Measurements of actual and pan evaporation in the upper Brue catchment UK: the first 25 years


Correspondence to: Colin Clark

About half of the solar radiation that is absorbed by the earth is used in evaporation (Trenberth et al., 2009). This is equivalent to a depth of evaporation of over 1000mm per year, or in excess of 0.5 million km3. It is, therefore, an important part of the hydrological cycle. In England measurements of evaporation have been made since the late nineteenth century (Symons, 1887); by the 1950s 16 tanks were in operation (Green, 1959; 1978) but the last of those stations closed in 2004 (Stanhill and Moller, 2008). The tanks were 1.8m square and 0.6m deep, and were sunk into the ground with a 6cm projection above the surface. In the 1960s, Green (1974) reported data from 32 irrigated lysimeters, a column of soil with a short covering of grass.

The importance of evaporation measurements cannot be overstated. The data are essential for monitoring the effects on the environment of changes in net radiation (Ohmura and Wild, 2002; Cong et al., 2009; Jung et al., 2010), land-use change (Law, 1957; Shukla et al., 1990; Clark, 1992) and other meteorological variables (Xu and Singh, 1998), for estimating the water balance of catchments (Penman, 1950; Dales and Clark, 1996), for better use of water resources (Allen et al., 1998) and for the provision of soil moisture deficit data for flood-warning systems (Pritchard, 1996; Clark, 2004). Given the widespread concern about climate change, the large decline in measurements of evaporation in Britain seems astounding, especially as decreasing potential evaporation is reported from many parts of the globe (Peterson et al., 1995; Liu et al., 2004; Moratiel et al., 2010; Roderick and Farquhar, 2004; 2005). Although many workers now calculate potential evaporation and soil moisture deficit from meteorological data, in England this has been shown to lead to serious errors (Clark, 2002; 2009). This paper will present the results of 25 years of evaporation measurements at a site in southwest England. In view of the near demise of direct measurements of evaporation, and bearing in mind the great interest in climate change, the results may serve as a benchmark for other studies which come as a result of this paper, or at least act as a comparison in the future.

It was Howard Penman (1948) who first combined the radiation and aerodynamic processes to produce an equation for evaporation. He compared his results with measurements made from sunken pans containing water and a grassed lysimeter. With the virtual cessation of measurements, the Penman equation, with modifications, is seen by many as the standard estimate. They forget that Penman used direct measurements to show his method gave reasonable results: his method needs measurements/estimates of radiation, windspeed, humidity, and temperature. The Institute of Hydrology (IOH), now the Centre for Ecology & Hydrology (CEH), has also made estimates of actual evaporation via rainfall and discharge measurements, and by using the eddy-correlation method at its experimental site at Plynlimon in mid-Wales since the late 1960s (Clarke and Newson, 1978), and this work is still in progress (Robinson et al., 2012). Another water vapour transfer method used to estimate evaporation is the Bowen Ratio, the ratio of sensible to latent heat. This is derived from the temperature and humidity gradients a few metres above the surface.

Before the measurement site in southwest England is described, it is vital to understand what the measurements of potential and actual evaporation mean and how they should be interpreted.

Potential and actual evaporation

Evaporation is the conversion of liquid water into water vapour. It can take place from the wet surfaces of lakes, bare earth, concrete, grass or any other vegetated surface. When water is lost from the breathing pores of plants – the stomata – the term transpiration is used, but since the processes of evaporation from non-living surfaces and also evaporation from the cuticle of plant surfaces is not precisely separated, the general term evaporation is used here. Direct measurements of evaporation can only be made over small sample plots. Evaporation from water, which is called potential evaporation when water is not limited, is usually measured using an evaporating pan, the most common being the class A pan: a galvanised pan 121cm in diameter and 25cm deep. It sits on a wooden platform resting on the soil. Water is kept in the pan to about 5cm below the rim. A raingauge, preferably kept at near ground level, is maintained at the same site and it is the difference in rainfall and change in height of the water surface that gives a measure of evaporation. Since the water depth is small compared to the surface area, and radiation can heat the sides of the pan, the results tend to be higher than the correct value. As there is no limit to the water supply, the results will depend on factors such at solar radiation, wind speed and humidity. There is no ‘memory’ of what has happened during the previous few days or weeks, although the relationship between monthly temperature and pan measurements does show some dependence on the recent past. Potential evaporation can also be measured using a well-watered drainage lysimeter; all data reported by Green (1974) are of this type. At the Rothamsted Experimental Station, in Hertfordshire, drainage lysimeters have existed since 1870, but they can only provide a reliable estimate of evaporation on an annual timescale since the full water balance is not measured.

Actual evaporation

Actual evaporation is best measured using a weighing lysimeter which consists of a column of soil with grass, held in a PVC tube with provision for drainage at the base. Rainfall will evaporate from the surface more when the grass is growing, much less when it is dormant. When the soil gets saturated the excess will drain out at the base. When the drainage ceases, the soil is said to be at field capacity and thereafter the soil will dry out until the grass wilts. By measuring the depth of rainfall, change in weight of the lysimeter, and any drainage that is collected in a vessel beneath, the full water balance of the lysimeter can be obtained. Although some authors believe that lysimeters are expensive (Allen et al., 1998) and need careful installation and maintenance (Henderson-Sellers and Robinson, 1986), they can be installed at low cost and little maintenance (Hinds, 1973; Rogowski and Jacoby, 1977; Jara et al., 1998; Clark, 2002). They have also been used to measure evaporation in upland areas (Calder et al., 1984; Hall, 1985). During springtime, when water is normally not limited, the rate of evaporation from the lysimeter should be very close to that from the evaporating pan. As the soil dries the plant roots find it increasingly difficult to get the moisture needed to keep the plant cells turgid and they close their stomata, thereby reducing the water consumed. At this stage the actual evaporation (AE) is less than the pan evaporation (PE). During wet weather in early summer AE can become higher than PE, because the latter depends heavily on solar radiation whereas plants will need moisture to grow and in so doing will evaporate more water. Plants only transpire during the day whereas evaporation from the pan will continue day and night. Measurement of AE from vegetated surfaces is much more important than PE because from the results growers can plan their schedule of irrigation if it is thought necessary: flood warnings can be made with much greater certainty and water balance of catchments can be better estimated on a daily, weekly or monthly basis. Over a period of years PE and AE (lysimeter) will respond in different ways to the controlling factors because of the limitation of soil moisture in the lysimeter. Many authors have struggled with the ‘evaporation paradox’, that is rising air temperatures yet decreasing PE (Brutsaert and Parlange, 1998; Ohmura and Wild, 2002; Liu et al., 2004; Cong et al., 2009; Zhang et al., 2011) without the use of AE data. At about the same time the notion of a complementary relationship between PE and AE was described (Hobbins and Ramirez, 2004), wherein PE decreased with increasing rainfall and AE increased.

It is expected that, arising from changes in climate, the hydrologic cycle will be intensified, with higher evaporation and rainfall, and more floods and droughts in some places. At present the evidence is by no means clear, with some of it contradictory (Huntington, 2006). This paper will present the results of 25 years on pan observations and 15 years lysimeter data gathered in the upper Brue catchment in southwest England. In view of the urgent need to understand what is happening to the hydrological cycle, this data set represents a significant addition to our knowledge. At this stage it should be made clear that it is not possible to use trends of the changes in PE and AE to predict the future. Any trends in the observations are only true for the time period over which the data have been gathered.

Site and methods of measurement

Charldon Hill Research Station (CHRS) is situated just outside Bruton in Somerset (Figure 1) at about 51°06’N and 2°27’W, near the centre of the upper Brue catchment of 135km2 at which riverflow is measured at the Lovington gauging station (NRFA 52010). Here a crump weir allows the measurement of low flows, while higher flows hare been measured with a current meter that measures flow velocity. Measurements made over a range of water levels allow the relationship between water level and discharge to be quantified. This gives good estimates of all except the highest floods. Further details of the catchment are given in Clark (2002). There are minimal human influences on the flow regime in the catchment, and it is essentially impermeable. Its importance will be considered below when the reliability of the AE data is considered. The instrumentation at CHRS is located on a small bench 76m above sea level on a southwest facing valley. The site is sheltered from northerly winds but exposed to southerly and southwesterly gales. Open pan evaporation has been measured since October 1985 using a 0.6m diameter PVC sunken pan which is about 60cm deep, with the water surface kept at about 5cm below the rim (Figure 2). A 50mm electrified wire mesh is placed over the pan in order to stop birds from drinking the water. The pan has shown no signs of deterioration after 25 years of service. There have only been two occasions in that time when the pan has been disturbed by larger mammals. The water level is measured on a daily basis using a steel angle-bar and screw pointer with a metric thread to give a reading to within 0.1mm; any errors made using this method are not cumulative. The advantages of a sunken pan are that it is not affected on the sides by direct and diffuse solar radiation and that it integrates the effects of ground heat storage.

Figure 1.

Location of the upper Brue catchment in east Somerset. L = Lovington.

Figure 2.

The open pan at CHRS. Note the ground-level raingauge and the 50mm wire mesh to keep off animals which might drink some of the water.

Since 1995 there have been two lysimeters at the site, made using a 0.3m diameter PVC pipe about 0.5m deep with a bevelled edge at the top. The soil column, which already had well-established grass present, was obtained by excavating the soil as a monolith and then sliding it into the plastic tube (Figure 3). A good seal at the base of the lysimeter was obtained by using galvanised steel held in by a wooden base and sealed with silicone sealant. A drainage hole at the base allows excess water to drain into a bucket (Figure 4). The lysimeters are neither grazed, watered or fertilised; each one sits over a plastic bucket in which drainage water is collected and weighed or, in the case of a low volume, measured volumetrically. The lysimeters are weighed daily using a portable electronic balance accurate to 0.02kg, which is equivalent to an accuracy of about 0.5mm. The site has been registered with the Met Office since 1990 for daily rainfall readings. The grass consists of rye-grass (Lolium perenne L.) and some white clover (Trifolium repens L.) and is representative of the surrounding grass; it is regularly cut so as to be no higher than 10cm, and the clippings are left on the surface. The surrounding area is treated in the same way. The advantages of this design of lysimeter are that it is cheap to make and install, it has shown no signs of deterioration over the 16 years of use, and the measurements of mass and drainage volume are simple and quick to achieve.

Figure 3.

Lysimeter at CHRS. Note the well-kept grass, with no overlapping blades at the boundary with the surrounding vegetation.

Figure 4.

Structure of the lysimeter, showing how drainage is captured. An alternative design could include the provision for runoff during a high-intensity storm.

The design of both pan and lysimeters takes into account the desirable features and precautions noted by Jones (1992) and Fu et al. (2009). An essential feature is to have a depth:diameter ratio close to or in excess of unity, which will avoid the water overheating found with a ratio of less than 1.0. Both lysimeter and pan are made from material with a low thermal conductivity (Brutsaert, 1982).

The water balance of the lysimeter is given by:

display math(1)


P = precipitation measured (kg)

D = drainage (kg)

ΔS = storage change (kg)

E = evaporation (kg), which is converted to mm by the lysimeter constant of 0.0437, the weight of 1mm rainfall over the lysimeter surface.

Rainfall measurements using the standard raingauge have been compared with a ground-level gauge (with no significant difference found).

The AE is calculated using:

display math(2)


AE = actual evaporation (mm)

W0  = weight of lysimeter at start time (kg)

W1  = weight of lysimeter at end time (kg)

P = precipitation measured (kg)

D = drainage (kg)

K = lysimeter constant

Soil moisture deficit (SMD) is the depth of rainfall needed to bring the lysimeter soil up to field capacity. Any additional water will lead to a loss due to drainage. Since 2004, the daily measurement of SMD has been recorded on a website accessible to the Environment Agency which can be used in a real time non-linear flow model. This was designed by the author for use as a flood warning method on the upper Brue for Bruton, which is located 2km downstream of a flood-detention dam (Clark, 2004). This method has vastly improved an earlier procedure of flood warning based on the rate of rise of floodwater behind the dam, which led to several false warnings in the past.

In addition to the above measurement, a standard Stevenson screen is present and maximum and minimum temperatures are recorded. At this point it is important to assess how representative the site is of a wider area. The crucial factor is the selective exposure and site in a tributary valley of the upper Brue. It might be expected that the temperature regime would be unrepresentative. Table 1 shows the Pearson correlation of CHRS and Central England Temperature (CET) monthly temperature data. All correlations are significant at greater than the 1% level; apart from May and June all correlations exceed 0.9, and the annual correlation is also above 0.9. Together, the results show that CHRS temperatures are comparable to CET, with typical annual temperatures being about 0.5 degC higher, a result to be expected for this part of southwest England. Bearing these facts in mind, if it can be shown that the results are representative of, say, a catchment whose area is around 100km2 then this would be an important result since heterogeneity often bedevils the value of site-specific surveys. At the same time it is reasonable to expect the results to represent at least the surrounding 1km2, but without many replicated sites this is not guaranteed.

Table 1. Correlation and regression analysis for CHRS and CET measurements 1986–2010
MonthCorrelationRegression slope (a)Constant (b)
  1. CET = CHRS a + b.


Data reliability

According to Shuttleworth (2008), weighing lysimeters provide the best level of accuracy for measuring AE. There have, though, been very few comparisons of pan and lysimeter evaporation on a daily or two-day time scale. Brutsaert (1982) reports on a comparison of mean monthly evaporation from a class A pan and evaporation from grass. A pan coefficient or correction factor of 0.8 was produced, and this raises a question regarding the effect of both pan design and site conditions on the pan coefficient. Soon after the start of the lysimeter measurements at CHRS the two-day data from both pan and lysimeter were compared. The latest of these comparisons is shown in Figure 5 which was based almost exclusively on two-day data, except where the lysimeter could not be read due to rainfall at 9am: there is a very close comparison between the two data sets. When the SMD reaches 70mm the supply of water in the lysimeter becomes depleted, and the rate of evaporation declines. When the SMD returns to below 70mm the rate of evaporation from the lysimeter continues very much as that from the pan in which water is not limited. The agreement between pan and lysimeter may be real or fortuitous, but the close comparison with the catchment losses (see below) suggests that the data are realistic – which means that the pan in use does not need a correction factor.

Figure 5.

Changes in cumulative evaporation as measured by the evaporating pan and lysimeter.

A further indication that the lysimeters represent a much bigger area and hence are of reasonable accuracy comes from direct observations of the behaviour of the River Brue at Bruton. It has been observed that, when both lysimeters drain, the river has risen by at least 0.1m. Smaller rises can take place when the SMD is significant due to runoff from roads, track-ways and other mainly impermeable surfaces upstream of Bruton. But when the lysimeters are at field capacity, large scale runoff from the catchment takes place following rainfall. Finally, as will be shown below, over the past 15 years the annual AE as estimated by the lysimeters is very close to the losses as estimated from riverflow and rainfall in the catchment at Lovington gauging station. There will always be some uncertainty in whatever measurements have been made, especially with measurements of rainfall. But when these have been considered so far, the measurements of evaporation at CHRS seem reasonable.

Changes in annual PE and AE

Figure 6 shows the changes in PE between 1986 and 2010. These show similarities to records of pan evaporation (Peterson et al., 1995), wherein a flattening of the trend of evaporation has taken place since the 1990s (Moller and Stanhill, 2007). The main difference with other published results is the positive trend of pan data from 1986 to 2000, (r = 0.51, sig. 5% level) when many sites, but not all, showed a decline. The nearest site for comparison is at Wellesbourne (Warwickshire) (Stanhill and Moller, 2008), where from 1986 to 1990 there was a similar increase in PE, with a flattening of the trend afterwards.

Figure 6.

Changes in open pan evaporation 1986–2010 at CHRS.

Figure 7 shows the changes in annual AE from 1996 to 2010, as well as the results of Jung et al. (2010). The time period was characterised by two low readings for the first and last years and when these are excluded from the temporal Pearson correlation, r = −0.70, sig. 1%. A global network (FLUXNET) of in situ measurements has only been set up during the past decade (Jung et al., 2010). They report a mean annual global land surface evaporation of 65.5 × 103km3, which compares with a reading of 75 × 103 km3 from data in Trenberth et al. (2009), using a figure of 2.4 MJkg−1 for the latent heat of vaporisation of water. The pattern of changes reported in Jung et al. (2010) shows a remarkable similarity with the measurements of AE at CHRS, that is an increase from 1996 until 1999, followed by a decline and then an increase. The recorded AE of Jung et al. (2010) goes up until 1998 and then declines until 2002. Thereafter it increases until 2005 and then declines. In summary, the trends of AE at CHRS are given in Table 2.

Figure 7.

Changes in actual evaporation 1996–2010 and the results of Jung et al. (2010).

Table 2. Trends in AE measured at CHRS
Time periodPearson's rRegression slopeRegression constant
  1. a

    p > 0.01. ns, not significant.

1996–2010−0.40 ns−4.18218827.3654
1997–2009−0.70a−5.549411 579.9340

Possible causes of the changes

A variety of causes may contribute to the changes in PE and AE at CHRS. In this paper only temperature, rainfall, mean maximum and minimum temperature, and the daily temperature range, are examined for annual data, while for monthly changes the effects of temperature and rainfall have been studied. Table 3 shows the Pearson correlation values for annual data. The results show that both PE and AE rise with increasing annual rainfall, although the correlation coefficient for PE is lower. Temperature has a positive effect on both PE and AE. Plant growth and hence AE appear to be favoured by higher night-time temperatures, more so than PE. The opposite seems to hold true during the day with PE being more strongly influenced by higher temperatures than AE. The daily temperature range (DTR) is often taken as a measure of solar radiation (Dai et al., 1999). The present results suggest that as DTR decreases with increased cloud cover and less solar radiation, AE increases. The opposite seems to apply to PE (Roderick & Farquhar, 2002). This is what has been referred to as the complementary effect (Hobbins and Ramirez, 2004). Increased cloud cover would suggest higher rainfall which fits in with the different effect of rainfall on PE and AE.

Table 3. Relationships of PE and AE with annual temperature, rainfall, mean min and max temperatures, and daily temperature range (DTR)
 TemperatureRainfallMean minMean maxDTR
  1. a

    p > 0.05;

  2. b

    p > 0.01;

  3. c

    p > 0.00. ns, not significant.

PE0.51b0.18 ns0.25 ns0.64c0.54b
AE0.77c0.58a0.79c0.56a−0.33 ns

Analysis of monthly observations


A better understanding of the processes that are involved in evaporation comes from an analysis of monthly data. The effect of temperature on PE and AE during the year is shown in Figure 8, illustrating the changes in regression slope and correlation between PE, AE and temperature respectively. In January the effect of temperature on AE is rather muted, but it becomes important by April with regression slopes of 6–7 mmdegC−1 and correlations in excess of 0.5, a significant value for both PE and AE, p > 0.05. The slope and correlation for PE peak in June and July respectively. In contrast to this, the effect of temperature on AE almost disappears during July. This illustrates the changes in soil moisture storage during some years as compared with others: the effects of the previous months’ lack of rainfall will show itself during July by reduced AE as compared with years when the storage of soil water is high enough not to limit evaporation. By September soil moisture levels recover as rainfall increases, so the effect of temperature on AE becomes more significant because soil moisture is not a limiting factor.

Figure 8.

Changes in the regression slope and correlation coefficient of monthly temperature, pan and lysimeter evaporation.


The correlation of PE and AE with monthly rainfall is shown in Figure 9. During the summer the effect of higher rainfall and therefore increased cloud cover tends to reduce PE. The effect of rainfall on annual AE is similar on a monthly timescale. This is because during the early spring the soil is at field capacity, and higher rainfall will reduce the rate at which the soil warms up as well as lead to reduced solar radiation. These are the limiting factors at this time of the year, but they appear to weaken during the summer since the effect of dry conditions becomes more important. By the end of a typical summer, water is often the limiting factor: this helps to explain the sudden increase in correlation with rainfall during September, an effect that does not last into the winter as the growing season comes to an end. The overall changes in the correlation can be explained by the fact that both PE and AE need solar radiation, which is reduced during wet weather.

Figure 9.

Changes in the regression slope and correlation coefficient for monthly rainfall, pan and lysimeter evaporation.

The regression slope of the relationship between rainfall, PE and AE (Figure 9) is less easy to interpret because there is much less variation in slope than for the temperature-based relationships. The regression slope for PE in winter is close to zero; during spring it decreases, reaching a low in June of −0.24mm PE per mm of rainfall. Thereafter it increases during the remainder of summer and by September is close to zero. The pattern of changes in regression slope for AE is less clear, but there is a tendency for the slope to decrease during the spring and early summer and thereafter to increase to a maximum of 0.09mm AE per mm of rainfall during September. The increasing importance of rainfall during the drier months of the year suggests that at this time rainfall can be the limiting factor in AE, especially if there has been a dry summer followed by a wet September, such as in 1999.

Daily temperature range (DTR)

The effects of DTR on PE and AE during the year can be explained by the changing average SMD at the end of each month (Figure 10). Warmer weather during the growing season will be associated with a higher DTR and normally lead to higher pan evaporation. This also applies to the lysimeter when soil moisture is not limited but its effect declines during the remainder of the growing season as the SMD rises. Higher rainfall in September might be expected to give a higher correlation between AE and DTR but higher rainfall is often associated with lower temperatures and reduced solar radiation.

Figure 10.

Changes in the value of Pearson's r for daily temperature range, pan and lysimeter. The vertical lines are SMD (mm) at the end of the month.

Sensitivity of AE to temperature changes

It is generally believed that with rising temperatures evaporation will increase (Budyko, 1982). The changes in AE may be more difficult to assess although some attempts have already been made (Arnell et al., 1990; Lockwood, 1993; Ziegler et al., 2005). The latest UKCP09 review of the potential impacts of climate change is not specific about changes in actual evaporation over the British Isles. The data set presented in this paper does allow an estimate of the likely changes in AE with a change of temperature. Partial regression equations were calculated for monthly temperature and rainfall in relation to AE for the 15 years of data. The presence of multicolinearity (Hauser, 1974) was checked and all correlations between the two independent variables were below 0.8. A value in excess of 0.8 is considered to be indicative of multicolinearity, which would exclude both rainfall and temperature being related to AE at the same time. The result gave an increase of 16mm per year for a change of 1 degC, which represents an increase of 3.6% in AE. When the same analysis was performed on annual data the result was +13%, which is an effect of the timescale used. The result of 3.6% is comparable with results of 4% (Budyko, 1982) and 3% (Lockwood, 1993) at a temperature of 10 °C. The greatest effect will be during the spring and early summer with the least effect during autumn and winter.

In order to examine the realistic effect of temperature on AE the analysis was extended to seasonal: April to September, and October to March data periods, and half- monthly data. The results are shown in Table 4. The reasons for the differences in the expected percentage change in AE with a change of 1 degC is on account of the larger data set for smaller time periods, which gives more information about changes on smaller and therefore more meaningful timescales. As the number of pairs of data increase the degrees of freedom also increase as shown in Table 4. Analysing data spanning timescales greater than one month leads to an averaging of the evaporation processes and their controls. Annual data represents the average conditions over the whole year, thereby obscuring the changes during the spring and summer when most of the evaporation takes place. Typically, in late June and July increasing temperature causes a decrease in AE because the soil has dried out: AE is less than PE. This feature is not found when either seasonal or annual data is studied. It appears that for AE a timescale of about one month leads to a stable result for the effect of a temperature change of 1 degC on AE.

Table 4. The effect of time period of analysis on the estimate of the effect of +1 degC on actual evaporation, with rainfall held constant
 % changeDegrees of freedom
Annual data1328
Seasonal (winter/summer)1156

A simple analysis of the trends of catchment losses and rainfall at Lovington showed that there are no significant trends through the period 1965–2009. This is in spite of an upward trend in CET of 0.2 degC per decade since 1965 and a Pearsoncorrelation for CET of 0.54, p > 0.01. On this basis the upper Brue does not seem to be as sensitive to a change of temperature as does the lysimeter.

Catchment evaporation

Since the measuring site is located close to the centre of the upper Brue catchment as monitored at the gauging station at Lovington (Figure 1), it is worthwhile comparing the results of the point measurements with the catchment as a whole. It should be emphasised that initially the lysimeters were installed just to measure evaporation. On completion of the non-linear flow model for the Brue used as a flood warning tool by the Environment Agency, the author was asked to provide, on a daily basis, the readings of SMD from the lysimeters. A measure of how representative the lysimeters are of the catchment as a whole comes from observations made of the behaviour of the Brue at Bruton: when both lysimeters drained, the river rose significantly. This behaviour has been checked for over 10 years in an attempt to establish how representative the lysimeters are of the catchment above Bruton, whose area is 31km2. At Lovington the area is about 135km2 so although the river will have risen on the same occasions as it did at Bruton, the question is whether the measurements of AE at CHRS are representative of the catchment at Lovington. Measurements of monthly rainfall from 1996 to 2008 at CHRS are representative of the catchment at Lovington, r = 0.99, p > 0.01: Lovington = 0.96777CHRS+ 2.908. It is not possible to get reliable catchment measurements of evaporation on a timescale much shorter than a year because the catchment cannot be weighed! There are unmeasured storage components and rainfall from one month to another which may not appear in the river until the following month. Therefore the best measure of comparison is the annual average evaporation compared with the losses from the Brue catchment via the water balance:

display math(3)

It is assumed that there are no changes in storage. This may not be true on an annual basis but, over a long period of time, should be a reasonable assumption since there are no imports or exports of water from the area and the catchment is mainly underlain by clay which prevents any significant leaks. Each water year starts on 1 October, when groundwater levels are normally at their lowest, and ends on 30 September.

To date there are 12 years data which can be used in the comparison, the data for 1998/1999 being incomplete due to work being done on the gauging station. The average annual catchment loss for the period 1996–2010, with the exception of 1998/1999, is 440mm. The average measured AE at CHRS for the same years is 452mm. These two observations, which will always be subject to measurement error, are highly comparable. The long-term loss (n = 43 years) for the Lovington gauging station is 446mm. Comparison of the losses in individual years is very similar at the two locations from 2001 onwards: the biggest difference was for the water year 1999/2000: 404mm losses and 536mm AE. This was mainly because of the high rainfall of August and September 1999, in the previous water year. This led to 81% runoff in October, making up about 50mm of the difference between AE and catchment losses. The rest of the difference is largely made up in June 2000 when the water balance gave 5mm losses while the lysimeter lost 95 mm. This greater loss was from the soil moisture storage, which cannot appear via the water balance. Hence, although the water balance can be used to assess evaporation on a long term basis, it is not sensible to use it for shorter timescale comparisons or assessments.


It has been the object of this paper to describe the results of 25 years of measurement of evaporation. By using a sunken pan and comparing the results with a weighing lysimeter, it was shown that a pan correction factor is not necessary. Both pan and lysimeter are low cost, need little maintenance and are simple to gather data from.

The main findings are:

  • Changes in PE have been rather muted over the past 25 years.
  • PE increased by about 50mm from 1986 to 1999 but thereafter declined by about 30mm from 2000 to 2010.
  • PE and temperature are correlated, and the lack of a recent trend in annual temperature has meant that PE has not increased significantly.
  • AE has declined by about 15% during the past 15 years: this seems to be related to a 9% drop in rainfall.
  • Changes in AE from 1996 to 2008 are reflected in the global FLUXNET data set. In this way the observations at CHRS may be representative of a wider area and so have greater significance than a single site normally attracts.
  • The annual average AE at CHRS is very close to the catchment average losses for the river Brue at Lovington. This result is significant because it shows that the measurements are representative at the catchment scale of around 100km2.
  • Since the temperatures are closely related to CET it is likely that the results are applicable to other clay catchments at least those with a similar rainfall regime.
  • The predicted changes in climate, via higher temperatures, of greater rainfall and AE are not currently apparent in this study.

The programme of data collection will continue in the future, with a short-rooting chalkland lysimeter being installed during 2011 and a similar lysimeter in the river Shreen drainage basin at Mere in Wiltshire. Further experimentation will include the provision for the collection of surface runoff from a new lysimeter, which may provide a better comparison between catchment losses at Lovington, and AE from a grassed lysimeter.