Design and construction of a large weighing lysimeter in an almond orchard

Effective water management is essential to ensure the sustainability of irrigated agriculture. The accurate determination of crop water requirements is the first step in this task. This paper describes the building of a one-tree weighing lysimeter (3 × 3 m and 2.15 m depth) located in an almond (Prunus dulcis cv. Guara) orchard, inside the experimental farm “Alameda del Obispo” in Córdoba, Spain, to measure orchard evapotranspiration (ETc). Following a review on lysimetry, the description of the construction of the weighing lysimeter is provided in detail, including considerations relative to system resolution and wind effects on the measurements. Finally, some preliminary results of the evaporation and transpiration of young almond trees are presented demonstrating that lysimetry in orchards provides accurate ETc values needed to determine irrigation water requirements. Additional key words: crop coefficients; evapotranspiration; irrigation; wind effects.


Introduction
Given the competition for water resources, irrigation must be managed with the aim of improving the efficiency of water use.Precise knowledge of the crop water requirements (evapotranspiration; ET c ) is the starting point for improving agricultural water re-sources management.There are many methods to measure or estimate ET c .Micro-meteorological techniques, such as the Bowen ratio and eddy covariance methods, are now commonly used for measurement of ET c (Todd et al., 2000;Payero et al., 2003;Testi et al., 2004).Even though measurements of ET c from weighing lysimeters have been conducted since long ago Design and construction of a large weighing lysimeter in an almond orchard separating evaporation (E) and transpiration (T) (Klocke et al., 1985;Qiu et al., 1999), determination of basal crop coefficients and water requirements for specific Northwest US irrigated crops (Wright, 1982), alfalfa (Benli et al., 2006) or garlic (Ayars, 2008), evaluation of methods to determine ET o (López-Urrea et al., 2006a,b), analysis of the relationship between evapotranspiration and soil water content (Brun et al., 1985), deficit irrigation studies in peach trees (Girona et al., 2002), analysis of the energy balance components (Aase & Siddoway, 1982), integration of Time-Domain Reflectometer (TDR) measurements and lysimetry (Young et al., 1997), and finally, correlation between canopy light interception and crop coefficients (K c ) in apple and pear trees (Girona et al., 2011).
In Spain, almond is an important tree crop, considering its cultivated area of 578,000 ha, of which only 6.7% is irrigated (Anuario de Estadística Agroalimentaria, 2010).Until recently, almond was considered a marginal crop in southern Spain, with average yields of around 200 kg ha -1 (Arquero et al., 2002), while under irrigated conditions yields over 2,000 kg ha -1 are possible (Arquero et al., 2002).These modern plantations are still scarce, but have high levels of productivity and profitability.Research related to almond irrigation management has been primarily carried out in California and Spain (Goldhamer et al., 2003(Goldhamer et al., , 2006;;Romero et al., 2004a,b;García et al., 2004;Girona et al., 2005).These analyses have been mainly focused on regulated deficit irrigation studies, but information about its water requirements is still very limited.
The objective of this study is to describe the construction and performance of a weighing lysimeter containing a single almond tree, which was constructed within a new experimental almond orchard facility located in Córdoba, Spain.In addition, some young almond orchard evapotranspiration measurements collected in the weighing lysimeter are reported.

Field characteristics
The field where the weighing lysimeter was built is located in the province of Cordoba, southern Spain (37°51.5'N, 4°48.1'W), at 91 m of elevation and close to the Guadalquivir River.The field is flat and is 5 ha in size.Most lysimeter facilities are located in fields that do not exceed this area (Allen & Fisher, 1990;Jia (e.g., van Bavel & Myers, 1962), they are still considered one of the most accurate methodologies to study the return of water to the atmosphere by the combined processes of evaporation and transpiration (Malone et al., 2000), and are the standard method to directly measure ET c (Reicosky et al., 1983;Payero & Irmak, 2008).
Weighing lysimeters have a long history of development, and different designs have been used.Pruitt & Angus (1960) presented one of the initial reports on lysimetry.Some of the first lysimeters consisted of a big shallow tank (Pruitt & Angus, 1960), small rectangular containers (van Bavel & Myers, 1962), or small cylindrical tanks weighed with scales (England, 1963).Subsequent developments led to an increase in lysimeter size which then required the use of weighing mechanisms with advanced counterbalance weights (Ritchie & Burnett, 1968;Armijo et al., 1972).Since that time, the majority of lysimeters incorporated balance beam and counterweight mechanisms (Malone et al., 1999) that offset the dead weight from the soil and the container.A number of lysimeters using this methodology have been built (Wright, 1982;Marek et al., 1988;Howell et al., 1995Howell et al., , 1998) ) been the majority devoted to determining the ET c of the major annual crops, such as alfalfa (Hunsaker et al., 2002), wheat (Dugas et al., 1985) or sorghum (Ritchie & Burnett, 1968).
More recently, improvements in design and cost of commercial load cells and data acquisition systems have enabled the design and installation of lysimeters which are completely supported by load cells without balance beam mechanisms or other moving parts (Allen & Fisher, 1990).The major drawback of the design as compared to mechanic designs is that the entire dead load (soil, container and crop) and live load of the tank are measured.Some lysimeters using load cells are described by McFarland et al. (1983), Schneider et al. (1998), Tyagi et al. (2000Tyagi et al. ( , 2003)), Barani & Khanjani (2002), Girona et al. (2004), Jia et al. (2006), Loos et al. (2007), and Payero & Irmak (2008).
Agronomic applications of weighing lysimeters have been numerous.Among them, comparisons and analyses of different evapotranspiration estimation methods (Parlange & Katul, 1992;Kashyap & Panda, 2001), verification of the reliability of the ET c estimates by means of the most recent updates of the FAO method (Lovelli et al., 2005), measurement and comparison of ET c in different cultivars (Ehlig & LeMert, 1976;Howell et al., 1998), analyses and validation of models et al., 2006;Payero & Irmak, 2008).In our field, the lysimeter is located approximately in the center of the plot.Howell et al. (1985) indicated that a fetch of at least 100 m would be desirable for temperature and vapour pressure profile stability, and other authors have insisted in this fact (Marek et al., 1988;Barani & Khanjan, 2002).Following these recommendations the minimum fetch in the "Alameda del Obispo" facility is 180 m in the predominant wind direction of West-East.
The soil in the experimental field "Alameda del Obispo" is uniform in the two upper meters due to the alluvial origin, and is described in Table 1.Deeper in the profile, an increase in sand content of the soil was detected.The soil is very deep (over 4 m) and there is no sign of a water table above that depth.
After completion of the construction of the lysimeter, the plantation was established in late February 2009 with almond trees (Prunus dulcis, cv.Guara).This self-fertile cultivar has a late flowering period adapted to frost that usually occurs in the area around late February.Although no pollinators were interplanted, the external rows of the field (around 200 trees) were planted with different almond varieties (Belona, Soleta, Vairo, Constantí, Marinada, Tarraco, Ferragnes, Ferraduel, Masbovera, Glorieta, Antoñeta and Marta).
Trees were planted at 7 m × 6 m spacing, for a total of around 1,300 trees in the field.A drip irrigation system was installed with a single drip line per row and emitters spaced 1 m with a discharge of 2.4 L h -1 .Every tree is thus irrigated with six emitters, including the one in the lysimeter.During summer 2009 measurements to determine the volume of the canopy and the fraction canopy cover for the tree located in the lysimeter were carried out using digital images software processing.The canopy volume in mid July was around 0.6 m 3 and the fraction cover around 4% (horizontal projection of the canopy was 0.34 m 2 ).
An automatic weather station (Campbell Scientific Inc., Logan, UT, USA) was used to collect standard meteorological data (wind speed and direction, solar radiation, relative humidity and temperature) at 10 min intervals over a 1.5 ha grass plot located 600 m away from the lysimeter.The grass plot was irrigated to meet full water demand (two irrigation events per week) and weekly mowed to obtain a homogeneous grass height of 0.12 m, following the recommendations of Allen et al. (1998) to satisfy the Penman-Monteith assumptions.The reference evapotranspiration (ET o ) was calculated using the Penman-Monteith equation (Penman, 1948;Monteith, 1965Monteith, , 1973)), applied on an hourly basis, using the normalised procedure of FAO (Allen et al., 1998), assuming a standard grass surface resistance of 69 s m −1 during the day and 500 s m −1 during the night, and calculating aerodynamic resistance as 208/u (s m −1 ), where u is the wind speed at 2 m height (m s −1 ).

Construction process
Lysimeter construction started on February 2008 with the excavation.Due to the special requirements of size, Design and construction of a large weighing lysimeter in an almond orchard the total dug volume was around 100 m 3 (Fig. 1 and Suppl.Fig. 1, pdf).The duration of the construction was approximately 1 year, and 3 months more were required for the installation and calibration of the electronic devices (load cells, dataloggers and pumping system).
When the soil is highly structured with distinct profile characteristics, having an undisturbed soil monolith is essential for project objectives (Armijo et al., 1972;Marek et al., 1988;Schneider et al., 1996).However, due to the homogeneity of the upper layers of the soil profile (Table 1), it was deemed not necessary to keep the soil undisturbed in the construction process.
Others authors proposed to remove the soil in small increments and then to refill the container maintaining the same characteristics as the surrounding soil (Ritchie & Burnett, 1968;Phene et al., 1991;Ayars et al., 2003;Johnson et al., 2005).However, in relation with soil representativeness, Khan et al. (1993) indicated that if roots are well developed and nutrients and water supply are unrestricted, dissimilar soil may not cause significant variation in water use and yield.
In our case, the soil was separated into two layers (surface layer of 20 cm and the rest).To achieve a bulk density similar to that of the original soil (around 1.55 g cm -3 ; Table 1), in the backfilling process, the soil was saturated with water every 30-40 cm depth increments, following a procedure described by Ritchie & Burnett (1968).
The buried control room contains the infrastructure for the load cell installation and instruments for pumping, data storage, etc.It was built with reinforced concrete with dimensions 4.3 m × 3.9 m with a depth of 4 m (Fig. 1 and Suppl.Fig. 1, pdf).
A drainage system is essential to avoid waterlogging in the container and it was installed before the soil refill process.Lysimeter drainage systems include vacuum-assisted systems (Phene et al., 1989), porous bricks (Bhardwaj & Sastry, 1979) or suction candles built with porous ceramic tubes (Kirkham et al., 1984).In our case, a 20 cm bed with pebbles with diameters around 1-2 cm (Suppl.Fig. 2b, pdf) and a pipe system of 2 cm of diameter inserted in the pebble bed were installed (Suppl.Fig. 2c, pdf).In order to avoid soil contamination of the pebbles-pipes system, a geotextile layer was placed between the soil and the drainage system (Suppl.Fig. 2a, pdf).The pipe system collects the drainage water that is stored in two plastic containers (Suppl.Fig. 3f, pdf) attached to the steel container and accessible from the control room.These containers have a capacity of 200 L each and are connected by a pipe to the lysimeter irrigation system (Suppl.Fig. 3e, pdf).A complete description of the drainage system and the access to the control room is shown in Figure 1.
The gap between the concrete wall and the stainless container is less than 1 cm to avoid alteration of the energy balance of the system (Barani & Khanjani, 2002).This gap has been covered in the surface with a flexible and impermeable PVC film (Suppl.Fig. 4, pdf).The container is supported and its weight measured by four load cells (Suppl.Fig. 3b, pdf).The installed load cells (HBM RTN) have an accuracy class C3 (according to OIML R60), with a nominal load of 10 t each and a very limited temperature effects and hysteresis and nonlinearity errors.The four load cells are connected in series and installed with radial symmetry around the centre of gravity.This setup allows to reduce the wind effects on the measurement because the unload of one load cell (caused by the wind) is compensated by the overload of the load cell located in the opposite place.The electrical signal from the four load cells is transmitted to a transducer device (model GI-308, Baxtran-Giropés, Girona, Spain) located in the control room (Fig. 1).This device, in addition to displaying the weight in a screen, transforms the signal from voltage (mV) to a digital signal that is collected by a datalogger (Model CR1000, Campbell Scientific Inc., Logan, UT, USA) through a RS-232 connection.The data collection protocol includes averaged weights, standard deviation (SD), signal and stability of the measurement.The maximum resolution of this device is 0.5 kg within a limited weight range (± 800 kg), so the system must be tared to have a weight around 800 kg at field capacity at the beginning of the irrigation season.The frequency of data collection was one measurement per second; averages were then calculated for each 5-min interval and stored in the datalogger memory.SD for the 300 instantaneous measurements was also calculated.This measurement frequency helped reducing adverse wind effects and is very high compared with previous studies of one measurement per minute (Payero & Irmak, 2008) or per 5 s (Dugas et al., 1985).Finally, in order to analyse possible temperature effects on load cells resolution, the temperature of the control room is also measured and stored.
Following the data provided by the manufacturers of the load cells and the transducer device, and considering the container surface area (9 m 2 ), the resolution of the lysimeter installed is equal to 0.056 mm.This value can be considered acceptable following the indications of other authors (Marek et al., 1988).In previous works, lysimeter resolution ranged between 0.01 mm for small lysimeters installed in the US Water Conservation Laboratory in Texas (van Bavel & Myers, 1962;Hunsaker et al., 2002) and 1 mm for a lysimeter in a peach orchard (McFarland et al., 1983).Since the relative error decreases with the magnitude of the ET c measured, the resolution obtained here is adequate for measuring daily ET c and longer periods, and even could be appropriate for shorter time steps.In order to ensure the precision of the measurement in the described lysimeter, six-monthly calibration checks by technical staff are carried out.
Measurement of crop evapotranspiration in conventional lysimeters is difficult when frequent irrigations events are required because the water added from an external source increases the mass of the lysimeter simultaneously with the decrease occurring from the evapotranspiration process.To avoid this, the lysimeter was equipped with a system that integrates its water supply which comes from containers attached under the lysimeter itself (Suppl.Fig. 3f, pdf).These water reservoirs are weighed with the lysimeter and refilled every day (preferably when ET c is near minimum).Similar devices, described previously by Phene et al. (1989), Girona et al. (2002Girona et al. ( , 2004) ) and Ayars et al. (2003) allow the automatic measurement of ET c without interruption (Phene et al., 1989).This approach is especially useful in orchard trees with drip irrigation that require frequent irrigation events.However, in order to avoid a possible increase of salinity in the long term, the tree located in the lysimeter can also be irrigated with the field irrigation system.
The weighing lysimeter required a buried control room to contain the load cells, support of load cells, drainage containers, pumping system for irrigation, datalogger and electric devices.Equally a pumping system (independent of the lysimeter) was installed in the control room in order to extract possible water filtrations.

Load cells installation
The load cells used in this lysimeter are extremely accurate but require a very careful installation under the stainless steel container.Initially six load cells were planned and installed but it was necessary to reduce its number to four due to elongation problems in the contact between the container and the load cells.This reduction in the number of load cells did not reduce the accuracy, although limited the maximum weight (from 60 to 40 t).In addition, four load cells increased the stability of the system, reducing the wind effects on the measurements.
The weight of the whole system (~ 35 t) caused significant deformations in the union between the container and the load cells, producing significant measurements errors as the weight values increased indefinitely without any physical reason.To avoid this problem, the base of the container was reinforced with an IPE steel beam support that avoid any deformation (Suppl.Fig. 3d, pdf), and a steel ball was inserted in the contact of the load cell with the support of the container (Suppl.Fig. 3c, pdf).On this support, the container was installed without any other type of union, allowing the free movement of the system.Design and construction of a large weighing lysimeter in an almond orchard

Electrical damages
Electrical devices installed under field conditions could be damaged by lightning.In this case electronic load cells and electricity devices such as dataloggers may be in danger and require the installation of lightning protection systems.For this purpose a pressurizedgas high-voltage circuit breaker was installed upstream the load cells transducer; additionally, the stainless steel container was grounded by welded copper cables (Suppl.Fig. 3a, pdf).
In addition, the power supply was equipped with traditional protections against voltage anomalies (differential circuit breaker), and the datalogger and load cells were equipped with an uninterrupted power supply system (UPS device).

Wind speed
The main wind effect on the lysimeter measurements is momentum flux, causing important measurement fluctuations.In our installation, measurements were collected every second and averaged every 5 minutes to mitigate this perturbation.
Wind greatly affects the scale performance and accuracy (Ritchie & Burnett, 1968;Howell et al., 1995;Malone et al., 1999;Vaughan & Ayars, 2009).For crops as sorghum 15-min SD was consistent at about 0.42 mm for winds less than 5 m s -1 , but increased with wind speeds about 5 m s -1 (Howell et al., 1995).Similar conclusions were reached for a weighing lysimeter with Festuca by Martínez-Cob & Baselga (1999) which obtained SD of 0.13 mm with 30-min periods for wind speed lower than 3.5 m s -1 , but was as high as 0.80 mm for wind speeds above 6 m s -1 .In order to avoid these vibration problems, some authors used filtering computer programs to reduce the SD (Vaughan et al., 2007;Vaughan & Ayars, 2009), or electronic filters to reduce the amplitude of oscillations (Ritchie & Burnett, 1968), although repeated readings has been the most used procedure to minimize the wind effects (van Bavel & Myers, 1962;Tyagi et al., 2003;López-Urrea et al., 2006b).Thus, in order to avoid the perturbation in the measurements produced by the wind, López-Urrea et al. (2006b) suggested to increase the frequency of measurement, averaging them in time intervals between 10 and 30 min, and Dugas et al. (1985) proposed to average 5-s sample values for a 10-min period obtaining measurements independent of wind.

Measurement resolution
The main limitation of lysimeters without counterbalance mechanisms to achieve good resolution is the accuracy of the load cell.However, with the use of new devices that weigh directly the whole system (container, soil and tree), an additional limitation results from the resolution of the datalogger (Payero & Irmak, 2008).Thus, the resolution and precision of evapotranspiration measurements could be limited more by the datalogger and multiplexer combination than by the load cells (Allen & Fisher, 1990).In our case, following the manufacturer specifications and considering the load cell and datalogger characteristics, the measurement resolution was equal to 0.056 mm (0.5 kg), as described above.However, if the voltage signal provided by the load cells was directly used without any amplification of the signal, the maximum resolution would be around 1.5 kg and 0.17 mm.
Efforts to increase the resolution with current technology could only be obtained by changing from a load cell system to a counterbalance system, because load cells with better sensibility are not yet available.However, the manufacturer advised against this option considering maintenance reasons.In any case, wind effects could disguise this improvement, as shown long ago by van Bavel & Myers (1962), especially in the case of a tree.These authors indicated that due to wind effects, there is little practical improvement in accuracy with lysimeters that are more sensitive than 0.01-0.05mm.
A range of resolution between 0.02 mm and 0.05 mm is required to obtain hourly ET c rates (Dugas et al., 1985;Howell et al., 1985;Schneider et al., 1998;Girona et al., 2004;López-Urrea et al., 2006a).Malone et al. (2000) proposed a maximum relative error of 10% for hourly ET c estimations, implying that for the lysimeter reported here a threshold ET c of 0.56 mm hour -1 is required to obtain reliable hourly values.
Temperature effect on sensitivity according to the manufacturer is very small ( ± 0.008% of sensibility for each 10 K).However if the range of variation in temperature is very high there could be significant errors (Allen & Fisher, 1990).In our case, these errors are minimized because the load cells are installed in the control room below ground, where temperature fluctuations are attenuated.Kirkham et al. (1984) detected temperature changes of less than ± 0.2°C day -1 in the buried scales located at 2 m depth in two weighing lysimeters built in Wash-ington State (USA), and were similar to the values measured in our lysimeter.

Effect of condensation on the measurements
Condensation was detected during the nights on the soil contained in the lysimeter and on the internal walls of the container.Thus, an increase in weight after sunset was measured in numerous days.The environmental conditions in the control room are characterized by high humidity and a stable, moderate temperature.These conditions may affect the measurements due to the occurrence of condensation on the walls and base of the steel container during the evenings, when temperature of the container decreases.Although the amount of condensed water is low (~ 1 kg day -1 ), it may represent a significant error in days of very low ET c .To avoid this effect, a dehumidifier capable of removing 20 L of water each 24 h was installed in the control room.For daily measurements during summer time, ET c and T computations were carried out as the difference in weight from 19:00 GMT of the day before and the current, while for winter time the computed time was considered from 16:00 GMT.The measurement time (19:00 and 16:00 GMT) was chosen to obtain data previous to the beginning of the condensation process, considering that at that moment the lysimeter conditions are similar during successive days.

Wind speed effects on ET c measurements
Wind speed effects on the measurements were analysed by calculating the SD of the total weight for 10-min periods, a period that matched the measurement output rate of the weather station.The SD of the measurements during summer 2009 had an average value of 0.36 kg (0.04 mm), with an average wind speed of 2.12 m s -1 .The lowest SD values (around 0.32 kg) were observed during the night with an average wind speed of 1.3 m s -1 , although in the periods with absence of wind no changes in the ET c measurement were detected (CV = 0).Considering the summer daytime only (from 6:00 GMT to 19:00 GMT), the SD increased to 0.39 kg, caused by a faster wind speed (2.68 m s -1 ), and the measurement changes due to the simultaneous evapotranspiration losses.To quantify the effects of ET c variations on measurement variability, we calculated the SD of the measurement after removing wind speed effects.During daytime in the summer, this SD was equal to 0.01 kg, implying that during daytime more than 97% of the observed variability was caused by wind speed.
A correlation between wind speed and variability in the measurement was detected (Fig. 2).For the whole day, a weak correlation (R 2 = 0.37) was found.However, if only the night time data were analysed, this correlation increased to R 2 = 0.7 (Fig. 2).

Evapotranspiration of a young almond tree
Evapotranspiration (the sum of tree T and E from soil) was measured every 5-min period with high accuracy.Figure 3 presents the evolution of cumulative ET c for 12 days in June, 2009.The 5-min records setup allow to reduce losses of data for long periods (30-min or 1-hour periods) when lysimeter maintenance is carried out (pruning, weed control, etc.).
The periods of irrigation (fall in ET c due to the increase of weight) and periods of drying are clearly detected.Due to the small tree canopy (see Material and Methods section), ET c depended on the time since last irrigation event in the sense that ET c is mainly evaporation from soil.Thus, for example in the second week of June, daily ET c one day after the irrigation event was equal to 1.6 mm day -1 (14.5 L day -1 ), while two days later, ET c was reduced down to 1.3 mm day -1 (11.6 L day -1 ).The ET c values reported here reflect directly the weight loss of the lysimeter and need to be corrected to reflect the fact that tree spacing (6 m × 7 m) is greater than the surface area occupied by the lysimeter (3 m × 3 m).
In order to eliminate the evaporation from the bare soil, the soil of the lysimeter was covered with transparent plastic and fixed to the borders to avoid loss of water from the soil surface, procedure described by Ayars et al. (2003).To keep the soil surface conditions similar to the rest of the field, the plastic was covered with a soil layer of around 2 cm, obtained from the surrounding area of the lysimeter.
Figure 4 presents lysimeter measurements of T for three measurement frequencies, compared against the ET o daily pattern.Using 5-min averaged data showed some disruptions due to variations in wind speed affecting the cumulative T during some measurements (Fig. 4a).This was corrected when the data was averaged in 30-min periods, with only two periods having questionable values.With hourly data the curve was very stable (Fig. 4c).For the three scales, the T curve behaved similarly to the ET o curve, suggesting that the T determination with the lysimeter was valid.Declines in the T curve were caused by condensation processes previously described.
detected.An example of condensation effects on the measurements are shown in Figure 5 with data collected after 20:00 GMT in the evening.A detectable increase in weight (~ 1 kg) was measured close to sunset as the container is cooling (causing condensation), with a similar weight reduction some hours later.Analysing hourly results, the period between 12:00 GMT and 13:00 GMT was the period with higher rate of transpiration (for 9 th of July this value was equal to 1.6 L h -1 tree -1 ; Fig. 6).Lower values were detected for the rest of the day, with a similar curve than for the solar radiation and ET o .T rates in the early morning (around 7:00 GMT) could be caused by the evaporation of the condensation water accumulated during the night on the bare soil of the lysimeter.As stated above, accurate hourly T values may be obtained under high ET o rates.
Evaporation from bare soil is important to establish a water balance (Loos et al., 2007).Here, E from soil was measured during autumn/winter at a time when there was no transpiration from the almond tree.Figure 7 shows the evaporation rates for the period after an abundant rainfall (34 mm on 20-22 October period).Daily evaporation values ranged from 0.6 to 1.9 mm, with K c values declining from 0.65 to 0.38 in a 13-day period after rainfall.In spite of the low absolute E values, the lysimeter was able to detect the small variations in evaporation and K c .

Discussion
Lysimetry for determining the ET c of orchards is a viable method, although it has a number of complexities as compared to simpler estimation methods.For instance, the construction of a precise weighing lysimeter is complex and requires careful consideration of numerous aspects if an accurate and efficient instrument for ET c determination is sought.The location, construction procedures, weighing mechanisms and effects of external factors such as the wind speed, must be studied in advance, specially with tree crops where some adverse effects are enhanced, to avoid malfunctioning  Design and construction of a large weighing lysimeter in an almond orchard that would, lead to inaccuracies, data gaps and important over costs.Due to these concerns, the majority of weighing lysimeters have been used to grow annual crops (Wright, 1982;Tyagi et al., 2000Tyagi et al., , 2003;;Liu et al., 2002;Lovelli et al., 2005;Miranda et al., 2006;López-Urrea et al., 2009a,b), while growing trees or vines in them has been less common.Lysimeters with trees have been reported by McFarland et al. (1983), Girona et al. (2002Girona et al. ( , 2004)), Yang et al. (2003), Ayars et al. (2003) and Johnson et al. (2005), among others, but to our knowledge, almond trees have not been planted in large, weighing lysimeters.One of the limitations of tree lysimeters, is the practical impossibility of having a surface area in the lysimeter identical to the tree spacing outside.This is required to exactly reproduce in the lysimeter the evaporation from the soil surface outside the lysimeter.In this case (6 m × 7 m), it would have meant building a lysimeter that would be about five times heavier, with the equivalent loss in precision.Thus, the E component of ET c must be estimated by other means.The resolution obtained with the weighing lysimeter (0.056 mm) provided sufficient accuracy to measure precisely daily ET c values, and under high evapotranspiration demand, accurate hourly ET c values may be determined.Thus, compared with others lysimeters with similar technology (load cells without balances and/or counterweights), the described lysimeter provided very accurate resolution relative to other lysimeters (resolution ranged from 0.053 mm for pears and apples described by Girona et al. (2004) to 1 mm in peach described by McFarland et al. (1983)).Equally, compared with lysimeters with balances and counterweights, the resolution was also remarkable (resolution of these lysimeters ranged from 0.01 mm obtained by van Bavel & Myers (1962) and Hunsaker et al. (2002), to 0.064 mm obtained by Evett et al. (2009)).
The methodology for collecting data (averaged 5-min with a frequency of one second) plus the new load cell technology and installation reduced significantly the wind effects on the quality of the measurement, obtaining very satisfactory results in reducing the variability in the measurement (0.36 kg).These results were even more accurate than previous studies carried out on lysimeters with short-crops (Howell et al., 1995;Martínez-Cob & Baselga, 1999).
The infrastructure described here due to its specific characteristics could constitute the base of numerous studies related with water use, remote sensing and crop modelling for orchards.Thus, thanks to the field size and the accurate resolution obtained in the lysimeter, validation of remote sensing techniques (such as energy balance models) could be carried out.Equally, the development of water balance models quantifying separately E and T will contribute to an accurate determination of irrigation water requirements for orchards.

Figure 1 .
Figure 1.Sketch of the lysimeter showing the weighing system, control room, drainage system and access to the control room.A: transducer device, screen and datalogger.B: pressurized-gas highvoltage circuit breaker.C: stainless steel container.D: drainage deposits.E: load cells.F: steel ball supports.G: pump for irrigation from deposits.H: pump for control room drainage.

Figure 2 .
Figure 2. Correlation between wind speed and coefficient of variation of the measurement in 5-min periods for the whole data set (a), night-time (b) and daytime (c).

Figure 4 .
Figure 4. Cumulative transpiration (mm) for 9 July with three different averaging periods: a) 5-min, b) 30-min and c) 1-hour.Cumulative ET o is also shown.

Figure 5 .Figure 6 .
Figure 5. Accumulated transpiration averaged in 30-min periods and accumulated ETo for the period between 7 th and 11 th July 2009.

Figure 7 .
Figure 7. Daily ET c values measured with the weighing lysimeter, ET o and K c values for the period from 27 October to 8 November 2009.

Table 1 . Soil properties (density and texture) in the experimental field "Alameda del Obispo" close to the lysimeter Depth (cm) Density (g cm -3 ) Depth (cm) Sand (%) Silt (%) Clay (%)
Figure 3. Accumulated ET c and ET o averaged in 5-min periods for the period between 9 th and 22 nd of June 2009, including irrigation and drying periods.