A 14‐Year Meteorological Dataset From a University Campus in the East Midlands of the UK

Abstract

ABSTRACT

The weather station on the Loughborough University campus underwent refurbishment and upgrade in 2007, and this contribution reports on the outcome of 14 subsequent years of meteorological data collection there, before a further episode of upgrading. Data collection is described, with emphasis on the continuity or lack of continuity of the variables monitored. Out of 136 instrument‐years deployment, only 36 are less than 90% complete, and 21 less than 75% complete. Data processing discusses the method of retrieving 0900‐0900 temperature maxima and minima and rainfall totals, to correspond to the standard UK and Ireland Climatological Day. As an independent check on the probable reliability of the campus weather dataset, values are correlated with and regressed against co‐located values extracted from the UK Met Office HadUK‐grid dataset. Campus temperatures are slightly, but consistently, higher than those indicated by HadUK‐grid, while HadUK‐grid rainfall is on average almost 10% higher than that recorded on the campus. Trend‐free statistical relationships between campus and HadUK‐grid data imply that there is unlikely to be any significant temporal bias in the campus dataset. The contribution concludes with a consideration of recent and potential future applications of the dataset.

Introduction

University campuses often feature weather stations of some description, which typically provide useful data for research, teaching, and estate management. For example, the Radcliffe Meteorological Station in Oxford has run continuously since 1813 and contributes to the Central England Temperature (CET) series, one of the longest instrumental climate records in the world (Legg et al. 2025). Similarly, the weather station at the Cambridge University Botanic Garden, established in 1904, recorded a new UK official maximum temperature of 38.7°C in 2019 (Sibley 2023).

The weather station on the campus of Loughborough University, in the East Midlands of the UK, had fallen into disuse and disrepair by the mid-2000s, but in 2007 the availability of infrastructure funding made it possible to re-establish regular weather observation with new equipment. The purpose of this contribution is to outline the dataset collected at this facility between 2008 and 2021, before another episode of refurbishment and upgrading took place. The collection site itself is described before data collection methods are outlined, specific issues in data processing discussed, and notable features of the dataset highlighted. The HadUK-Grid dataset (Hollis et al. 2019) is used as an independent check on the reliability of the data collected on the campus. Finally, recent and potential future applications of the dataset, which is archived in the institutional repository, are considered.

Site Description

Loughborough University campus occupies a 2.12 km² site on the western side of the East Midlands town of Loughborough. With over 19,000 students and 3400 staff, the campus consists of a mixture of academic and support services buildings, a science and enterprise park, student halls of residence, sports facilities, and various green spaces. The town of Loughborough occupies 16.5 km², with a population of 65,000 (2021 census), and is situated in Charnwood, NW Leicestershire, between the 245 m Beacon Hill to the southwest and the floodplain of the River Soar to the northeast. The campus weather station is located at latitude 52.7632°, longitude −1.235° and 68 m a.s.l., in a dedicated paddock on a green space near the centre-east boundary of the campus (Figure 1). A cabin, which houses power and network points, sits 10 m to the northeast of the main meteorological instrument tower. The paddock is otherwise mostly open on an arc from the northwest to the northeast, but on the other sides there are fruit trees (mainly varieties of prunus domestica) at distances of 13–16 m, forming part of the university's ‘Fruit Routes’ biodiversity initiative.

[IMAGE OMITTED. SEE PDF]

Data Collection Methods

Instruments (Table 1) were fixed to a 3 m lattice mast (Figure 1), which is concreted into the ground in the centre of the paddock described above. Up to late July 2013, the instruments were controlled by a solar-charged, battery-powered Campbell Scientific CR1000 data logger and periodically manually downloaded. From early November 2013, this logger was replaced with a Campbell Scientific CR3000, run from the mains power supply from the cabin and connected to the campus network by ethernet. At the same time, the station's Young 01503 Wind Monitor was replaced by a Gill WindSonic ultrasonic anemometer. This combination remained in place for the rest of the measurement period described here. Frustratingly, the CS215 temperature/relative humidity sensor failed shortly before the peak of the 2018 heatwave and had to be replaced with another CS215. Likewise, the ARG100 rain gauge was replaced in 2011 and 2016. The main cause of data gaps is the unreliable power supply from the cabin, particularly in 2013 and 2021 (the latter leading to the complete replacement of the cabin and all other equipment). Furthermore, even though the post-2013 CR3000 logger had a backup battery, it sometimes failed to restart after mains power was lost, yielding data gaps until it was manually restarted. Nevertheless, out of 136 instrument-years deployment (Table 2), only 36 are less than 90% complete and 21 less than 75% complete.

TABLE 1 Variables measured at the campus weather station, their units, instruments, and accuracy.

Measurement label	Variable	Interval	Unit	Instrument	Accuracy
AirTC_Avg_C	Air temperature	15-min average of 60-s samples	°C	CS215 (in unaspirated radiation shield)	±0.4°C
RH_Smp_%	Relative humidity	15-min sample	%	CS215	±4%
Slr_Avg_W/m2	Solar radiation	15-min average of 60-s samples	W m−2	SP-LITE	< ±2%
Slr_Total	Solar radiation	15-min total	MJ m−2	SP-LITE	< ±2%
NR_Avg_W/m2	Net radiation	15-min average of 60-s samples	W m−2	NR-LITE	±5%
WS_Avg_m/s	Wind speed	15-min average of 60-s samples	m s−1	Young 05103 WindSonic	±0.3 m s−1 (05103) ±2% (WS)
WD_Avg_Deg	Wind direction	15-min average of 60-s samples	°	Young 05103 WindSonic	±3 degrees (both)
WD_SD_Deg	Wind direction standard deviation	15-min average of 60-s samples	°	Young 05103 WindSonic	±3 degrees (both)
WS_Max3SGust_m/s	Wind speed maximum 3 s gust	15-min maximum from running 3-s samples	m s−1	Young 05103	±0.3 m s−1
Rain_Total_mm	Rainfall total	15-min total	mm	ARG100	< ±4%
BP_Avg_hPa	Barometric pressure	15-min sample	mb ≡ hPa	CS100	±1 hPa

Source: Campbell Scientific Ltd., 1998. SP-Lite Silicon Pyranometer. User Guide. Campbell Scientific Ltd., 2005. CS100 Barometric Pressure Sensor. Instruction Manual. Campbell Scientific Ltd., 2010. ARG100 Tipping Bucket Rain gauge. User Guide. Campbell Scientific Ltd., 2010. NR-LITE Net Radiometer. Instruction Manual. Campbell Scientific Ltd., 2010. WindSonic Ultrasonic Wind Sensor. User Guide. Campbell Scientific Ltd., 2012. CS215 Temperature and Relative Humidity Probe. User Guide. R.M. Young Co., 2000. Wind Monitor Model 05103. Instructions.

TABLE 2 Data completeness expressed as % of maximum 15-min values over a year, except for Rain Total, which is a % of maximum daily values over a year (cf. Table 4). Measurement labels (column headings) are explained in Table 1.

	AirTC	RH	SlrAvg	SlrTotal	NR	WS	WDAvg	WDSD	WSGust	RainTotal	BP
2008	100	100	100	—	100	100	100	100	100	75	4.2a
2009	100	100	100	—	100	100	100	100	100	75	17a
2010	100	100	100	—	100	100	100	100	100	58	—
2011	89	89	100	—	100	100	100	100	100	62	62
2012	96	96	100	—	100	100	100	100	100	80	96
2013	70	70	71	16a	71	71	71	71	55a	70	70
2014	100	100	100	100	100	100	100	100	—	100	100
2015	92	92	92	92	92	92	92	92	—	51	92
2016	73	73	73	73	73	73	73	73	—	48	73
2017	100	100	100	100	100	100	100	100	—	100	100
2018	92	100	100	100	100	100	100	100	—	100	100
2019	99	99	99	99	99	99	99	99	—	99	99
2020	87	87	99	99	99	99	99	99	—	91	99
2021	76	79	79	79	79	79	79	79	—	66	79

Data Processing

Data retrieved manually or downloaded remotely were filtered for invalid measurements. The 15-min data were then processed to daily and monthly values, using the pivot table function in Microsoft Excel. Most variables could be output simply as midnight-to-midnight daily means (e.g., solar and net radiation, wind speed). However, certain variables needed to be referred to the UK and Ireland standard ‘Climatological Day’ (Burt 2012, 272), 0900-0900: namely, air temperature minimum and maximum, plus rainfall total. The procedure for this follows Burt (2012; ) and requires the insertion of additional date columns into the spreadsheet to define two further, separate ‘Climate Dates’ for maximum temperature and rainfall total (the 24 h commencing at 0900 on the date given, ‘ClimateDateMax’), and for minimum temperatures (24 h ending at 0900 on the date given, ‘ClimateDateMin’). For the archived data, in the spreadsheet tabs labelled ‘Output—Daily 09-09 minima’, the pivot table function derives daily minimum temperatures by the correct 0900-0900 date, given by the ClimateDateMin variable. Similarly, in the tabs labelled ‘Output—Daily 09-09 maxima’, the pivot table function derives daily maximum temperatures and daily rainfall totals by the correct 0900-0900 date, given by the ClimateDateMax variable. Then, in the tabs labelled ‘Output—Daily 00-00 means’, variables with midnight-to-midnight means use the unmodified date variable. To take into account the effect of missing data, the tab ‘Completeness’ again uses a pivot table to count the numbers of daily and monthly observations where the 15-min data are not at least 99.99% complete. Values are only entered into the ‘Daily data’ tab of the archived spreadsheets where 15-min data are at least 75% complete; values are only entered into ‘Monthly data’ tabs where daily data are at least 75% complete (Table 2).

Wind directions are particularly important in UK meteorology because they indicate the origin of air masses with potentially contrasting characteristics. But wind directions are not averaged in the same way as other variables, as they are measured on a circular scale. Instead, 15-min wind direction data in degrees are converted to 16 compass points, and a pivot table is used to summarise these into wind speed categories, giving the frequency and strength of winds by compass point.

In order to evaluate the reliability of the collected dataset, it was compared to equivalent variables from the HadUK-Grid dataset (Hollis et al. 2019). HadUK-Grid is a collection of gridded climate variables derived from the network of UK land surface observations, which have been interpolated from meteorological station data onto a uniform grid to provide coherent coverage across the UK at 1 km × 1 km resolution. Daily and monthly air temperature and rainfall variables from the HadUK-Grid v1.1.0.0 Met Office (2022) were downloaded from the Centre for Environmental Data Analysis (CEDA) archive (). Then the grid square containing the campus weather station was identified using the Point Subset Tool of the NOAA Weather and Climate Toolkit () in order to retrieve data from that specific location. Co-located campus and HadUK-Grid data were then compared by correlation and regression (Table 3).

TABLE 3 Correlation and regression slope and intercept between air temperature (°C) and rainfall (mm) from this dataset and the equivalent HadUK-Grid values. Values in italics are from years with missing (< 75%) data (Table 2).

	Correlation	Slope	Intercept
AirT mean	AirT max	AirT min	Rainfall	AirT mean	Rainfall	AirT mean	Rainfall
2008	0.999	1.000	0.999	0.992	0.987	0.834	0.506	3.828
2009	0.999	0.999	1.000	0.974	0.983	1.206	0.545	−16.057
2010	0.999	1.000	0.999	0.959	0.995	1.032	0.412	−8.341
2011	0.997	0.999	0.994	0.971	0.979	0.865	0.689	−3.594
2012	0.999	0.999	0.999	0.988	0.979	0.988	1.265	−0.324
2013	0.998	0.996	0.997	0.958	0.994	0.930	0.894	−2.952
2014	0.999	1.000	0.998	0.991	1.015	0.939	0.183	−0.795
2015	0.999	1.000	0.999	0.970	1.000	1.205	0.336	−22.729
2016	0.997	0.999	0.999	0.994	0.988	0.888	0.524	−0.428
2017	0.999	1.000	0.999	0.810	0.992	0.808	0.612	3.921
2018	0.999	1.000	0.999	0.994	1.009	0.985	0.320	3.819
2019	0.999	0.999	0.999	0.944	0.967	1.112	0.734	2.014
2020	0.999	1.000	0.999	0.986	0.973	0.958	0.707	3.938
2021	0.999	1.000	0.999	0.980	1.005	0.921	0.517	0.940

Summary Meteorology

Monthly maximum and minimum air temperatures for the 2008–2021 period from both campus and HadUK-grid data are shown in Figure 2. The maximum recorded monthly air temperature was 24.6°C in July 2014 (Table 4). A higher maximum was probably attained in the hot summer of 2018 (Kendon et al. 2019; Met Office 2018), but data were lost in July and August of that year when the CS215 sensor failed (Tables 1 and 2). The HadUK-grid series indicates that July 2018 was indeed the warmest month of this 14-year period, at 26.2°C. The minimum recorded monthly air temperature was −3.3°C in December 2010 (Table 4), which coincides with the coldest month in the HadUK-grid series, at −3.7°C, and with a period of significant snowfall and exceptionally low temperatures nationally (Prior and Kendon 2011). The figure shows that campus temperatures are slightly (close to the ±0.4°C accuracy of the sensor; Table 1), but consistently, higher than those indicated by HadUK-grid.

[IMAGE OMITTED. SEE PDF]

TABLE 4 Annual air temperature (°C) and rainfall (mm) summary. Values in italics are from years with missing (< 75%) data (Table 2). HadUK equivalent is the total from the same months that are complete in the campus dataset when this is less than 12.

	AirT mean	AirT monthly max	AirT monthly min	Rainfall total	(Rainfall no. complete months)	HadUK equivalent
2008	10.5	22.6	1.1	610.8	9	690.7
2009	10.7	22.7	0.9	636.4	9	647.3
2010	9.4	22.9	−3.3	322.4	7	369.1
2011	11.6	21.8	2.0	239.0	7	305.3
2012	10.3	22.9	2.0	802.8	9	815.4
2013	7.6	19.5	0.5	397.6	8	452.7
2014	11.4	24.6	3.1	835.6	12	899.7
2015	11.0	22.3	2.1	240.2	6	312.5
2016	12.5	23.4	3.6	239.8	6	273.0
2017	11.3	22.9	1.7	619.4	12	708.2
2018	9.7	22.5	0.5	677.6	12	641.2
2019	10.9	23.4	2.5	1045.6	12	918.8
2020	10.4	23.3	3.1	765.0	11	753.5
2021	8.6	21.9	1.3	330.2	6	352.2

The statistical relationships between campus and HadUK-grid data are displayed in Table 3: the (non-detrended) monthly mean, maximum, and minimum air temperatures are extremely closely correlated, not falling below r = 0.996 for any month of the series. The regression slope between campus and HadUK-grid data is consistently close to 1 (0.967–1.015; Table 3) with no trend over the monitoring period. The mean regression intercept is around 0.6 (0.183–1.265; Table 3) and again exhibits no trend. The consistent relationships between campus and HadUK-grid data imply that there is unlikely to be any significant temporal bias in the campus data series.

Daily total rainfall for the 2008–2021 period from both campus and HadUK-grid data is shown in Figure 3. The rainfall series is less complete than the temperature series (Table 2), but in 11 of the 14 years, campus records lower total rainfall than HadUK-grid for months with complete data. HadUK-grid rainfall is 88%–130% (average 109%) of that recorded on the campus for complete months: this exceeds the < ±4% accuracy of the ARG100 rain gauge (Table 1). The wettest year recorded in the campus series is 2019, with 1045.6 mm rainfall, which is also the wettest year recorded by HadUK-grid. Missing data gives more uncertainty concerning the driest year, with 2011, 2015, and 2016 all recording similarly low monthly totals, but being only 6–7 months complete in each year. HadUK-grid data suggests that 2016 was the driest of these years, and this was a year in which rainfall was slightly below average nationally (Kendon et al. 2017).

[IMAGE OMITTED. SEE PDF]

The regression slope between campus and HadUK-grid data is again consistently close to 1 (0.808–1.206; Table 3) with no trend over the monitoring period. The mean regression intercept is around 0.6 (but with greater variability than temperature; Table 3) and again exhibits no trend. Again, this implies that there is unlikely to be any significant temporal bias in the campus data series. There is an episode of notable rainfall under-catch (Dunn et al. 2025) in June 2017, which leads to a relatively low correlation between campus (recording 33 mm) and HadUK (recording 71 mm) data (Table 3). Closer inspection of the daily series indicates that the campus rain gauge doesn't register a 12 mm fall on 2 June, nor a 29 mm total from 5–8 June, although it does then accurately register a daily fall of 15 mm on 27 June. The reason for this under-catch is obscure, particularly since the gauge appears to rectify itself later in the month and subsequently functions satisfactorily.

The location of the campus weather station predisposes it to relatively low wind speeds, as a result of the proximity of various buildings and trees (Figure 1). Speeds are generally around 1.0–1.5 m s⁻¹ in the summer months and 1.5–2.0 m s⁻¹ in winter. There is no significant trend in wind speed over the 14-year period, indicating no detectable impact from the growth of ‘Fruit Routes’ trees. However, there does appear to be a shift in dominant wind direction (Figure 4). 37% of winds overall are from the WSW–W–WNW direction. But the proportion of SW–WSW–W direction winds decreases from 11%–21% to 6%–13% over the 14-year series, while the proportion of WNW–NW winds increases from 3%–5% to 6%–14%, that is, the wind direction veers detectably during the monitoring period. There has been no significant change in buildings or in tree distribution during this time; the ‘Fruit Routes’ trees to the west of the weather station paddock are already in the lee of taller, pre-existing, mature trees (Figure 1). The magnitude of the veering is greater than the precision of both instruments used (Table 1) and the switchover between them occurs much earlier in the record. A notable succession of named storms crossed the UK in late winter and late summer 2020 (Kendon et al. 2021), and February of that year is the windiest month in the campus record, averaging 2.3 m s⁻¹, but the following year was unremarkable in this respect, nationally and locally. There is therefore no obvious explanation in the instrumentation or synoptic situation. The persistence or otherwise of this effect should therefore be examined closely in analysis of post-2021 data at the campus site (see following section).

[IMAGE OMITTED. SEE PDF]

Applications of the Dataset

Thus far, data from the campus station have been used to support building energy research: notably, wind speed and direction and solar radiation data included in this dataset were used in an experimental study of summertime overheating in residential dwellings, which compared different dynamic thermal models and measurements in synthetically occupied test houses (Roberts et al. 2019, 2022).

A further application of the dataset is to establish meteorological baselines for the campus location from HadUK-grid and UKCP. Subsequently, the data and its future extensions will be used to generate an air-mass transport climatology for the campus, where a network of EarthSense Zephyr air quality monitors was established in November 2022 to provide multi-site, real-time pollutant data. This will involve quantifying air-mass-type frequencies, their associated pollutant profiles, and whether and how these vary seasonally. Continued monitoring of the potentially changing distribution of westerly winds will be part of this approach. The combination of location-specific climatological baselines and a wind transport climatology will provide a basis for evaluating the impact of climate change on air quality: for instance, Hodgson and Phillips (2021) observed that high pollutant concentrations were associated with anticyclonic air-mass types, which implies potentially worsening air quality under more persistent future summer ‘Omega block’ heatwaves (Woollings et al. 2018; Nabizadeh et al. 2019).

From late 2021, the Loughborough University campus weather station transitioned to a new set of instruments to take advantage of fully wireless connectivity, an expanded range of ecological sensors, additional haptic rainfall measurement, an updated web interface, and automatic data archiving. Experiences and analyses from the upgraded station will be the subject of future publications.

Acknowledgements

The constructive comments of two anonymous reviewers are gratefully acknowledged.

Funding

The instrumentation used here was acquired through 2007 HEFCE Research Capital Funding to the Department of Geography, Loughborough University.

Data Availability Statement

The data that support the findings of this study are openly available in Loughborough University Research Repository at .

References

Burt, S. 2012. The Weather Observer's Handbook. Cambridge University Press. https://doi.org/10.1017/CBO9781139152167.

Dunn, R., H. Fowler, A. Green, and E. Lewis. 2025. “Tipping‐Bucket Rain Gauges: A Review of the Undercatch Phenomenon, and Methods for Its Reduction and Correction.” Weather 80: 196–205. https://doi.org/10.1002/wea.7736.

Hodgson, E., and I. Phillips. 2021. “Seasonal Variations in the Synoptic Climatology of Air Pollution in Birmingham, UK.” Theoretical and Applied Climatology 146: 1421–1439. https://doi.org/10.1007/s00704‐021‐03779‐7.

Hollis, D., M. McCarthy, M. Kendon, T. Legg, and I. Simpson. 2019. “HadUK‐Grid—A New UK Dataset of Gridded Climate Observations.” Geoscience Data Journal 6: 151–159. https://doi.org/10.1002/gdj3.78.

Kendon, M., M. McCarthy, S. Jevrejeva, and T. Legg. 2017. State of the UK Climate 2016. Met Office. https://www.metoffice.gov.uk/binaries/content/assets/metofficegovuk/pdf/weather/learn‐about/uk‐past‐events/state‐of‐uk‐climate/mo‐state‐of‐uk‐climate‐2016‐v4.pdf.

Kendon, M., M. McCarthy, S. Jevrejeva, A. Matthews, and T. Legg. 2019. “State of the UK Climate 2018.” International Journal of Climatology 39: 1–55. https://doi.org/10.1002/joc.6213.

Kendon, M., M. McCarthy, S. Jevrejeva, A. Matthews, T. Sparks, and J. Garforth. 2021. “State of the UK Climate 2020.” International Journal of Climatology 41, no. S2: 1–76. https://doi.org/10.1002/joc.7285.

Legg, T., S. Packman, T. Caton Harrison, and M. McCarthy. 2025. “An Update to the Central EnglandTemperature Series—HadCET v2.1. Temperature Series—HadCET v2.1. Temperature Series—HadCET v2.1.” Geoscience Data Journal 12: e284. https://doi.org/10.1002/gdj3.284.

Met Office. 2018. “Summer 2018.” https://www.metoffice.gov.uk/binaries/content/assets/%20metofficegovuk/pdf/weather/learn‐about/uk‐past‐events/interesting/2018/summer‐2018‐‐‐met‐office.pdf.

Met Office, D. Hollis, M. McCarthy, M. Kendon, and T. Legg. 2022. HadUK‐Grid Gridded Climate Observations on a 1km Grid Over the UK, v1.1.0.0 (1836–2021). NERC EDS Centre for Environmental Data Analysis. https://doi.org/10.5285/bbca3267dc7d4219af484976734c9527.

Nabizadeh, E., P. Hassanzadeh, D. Yang, and E. Barnes. 2019. “Size of the Atmospheric Blocking Events: Scaling Law and Response to Climate Change.” Geophysical Research Letters 46: 13488–13499. https://doi.org/10.1029/2019GL084863.

Prior, J., and M. Kendon. 2011. “The Disruptive Snowfalls and Very Low Temperatures of Late 2010.” Weather 66: 315–320. https://doi.org/10.1002/wea.874.

Roberts, B., B. Abel, D. Allinson, et al. 2022. A Dataset From Synthetically Occupied Test Houses for Validating Model Predictions of Overheating. Loughborough University. https://hdl.handle.net/2134/19412207.v1.

Roberts, B., D. Allinson, S. Diamond, et al. 2019. “Predictions of Summertime Overheating: Comparison of Dynamic Thermal Models and Measurements in Synthetically Occupied Test Houses.” Building Services Engineering Research and Technology 40, no. 4: 512–552. https://doi.org/10.1177/0143624419847349.

Sibley, A. 2023. “History of Weather Observing Sites in the Cambridge Area From the Mid‐NineTeenth Century to the Present‐Day.” Weather 78: 212–219. https://doi.org/10.1002/wea.4346.

Woollings, T., D. Barriopedro, J. Methven, et al. 2018. “Blocking and Its Response to Climate Change.” Current Climate Change Reports 4: 287–300. https://doi.org/10.1007/s40641‐018‐0108‐z.