anemometers on a single mast resulted in sheltering of the anemometers over narrow sectors. This resulted in larger wind speed differences, of order 0.5–0.8 knots, over relatively narrow sectors of wind direction. However, removal of data for which the wind direction lay within the affected sectors resulted in negligible change in the overall mean wind speed difference or the standard deviation. Therefore, the results demonstrate that, providing adequate horizontal separation of the anemometers is ensured, inter-comparison of anemometers on a single mast can yield reliable results.


The authors wish to thank staff of the Met Office Engineering Services for set-up of the instruments at Camborne, without which this inter-comparison would not have been possible. In addition, thanks are due to Charlie Pethica and Alan Hewitt for their contribution to the data analysis.