Field measurement of groundwater recharge under irrigation in Canterbury, New Zealand, using drainage lysimeters

Agricultural Water Management 166 (2016) 17–32

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Field measurement of groundwater recharge under irrigation in Canterbury, New Zealand, using drainage lysimeters M.J. Duncan, M.S. Srinivasan ∗ , H. McMillan National Institute of Water and Atmospheric Research, P.O. Box 8602, Riccarton 8011, New Zealand

a r t i c l e

i n f o

Article history: Received 8 June 2015 Received in revised form 26 November 2015 Accepted 1 December 2015 Keywords: Actual evaporation Irrigated pasture Free draining shallow soil Deep loam soil Landscape position Centre pivot irrigators

a b s t r a c t Irrigation using groundwater in Canterbury, New Zealand, is reaching sustainable limits and to assist with water allocation a better understanding of groundwater recharge from irrigated agriculture is required. To help characterise groundwater recharge from irrigated pasture, three sets of three drainage lysimeters were installed in three irrigated dairy farms in Canterbury, New Zealand. Two farms have free draining, shallow, stony soils over gravel and the third site has a deep silt loam. The sites are spread across three landscape positions within the Canterbury Plains–foot-hill, mid plains and coastal plains. Average annual rainfall during the study period (2010–13) at the sites varied between 633 mm (coastal plain) and 891 mm (foothill). Irrigation management varied among the farms. Irrigation applications increased as actual evaporation increased and ranged from 144 to 445 mm/season (September–April). Drainage tended to increase with annual rainfall and most (70%) occurred in the winter (May–August). Drainage from the shallow stony soils and deep silt loams averaged 33 and 18% respectively of total precipitation (irrigation plus rainfall), a similar percentage to those reported from dryland lysimeters studies in this region. However, as the total precipitation on the irrigated sites is greater than rainfall in the dryland studies, irrigated agriculture had more drainage. This implies that irrigation of dryland will result in more recharge, but in much of Canterbury efficient centre pivot irrigators have replaced border dyke flood irrigation that has very high recharge rates, so there may be an overall reduction in recharge. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Groundwater reserves are important for many reasons: they provide vital municipal water supply, enable irrigation, sustain river base-flows and support aquatic habitats. However, in many parts of the world, an increasing use of groundwater threatens to reduce the stored water volume and, with it, the services this water provides (Zhang et al., 2003; Rodell et al., 2009). It is necessary to actively manage water use in order to sustain groundwater stores. To make accurate estimates of the consumptive use that an aquifer can sustain, the rate of recharge is a key variable that must be quantified (de Vries and Simmers, 2002; Scanlon et al., 2002, 2006). Groundwater recharge typically derives partly from the land surface and partly from rivers. Land surface recharge is controlled by precipitation depths and patterns, soil structure and antecedent conditions, which themselves depend on climate and vegetation (Bethune et al., 2008). Where land is used for agricul-

∗ Corresponding author. E-mail address: [email protected] (M.S. Srinivasan). http://dx.doi.org/10.1016/j.agwat.2015.12.002 0378-3774/© 2015 Elsevier B.V. All rights reserved.

ture, irrigation can dramatically change total precipitation depth (rainfall + irrigation) and soil wetness conditions, and therefore can potentially alter recharge. It is therefore important to quantify the effects of irrigation on recharge, to ensure that recharge estimates are valid in irrigated landscapes, and establish whether recharge observations from dryland (non-irrigated) landscapes are applicable. A better understanding of groundwater recharge under irrigation can also be used to inform farm management practice. In New Zealand, intensive dairy farming leads to recharge associated nutrient leaching and subsequent contamination of groundwater and eutrophication of waterways (Di and Cameron, 2002; Monaghan et al., 2014). Where recharge is well understood, irrigation management can be optimized to ameliorate water stress on crops or pasture without causing recharge and leaching (Bryla et al., 2010; Martin et al., 1994). Use of groundwater for irrigation in Canterbury, New Zealand is approaching sustainable limits due to a relatively dry climate combined with recent expansion of irrigated dairy farms. This expansion drove the irrigated area in the region up to 6800 km2 or an increase of 65% in the period between 1990 and 2010, with 42% of allocated water sourced from groundwater (Rajanayaka et al.,

18

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 1. Study locations.

2010). There is currently high uncertainty in recharge estimates for the region due to a bias in studies toward unirrigated land uses. There is a need for more measurements under irrigation to determine the effects on recharge given local climate and soil characteristics. The shallow, stony soils of the region (described in more detail below) are typically fast draining, implying that recharge from the land surface could be an important source for the underlying groundwater. Improved knowledge could potentially allow more efficient allocation of water, reducing water use and enhancing the economic benefits of irrigation. This study presents initial results of recharge from irrigated dairy farms on the Canterbury plains, in eastern South Island New Zealand (Fig. 1). The “Canterbury Lysimeter Network” currently includes lysimeter installations at four sites, each of which includes three drainage lysimeters and a climate station. Three sites are located on irrigated dairy pasture. A further site is located on a cropping farm and is not included here. The aims of the network are to provide measurements of recharge under irrigation in Canterbury, and to allow comparison with previous dryland recharge measurements in the region made using lysimeters similar to those

in this study. In this paper we investigate sources of variability in the rate of groundwater recharge, including between-site variability due to soil type, climate, and irrigation management practices, and within-site variability due to small-scale soil and vegetation heterogeneity. Our long-term aim is to improve region-wide estimates of recharge, and inform irrigation management practice.

1.1. Lysimeter design Measurement of recharge through soils must account for the dual pathways of matrix flow and macropore flow (Greve et al., 2010; Schoen et al., 1999). Instruments should therefore be designed to cover sufficient area to include a representative sample of macropores resulting from plant roots, earthworm or other animal activity, or by soil cracking. The most common method is through the use of a lysimeter; a controlled volume of soil through which water movement can be measured (Goss and Ehlers, 2009). Two types of lysimeter are commonly used: “weighing lysimeters” (e.g., Campbell, 1989; Evett et al., 2009; Marek et al., 2006; Yang et al., 2000; Young et al., 1996) which continuously weigh

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

19

Fig. 2. Schematic layout of a lysimeter site.

Fig. 3. Daily rainfall, irrigation, soil water content and drainage at the Methven site. Totals for the period of record are inset. For soil water content, the distance of the sensors below the soil surface is indicated.

a known soil volume to determine moisture content and evapotranspiration, and “drainage lysimeters” (e.g., Arauzo et al., 2010; Bethune et al., 2008; Gee et al., 2003; Kim et al., 2011) in which water draining out of the soil column is collected and measured. The different primary purposes have resulted in quite different

designs. Lysimeters to measure evapotranspiration are usually very large (several meters in diameter or along the side of rectangular lysimeters and as much as 2.5 m deep), and are set on sensitive load cells that enable small changes in weight to be measured and allow changes evaporative losses of the order of 0.1 mm to be

20

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 4. Daily rainfall, irrigation, soil water content and drainage at the West Eyreton site. Totals for the period of record are inset. For soil water content, the distance of the sensors below the soil surface is indicated.

recorded (Campbell, 1989; Evett et al., 2009). They can also measure drainage, but that is not their main purpose. There is often difficulty in obtaining undisturbed soil cores for such lysimeters because of the difficulty in lifting such large and heavy soil samples and their containers. Consequently many contain reconstituted soil (Evett et al., 2009). Because of their size and the need to provide infrastructure such as a pit to contain the lysimeter and instrumentation, they are expensive. Drainage lysimeters are commonly less than 1 m in diameter and less than 2 m deep, and are more likely to contain undisturbed soils. Their smaller size enables replication at a site to measure variability. With permanent vegetation, the lysimeter rim can protrude slightly above the soil surface preventing surface water from entering or leaving the lysimeter, thereby ensuring the accuracy of the water balance. Drainage can either be directed to a nearby pit containing tipping bucket mechanisms or containers for measuring drainage or have a mechanism underneath the lysimeter to measure and/or collect the drainage such as the Gee passive capillary lysimeter (Arauzo et al., 2010; Gee et al., 2003; Kim et al., 2011). Recharge over wide areas is often estimated using indirect methods such as recharge models (Kendy et al., 2003) driven by soil moisture estimates from remote sensing (e.g., satellite microwave sensing; (Li et al., 2008)) or soil water balance models (Barringer et al., 1995). However, these approaches typically require ground truth data from direct measurements of both soil moisture and recharge. No special adaptation is required for passive drainage lysimeters in irrigated fields. However, in the case of sprinkler irrigation, steps need to be taken to ensure that the rain/irrigation gauge and the lysimeters all receive the same part of the spray pattern as it is

spatially variable. Data from a nearby unirrigated rain gauge can help distinguish rainfall from irrigation in the combined gauge. 1.2. Previous rainfall recharge studies in Canterbury Groundwater recharge from rainfall has previously been measured with lysimeters at four sites in mid-Canterbury (Christchurch Airport, Lincoln, Winchmore and Hororata, Fig. 1). At Winchmore there were six lysimeters made from concrete pipes 1220 mm in diameter and 900 mm deep surrounding an undisturbed soil monolith. The other sites had pairs of lysimeters (undisturbed soil cores) very similar in design to those used in this study and described below. Lysimeter output from the Winchmore lysimeters was initially (1953–1978) collected in large calibrated tanks and nominal daily values read manually from attached sight glasses. For 1994–1997 lysimeter output was measured with 0.5 mm tipping bucket rain gauges (Thorpe and Scott, 1999). The Winchmore rain gauge, in common with most rain gauges with orifices 40 cm above ground level, has been found to under-catch by about 7% (White et al., 2003) to 10% (Thorpe and Scott, 1999) compared with rain gauges with orifices at ground level. The other sites had ground level tipping bucket rain gauges that measured rainfall at 5 min intervals (White et al., 2003). At Winchmore, based on data from 1955/56–1959/60 and 1995–1997, 37% of precipitation on grassed lysimeters became recharge (Thorpe and Scott, 1999). For 1995–1997, the three lysimeters at Winchmore had recharge of 25% of rainfall for a year when there was no irrigation and recharge of 46% of rainfall plus irrigation over two years when they were irrigated (Thorpe and Scott, 1999). There were five spray irrigations of 102–108 mm each during the two years.

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

21

Fig. 5. Daily rainfall, irrigation, soil water content and drainage at the Dorie site. Totals for the period of record are inset. For soil water content, the distance of the sensors below the soil surface is indicated.

From May 1999 to April 2003, for Christchurch Airport, Winchmore and Hororata where the soils are shallow and stony (Lismore or Lismore-like soils) average annual rainfall was 762 mm and drainage 261 mm (34%). At Lincoln, where the soil is a deep silt loam (Templeton), average annual rainfall was 676 mm and drainage 96 mm (14%). Rainfall recharge is commonly zero or low in summer (White et al., 2003). 2. Methods 2.1. Site description—irrigated lysimeter sites The Canterbury Plains (Fig. 1) are formed from a series of large, coalescing Pleistocene glacial outwash fans deposited by rivers that have catchments in the mountains to the west. The alluvial fans consist of greywacke gravels and cobbles in a highly variable matrix of sand and silt (White et al., 2003). A thin veneer of loess soil covers much of the fan surface (White et al., 2003). Most of the soils on the Canterbury Plains are shallow and stony. Kear et al. (1967) describe the soils covering about 60% of the Plains as “somewhat excessively “ or excessively drained, so recharge from rainfall and irrigation to groundwater occurs quite readily in most areas. Lismore soils are the most widespread, covering about 25% of the Plains. The study sites were selected to cover a range of climate, soils and irrigation management regimes typical of dairy farms in Canterbury (Fig. 1). Two of the sites (Methven and West Eyreton) are on shallow (300–400 mm deep) stony (Lismore like) soils overlying coarse sandy gravels, typical of new dairy farms in Canterbury, with a profile available water (PAW) of less than 115 mm (Table 1). The Dorie site has Templeton silt loam soil which has over 700 mm of

soil and subsoil above sand over a sandy gravel, typical of the heavier and imperfectly drained soils on the eastern edge of the Plains. Dorie has low rainfall and mid-range potential evapotranspiration (Table 1). The ground cover at all sites is ryegrass/white clover dominated pasture that is grazed about every three weeks during spring, summer and autumn and averaging four weeks during winter. The maximum ground-water levels at Methven and Dorie are 39 m and 6 m below the surface respectively so the water table is too low for to have any capillary rise effect on those lysimeters. At West Eyreton the water levels during winter can rise to within ∼1.2 m of the ground surface, but because the subsoil is poorly compacted sand and gravel, capillary rise is unlikely to have any effect on the lysimeters there. So for all sites, the lysimeters represent soil water movement correctly with respect to capillary rise. 2.2. Lysimeters and instrumentation At each site, three cylindrical, non-weighing, drainage lysimeters 0.5 m in diameter and 0.7 m in depth were installed. Three lysimeters were installed at each site to measure drainage variability. The major components of the lysimeters are hot-dip galvanised mild steel to ensure longevity. The dimensions of the lysimeters were matched to those of previous lysimeter studies within the region (White et al., 2003) to enable a direct comparison. Fig. 2 shows a schematic layout of a lysimeter site. The lysimeters are “filled” with an undisturbed column of soil by easing the lysimeter cylinder down the soil column while excavating around it. When the soil column is completely enclosed by the lysimeter, it is cut off using a hydraulically-operated blade and enclosed with a base plate. Any gaps between the soil column and

22

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 6. The range of monthly drainage for each site from three lysimeters for 3 seasons. Note the shorter time periods shown for West Eyreton.

the lysimeter barrel were filled with petrolatum to prevent edge flow. The lysimeters are installed so either the lysimeter lip protrudes above the surrounding soil by 1–2 cm or the lysimeter is fitted with a rubber cuff that protrudes above the soil surface. This ensures that there is neither surface runoff onto nor from the lysimeter, ensuring the accuracy of the water balance. Drainage reaching the base of each lysimeter is guided by grooves in the lysimeter base to the centre from where it is piped to a tipping bucket gauge (0.2 mm cup size) that measures the drainage. Each tip of the bucket represents 0.032 mm of drainage with respect to the diameter of the lysimeter. Where there is a predominantly gravel horizon at the bottom of the lysimeter (Methven and West Eyreton), any gaps between the gravel and the lysimeter

bottom plate are filled with sand. For fine textured sub soils (Dorie) a fibre-glass wick is spread over the bottom plate and extends below the lysimeters to provide a suction force to the bottom of the soil profile. This allows soil water to be drawn out of the profile, to more closely mimic the natural condition of a continuing soil column. If wicks were not installed, water would sit at the bottom of the profile until it was saturated, delaying the drainage. To ensure that the lysimeters, the tipping bucket rain gauge and check gauge that measure irrigation and rainfall receive the same input, the lysimeters are aligned so that they are irrigated by the same spray nozzles of the centre pivot irrigators. The rain gauge orifices were installed at ground level to measure true rainfall and irrigation and were surrounded by a ground level grid to prevent splash. All the study sites have centre pivot irrigators, where the

Table 1 Soil type, texture, profile available water (PAW), mean annual rainfall, mean annual Penman potential evaporationa , and elevation for the lysimeter sites. Site

Soil type

Soil texture

PAW** (mm)

Rainfall (mm/y)

Penman PET (mm/y)

Elevation (m amsl)

Methven West Eyreton Dorie

Hororata Eyre-Paparua Templeton

Shallow stony soil over gravel Shallow stony soil over gravel Deep silt loam

111 108 148

915 751 661

731 917 851

308 89 54

a

Rainfall and potential evaporation from NIWA’s virtual climate network (Tait and Woods, 2006).

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 7. The effect of summer rainfalls on soil water content levels and drainage at Methven.

Fig. 8. A winter rain storm and resulting drainage from each lysimeter at Methven.

23

24

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 9. A spring rain storm and resulting drainage from each lysimeter at Methven.

lysimeters are positioned midway between gantries and about midway along the irrigator span, to avoid the portion near the pivot point where over-irrigation can occur and the outer spans where high instantaneous application rates can result in surface runoff and ponding. All tipping buckets are field-calibrated annually. An issue at one site (West Eyreton) was the appearance of groundwater in the pit to a depth sufficient to drown and disrupt the tipping-bucket gauges for 76 days. The problem was eliminated by enclosing the tipping-bucket gauges in a plastic box and replacing the timber lined pit with a prefabricated concrete box. A daily water balance model, calibrated against the record, was used to estimate drainage for the 76-day period when the pit was flooded. There is a climate station at each site located in a non-irrigated area within 0.9 km of the lysimeters. The parameters measured are wind speed, wind direction, air temperature, relative humidity, solar radiation, rainfall (ground level orifice), and soil water content and soil temperature at 20 cm depth. Data are averaged and logged every 10 min. The climate station rainfall is used to help partition the precipitation at the lysimeter between rainfall and irrigation. Volumetric soil water content is measured periodically using a TroxlerTM neutron probe at 150, 350, 450 and 550 mm depths from the soil surface via access tubes inserted into the centre of each lysimeter (Fig. 2). Measurements are carried out approximately weekly during the irrigation season (September–April) and less frequently (once in three weeks to monthly) at other times (May–August). The neutron probe instruments are recalibrated annually. Additional continuous volumetric soil water content, measured using AcclimaTM electronic sensors (http://acclima.com), is averaged over 0.45 s and logged every 10 min. Acclima sensors use the time domain travel principle. For the Dorie site, with fine textured soils one lysimeter in each set of three is fitted with sensors located

at the mid-point of soil horizons at 95, 300 and 540 mm below the soil surface. At Methven and West Eyreton with stony soils, all three soil water content sensors were placed between 100 and 200 mm below the soil surface outside of, but in the vicinity of, the lysimeters. The sensors were not placed deeper or within the lysimeters because of the difficulty in ensuring close soil contact with the sensor elements in stony soils, especially when installed within lysimeters. Drainage, soil water content, rainfall, climate parameters and irrigation are registered every 10 min on a data logger, and every hour the data are telemetered to a central database. The data are immediately available on line and via e-mail to researchers and farmers. Every day data from all the sensors are automatically plotted and checked for errors and continuity. 2.3. Irrigation management Irrigation management influences the amount of drainage. Irrigation scheduling at Methven and Dorie is based on soil water content measurement from the lysimeters and weekly neutron probe soil water content measurements supplied by a commercial service, with irrigations postponed at Methven if rainfall is forecast. Irrigation water supply at Methven is from a piped community scheme, and at Dorie from groundwater. Both sources are very reliable. The centre pivot irrigators apply small amounts (4–10 mm/application) every one to four days to provide enough soil water for a good pasture growth, but still allowing sufficient soil water storage to absorb most rainfalls, thus reducing the potential for drainage. At West Eyreton, the run-of-river irrigation supply is less reliable than at the other two sites, so soil water is kept topped up with frequent irrigations to guard against reductions in supply. When river flows fall below a pre-set threshold, the irrigation

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

25

Fig. 10. A summer rain storm resulting in no lysimeter drainage at Methven.

supplies may be fully or partially restricted. Thus, more water was applied than there would have been with a reliable water supply and thus there is less soil water storage available to absorb rainfall. 3. Data analysis Irrigation was determined by the difference in precipitation within and outside the central pivot irrigator. Drainage and neutron probe soil water content are averaged from the three lysimeters at each site. Actual evapotranspiration is the residual when drainage and change of soil moisture storage is deducted from precipitation (rainfall + irrigation). The changes in soil moisture between the start and finish of the irrigation season were relatively small and average absolute values were 10.6 mm, 9.8 mm and 20.8 mm for Methven, West Eyreton and Dorie, respectively. Pasture transpiration is unlikely to be constrained at soil water contents greater than 20% (140 mm in the 700 mm deep lysimeters) for these soils. The water balance was calculated for irrigation and nonirrigation seasons and for whole years. Differences in drainage between and within sites were explored. The effects of irrigation management and seasonal evapotranspiration on drainage were examined. The relationship between annual rainfall and drainage were also explored. 4. Results 4.1. Drainage Drainage for the two year common period (2011–2013) was 815, 726 and 350 mm for Methven, West Eyreton and Dorie, respectively. The drainage at Methven and West Eyreton were similar

though Methven received more rainfall and less irrigation than West Eyreton. The lower drainage at Dorie is attributed to its lower rainfall and higher soil PAW. The average drainage at Methven, West Eyreton and Dorie for complete years was 32, 34 and 18% of total precipitation, respectively (rainfall plus irrigation, Table 2). For 2011–2012, total precipitation, evaporation and drainage were highest at Methven, even though there was careful irrigation management, as soil water levels remained high because of high rainfall. The drainage fraction of total precipitation was similar to that of the West Eyreton site that has similar shallow stony soils and where more irrigation occurs because of the unreliability of the irrigation water supply. The Dorie site with soils with much higher PAW (Table 1) and conservative irrigation practice had much less drainage, especially during the irrigation season At each site, the relatively even temporal distribution of rainfall through the year contrasts with the seasonal irrigation distribution (Figs. 3–5). While the soil water contents vary, the base levels are relatively high throughout the year as would be expected for irrigated land. Drainage at the Methven and Dorie sites occurs predominantly during the winter. Unlike other two sites, at West Eyreton drainage occurs both in winter and during the irrigation season. The temporal distribution of drainage at West Eyreton is different from the other sites as the irrigation practice there is to keep soil water content levels high in case irrigation water is unavailable from the run-of-river supply, which has historically been unreliable. The Dorie site (Fig. 5) shows very little drainage during irrigation season even though there are spikes of high soil water levels in the lowest sensor (@ 54 cm below surface), whereas in winter there is drainage when the spikes are smaller, however the base-

26

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

Fig. 11. Rainfall, irrigation, drainage and soil water status for the 2011–2012 irrigation season. (A) Methven, (B) West Eyreton, (C) Dorie.

line levels of soil water are lower in the summer than the winter. An explanation of the behaviour of the soil water levels at 54 cm is that in summer when the soil dries out a little, the subsoil cracks a little creating macro pores that fill with water after irrigation or rainfall, whereas in winter the soil is wetter and the soils swell and reduce pore sizes. There is little drainage in summer as the soil is dry enough to absorb the macro pore water where as in winter the total soil body becomes wet enough to allow drainage.

4.2. Drainage variability within and between sites At Methven, for three years from 1 September 2010 to 31 August 2013, the three lysimeters each had 1206 mm, 776 mm and 1063 mm drainage, averaging 1015 mm. The range of absolute monthly drainage for each lysimeter set and each year are shown in Fig. 6. While the relative response to rain of each lysimeter varied, it was generally of the same pattern as for the period totals. For

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

27

Table 2 Water balance for three lysimeter sites for three years. Site Methven

West Eyreton

Dorie

Irrigation season Non-irrigation season

2010–2011

Rain (mm) Irrigation (mm) Actual evaporation (mm)a Penman PET (mm) Change in soil water content (mm)b Drainage (mm) Drainage/(rain + irrigation) (%)

448 144 538 757 −12 66 11

2011–2012 2011

Rain (mm) Irrigation (mm) Actual evaporation (mm) Penman PET (mm) Change in soil water content (mm) Drainage (mm) Drainage/(rain + irrigation) (%)

859

Rain (mm) Irrigation (mm) Actual evaporation (mm) Penman PET (mm) Change in soil water content (mm) Drainage (mm) Drainage/(rain + irrigation) (%)

367 303 658 777 1 11 2

207 67 101 6 134 65

118

220 127 98 -4 97 44

2012–2013 2012

678 173 658 700 −7 200 24

310

456 225 514 755 1 166 24.4

299

442 150 631 691 -55 16 3

218

134 84 −21 197 64

86 103 55 158 52.8

90 91 35 93 43

2013 618 168 584 772 7 195 25

411

346 445 557 846 26 208 26.3

344

278 253 534 765 -9 6 1

373

185 101 3 223 54

125 104 25 194c 56.4

134 90 4 235 63

Average complete Years 891 162 722 838 338 32 723 335 641 904 363 34 633 235 725 837 153 18

a

Actual evaporation was calculated from the lysimeter water balance. Change in soil water content is the difference between soil water content on the first and last days of the season. Record not complete because the pit was flooded. Daily water balance modelling indicates that drainage was unlikely during this time as during the 76-day flooding period, potential evapotranspiration of 186 mm exceeded rainfall of 140 mm. Pit flooding occurs because of a seasonal increase in the water table level and is not directly related to the onsite water balance. b c

isolated storms lasting two days with daily rainfall of the order of 25 mm or more, most of the drainage occurred during the day of the rain or the day after. Drainage usually ceased by the third day following such storms. Rain or irrigation occurred on 47% of days and drainage occurred on 45% of days. Daily drainage ranged from 0 to 47.4 mm and averaged 2.2 mm per drainage day. At West Eyreton from 1 September 2011 to 28 August 2013, the three lysimeters had 769, 805 and 702 mm drainage, averaging 725 mm. The range of drainage for each lysimeter set for each month is shown in Fig. 6. Again, while the relative response to rain of each lysimeter varied, it was generally of the same pattern as for the period totals. For isolated storms lasting two days with daily rainfalls of the order of 25 mm or more, most of the drainage occurred during the day of the rain or the day after, but in contrast to the Methven site, small amounts of drainage continued for a long time, so that there was some drainage during 76% of days. Drainage ranged from 0 to 33.2 mm/day and averaged 1.3 mm per drainage day. The greater number of drainage days is attributed to the excess irrigation. Rain or irrigation occurred on 44% of days. At Dorie, from 1 September 2010 to 31 August 2013, the three lysimeters had 508, 479 and 382 mm drainage to average at 456 mm. The range of drainage for each lysimeter set for each month is shown in Fig. 6. Again, while the relative response to individual rain events of each lysimeter varied, it was generally of the same pattern as for the period totals. However, the lysimeter with the largest total would sometimes drain when the others did not. Storms lasting two days with daily rainfalls in excess of 45 mm where required to initiate drainage in all three lysimeters. While most of the drainage occurred during the day of the rain or the day after, the amounts were much smaller than for the other two sites and drainage continued for 2–3 weeks after the storm. Drainage occurred during 33% of days. Drainage ranged from 0 to 41.0 mm/day and averaged 1.3 mm per drainage day. The difference in drainage behaviour between this site and the other two is attributed to the higher PAW of the deep silt loam soils. Rain or irrigation occurred on 57% of days.

Fig. 6 shows the drainage range from the lysimeters for each month and site. Overall there is more drainage in winter when the soil is wet and evapotranspiration is less than rainfall, but there are considerable differences in drainage for the same month in different years as the lysimeters respond to the month’s rainfall. In June 2012 there was heavy widespread rainfall throughout Canterbury falling on soils that were saturated, as indicated by the drainage for May 2012, and all sites had similar substantial amounts of drainage. There are differences in monthly drainage between sites that reflects both the rainfall regime and soil texture. Methven tends to have more drainage than the other two sites because of its higher rainfall. Dorie with its high PAW silt loam soils has much less drainage during all seasons than the other sites with their shallow stony soils. Methven has a greater range of drainage from its lysimeters than the other sites and this is attributed to the variability in soil texture and stone content between the lysimeters. The other sites have less soil variability between lysimeters and less variability in monthly drainage. While the lysimeters at each site were subjected to the same evapotranspiration demand and received almost the same rainfall and irrigation, the drainage from the individual lysimeters varied, and is attributed to differences among the lysimeters in the amount of grass cover and soil heterogeneities resulting from stone content, and amount and connectivity of soil pores, especially macropores.

4.3. Seasonal drainage variability The Methven and Dorie sites have been active for three years and provide the opportunity to examine seasonal variability of drainage during that period. During three summers (September–April) at Methven, drainage averaged 20% of rainfall plus irrigation (range 11–25%) whereas over three winters (May–August) drainage averaged 61% of rain (range 54–65%). From the deep silt loam at Dorie, drainage averaged 2% of rain plus irrigation (range 1–3%) over 3 summers and over the winters 50% of rain (range 44–63%). The higher rainfall and shallower stony soils at Methven resulted in

28

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

more drainage during both seasons than at Dorie. At both sites, winter drainage as a proportion of rainfall shows a low variability, indicating that once the soil water store is full any additional rain results in drainage. At Methven in particular, summer drainage varied widely and is probably reflective of the frequency of large rainfall events and their coincidence with high soil water levels due to irrigation prior to such rainfall events. Fig. 7 shows regular irrigations at Methven kept the soil water content at just below 40% by volume with no drainage. Two days after the last irrigation there was rainfall of over 20 mm/d for two days, which resulted in elevated soil water values and drainage. Even if soil water levels were kept lower and the irrigation on 15 March cancelled, the 52 mm of rainfall during 17–19 March would have exceeded field capacity resulting in drainage. Examination of drainage behaviour helps explain the differences in drainage between seasons. Figs. 8–10, show storm rainfall and lysimeter drainage at the Methven site for three seasons. Fig. 8 shows a winter storm with similar drainage from each lysimeter and an average drainage of 89% of rainfall. The initial PAW was estimated at approximately 160 mm, based on weekly neutron probe measurements and electronic soil moisture sensors. Increases in drainage commenced soon after the rain started. Fig. 9 shows a spring storm that shows some variability in the amount of drainage between the lysimeters and, even though it is a relatively large storm (55 mm), the average drainage was less at 55% of rainfall. The initial PAW was approximately 132 mm before the rain. Fig. 10 shows a summer storm of 36 mm with no drainage recorded. The initial PAW was approximately 136 mm. In general, annual drainage increased with total annual precipitation (rainfall plus irrigation, Fig. 6). There is a suggestion that this relationship may not hold for the deeper silt loam soils at Dorie, but for the two years with low amounts of drainage it is possible that some precipitation falling on the lysimeter was not contained within the lysimeter as the lysimeters initially did not have a lip preventing runoff.

the higher total precipitation there. The other two sites had similar rainfall amounts but there was much more drainage at West Eyreton resulting in less AET there. Irrigation applied was similar at Methven and Dorie with 341 and 403 mm, respectively. The slightly higher irrigation application at Dorie is partly explained by the much lower irrigation-season rainfall there. However, the difference in irrigation applications is much less than the difference in evaporation between the sites and probably also reflective of larger PAW (Table 1) at Dorie, which allows more soil water storage at the start of the irrigation season, and provides the potential to store more rainfall and have less drainage during the irrigation season, thus lowering the overall irrigation requirement. 4.5. Irrigation There was more irrigation (335 mm/year (Table 3)) at West Eyreton than at Methven (162 mm/year) and Dorie (253 mm/year), even though there was more rainfall at West Eyreton during the irrigation season than at Dorie (Table 1). Table 3 shows that the number of irrigations at each site for each season are roughly similar, but the application rates at West Eyreton were much higher than at the other two sites. The average application rate at West Eyreton of 8.4 mm/irrigation is over twice the average Penman PET daily average rate for November to February and the summer drainage shows that was more irrigation than required. Irrigation at West Eyreton also spans a longer period than at the other sites with Methven having a significantly shorter season (Table 3). The lower PAW (Table 1) at West Eyreton that may have necessitated a longer irrigation season, but if the application rate was lower, summer drainage may have been reduced. Also, Dorie had more frequent, smaller irrigation applications than at West Eyreton, thus reducing the likelihood of drainage. The lack of increases in soil moisture at deeper depths (e.g., 540 mm) despite irrigations (see Fig. 5 soil moisture and irrigation application depth data) shows that evapotranspiration and irrigation were balanced and well managed.

4.4. Evapotranspiration 4.6. Effectsof irrigation management on drainage Leaf area index (LAI) and the grazing regime have the potential to affect actual evapotranspiration and recharge. At Methven the LAI ranges from 1 to 3–5 in summer and 1 to 2–3 in winter with grazing every 3 weeks over most of the summer (October–April) and averaging 4 weeks in winter (Graham et al., 2015). As all the sites are high producing dairy farms with similar ryegrass/white clover pastures the other sites will have similar LAI’s and grazing intervals. AET from the pasture at Methven, measured using eddy covariance, is very similar to that from a ryegrass seed crop when that crop is actively growing, i.e., before drying off for harvest. This indicates that the LAI does not seem to have a material effect on AET in this instance (Graham et al., 2015). This effect has been attributed to evaporative losses from the soil surface when the grass is short because the soil surface is almost always damp because irrigation and rainfall keep soil water content levels high. Furthermore, AET from the eddy covariance tower at Methven is similar to Penman PET which also indicates that grazing in this case is not materially affecting AET (Graham et al., 2005). In the context of the Canterbury Plains, the sites have different rainfall and evaporation regimes, and there are differences in soil PAW between sites. For the two year period of common record for all sites (September 2011–August 2013), the Methven, West Eyreton and Dorie sites had rainfalls of 2017, 1445 and 1311 mm, respectively, and irrigation season rainfalls were 1296, 802 and 720 mm, respectively. Actual evapotranspiration (AET) was calculated from the lysimeter water balance. At Methven, the 2-year AET was 1561 mm, whereas it was 1282 and 1389 mm at West Eyreton and Dorie, respectively. The higher AET at Methven is attributed to

Fig. 11A–C shows rainfall, irrigation, drainage and soil water status for the 2011–2012 irrigation season for each site. Drainage at Methven (Fig. 11A) with its higher rainfall occurred mainly after large rainfalls. Irrigation occurred well after large rainfalls and only when soil contents were falling, except between 23 January and 8 February 2012. During this period, after a 29.6 mm of rain on 22 January, there was 19 mm of rain and 88 mm of irrigation, even though soil water levels were high. During this time there was 25.4 mm of potentially avoidable drainage. West Eyreton (Fig. 11B) shows a similar pattern, but there are some occasions when rain and irrigation coincided or irrigation occurred soon after rain and this resulted in drainage. Dorie (Fig. 11C) shows careful irrigation management with almost no irrigation after large (<15 mm) rainfalls, and only sufficient irrigation to keep soil water levels above the critical 20% irrigation trigger level (soil water content below which the crop may come under stress and become less productive) These examples show that active irrigation management by postponing irrigation after large rainfalls can minimise drainage while maintaining soil water above stress levels. However, if soil water levels are not constantly monitored then over irrigation can occur resulting in avoidable drainage. More generally, more irrigation might be expected in dry summers, but that is not supported by the data as is shown in the relationship between rainfall and irrigation during the irrigation season, categorised by site (Fig. 12). Curiously at Dorie there is a suggestion of a positive relationship between irrigation season rainfall and irrigation application whereas the opposite would be expected.

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

29

Table 3 Seasonal irrigation totals, number of irrigations and average application rates for three lysimeter sites for three years. Site

Irrigation season

2010–2011

2011–2012

2012–2013

Average

Methven

Irrigation total (mm) No. of irrigations Average application) rate (mm/application Start date Finish date

144 62 2.3 28/09/2010 4/02/2011

173 31 5.6 28/11/2011 12/02/2012

168 42 4.0 2/11/2012 26/03/2013

162 45 3.6

West Eyreton

Irrigation total (mm) No. of irrigations Average application rate (mm/application Start date Finish date

225 34 6.6 5/10/2011 2/04/2012

445 46 9.7 26/09/2012 3/04/2013

335 40 8.4

Dorie

Irrigation total (mm) No. of irrigations Average application rate (mm/application Start date Finish date

150 40 3.8 2/11/2011 6/04/2012

253 56 4.5 30/09/2012 5/04/2013

235 53 4.4

303 63 4.8 14/09/2010 15/03/2011

Fig. 12. The relationship between irrigation season rainfall and irrigation application.

Fig. 13. The relationship between irrigation season actual evapotranspiration and irrigation application. The markers show the average actual evaporation from each lysimeter set and whiskers the range of actual evaporation from the individual lysimeters in the set.

At West Eyreton there is suggestion of a negative relationship, but that site received more irrigation than at the other sites for similar seasonal rainfall amounts. At Methven, there appears to be no relationship between irrigation season rainfall and irrigation. However, Fig. 13 shows that the irrigation was applied in response to the measured evapotranspiration. Fig. 13 shows the relationship between irrigation season evapotranspiration calculated from lysimeter water balance and

Fig. 14. Irrigation season actual evapotranspiration vs Penman PET. The markers show the average actual evaporation from each lysimeter set and whiskers the range of actual evaporation from the individual lysimeters in the set.

irrigation application. During the irrigation season, the soils receive sufficient rainfall and irrigation to ensure that the pasture is not water stressed, and so it is assumed that the pasture is transpiring at potential rates. Fig. 3 shows the soil water content is between 25 and 45% (proportion of the soil volume that is water (v/v)) which is above stress levels. Overall, the evidence is weak that more irrigation is applied as measured evaporation increases. Only at the Methven site, where irrigation scheduling is based on measured soil moisture did the amount of irrigation increase monotonically with increased evapotranspiration calculated from the lysimeter water balance. At West Eyreton, it appears that more irrigation was applied than necessary as described earlier. Fig. 14 shows irrigation season evapotranspiration obtained from lysimeter water balances vs Penman (1963) PET obtained from the National Institute of Water and Atmospheric Research’s (NIWA) Virtual Climate network (Tait and Woods, 2006). It shows a large variability in the actual evapotranspiration relative to the Penman (1963) PET and the overall and individual site relationships are poor. For most irrigation seasons Penman (1963) PET is greater than AET. This is understandable as Penman (1963) calculates open water surface evaporation rather than evapotranspiration from pasture. Furthermore, the lysimeters are grazed periodically, immediately after which the leaf area index will be low and the pasture may be transpiring at lower rates than those from the Penman (1963) calculations. Another consideration is whether the soil is moist enough to allow the pasture to transpire at potential rates. Farmers at all sites employ a consultant to measure soil moisture and to recommend irrigation schedules that are designed to keep

30

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

soil moisture above stress levels (∼25% v/v) (Figs. 3–5). Thus it is valid to assume that soil water contents were kept high enough to allow transpiration at potential rates.

ances for each site come from using the combined data from the site. 5.3. Evapotranspiration

5. Discussion 5.1. Drainage This paper has shown that average drainage of 153 mm/year from an irrigated deep silt loam soil was 18% of average annual precipitation (rainfall + irrigation). Average drainage of 331 mm/year from two irrigated sites with shallow stony soils was 33% of average annual precipitation. Previous research (Thorpe and Scott, 1999; White et al., 2003) found that drainage from an unirrigated deep silt loam of 96 mm/year was 14% of annual average precipitation, while drainage of 205–298 mm from unirrigated, shallow stony soils in Canterbury was 32–36% of annual precipitation (639–841 mm/year). Recharge from flood irrigated lysimeters over two years at a site with shallow stony soils, was 46% of rainfall plus irrigation (Thorpe and Scott, 1999). For the shallow stony soils, the drainage as a percentage of total precipitation under irrigation in this study was similar to the 32–36% of precipitation from the dryland lysimeters (White et al., 2003). However, for irrigated shallow stony soils, the drainage as a percentage of precipitation was much less than found by Thorpe and Scott (1999), possibly because the flood irrigations at Winchmore were large at 102–108 mm each and some occurred in May and August when soil water content might be expected to be high, whereas at Methven and Dorie only small (3–10 mm) irrigation applications were made when there was sufficient soil moisture storage to absorb it without causing drainage. Fig. 15 shows the annual precipitation and drainage from the shallow stony soils from this study and those by Thorpe and Scott (1999) and White et al. (2003). The figure suggests a tendency for the relative drainage from the current study to be less than from the earlier studies. We attribute that to firstly, greater PET in the current study where the fields have a vigorous sward all year round whereas the unirrigated sites have a poor cover during the summer and it takes some time in the autumn for the soil water content to increase enough to support vigorous growth. Secondly, with careful irrigation management irrigation season drainage can be minimised and become a small proportion of total, winter dominated, drainage. The primary influence on drainage quantities appears to be annual precipitation as the site with the highest precipitation had the most drainage even though it was irrigated less than the site with similar soils and 27% less rainfall. The Dorie site with the lowest rainfall has the lowest amount of drainage because it has higher PAW, stone-free soils and good irrigation management. With good irrigation management, irrigation season drainage can be minimised, and reduced to a few occasions when large rainfalls occur.

Evapotranspiration (ET) calculated from the lysimeter water balance from the average of complete years is less than potential evapotranspiration (PET) calculated using Penman by 13.8%, 29% and 13.4% for Methven, West Eryeton and Dorie respectively (Table 1). However, there is a seasonal variation with many winters having higher lysimeter-based evapotranspiration than that estimated from Penman. There are many reasons for the differences as there are uncertainties in all methods used for estimating evapotranspiration. For the lysimeters there is a possibility that intense rain will runoff the lysimeters resulting in less water available for drainage and an increase in calculated ET. Tipping buckets are used to record both precipitation into, and drainage from, the lysimeters. Systematic error is the most significant source of tipping bucket error (Sevruk and Lapin, 1993) and includes losses due to wind, wetting, evaporation and splashing with wind-induced error being the largest component (Habib et al., 2001). The funnels for the tipping buckets measuring drainage are enclosed, so there are no wind or splashing losses and evaporative losses would be minimal. As the rain gauge orifices are at ground level, wind induced losses will be small and the main cause of any under catch will be evaporation of light rain in the funnel or from the bucket, but with 0.2 mm tipping buckets this error is likely to be minimal. The rain gauge pit and anti-splash grid minimise splash into the rain gauge and because rainfall intensities are relatively low in Canterbury (Thompson, 2011) there will be little splash out of the collection funnel. Habib et al. (2001) conclude that tipping bucket sampling error is negligible when sampling intervals are greater than 15 min and where the bucket size is less than 0.254 mm. Any under catch would result in a decreased ET estimate. Poor or incomplete vegetative cover on the lysimeter would also reduce ET, but observations of the lysimeters show there is full cover. Low soil water levels would reduce ET, but soil water measurements suggest that soil water levels are sufficient to support transpiration at potential rates. Eddy covariance tower (EC) based AET at Methven is about 7% higher than Penman suggesting that ET from the paddock as a whole is at potential rates Graham et al. (2015). Apart from uncertainties in EC AET, some of the difference can be accounted for by the EC tower measuring evaporation of dew that is not accounted for by the Penman method. While the Methven Penman PET was derived from the meteorological station on site, the PET for the other two sites is sourced from the VCSN that uses a splined surface fitted through climate stations (Tait and Woods, 2006). The nearest climate stations to West Eyreton and Dorie are 15 km and 27 km away respectively. Thus there is uncertainty in the VCSN PET estimates for these sites. So given the uncertainty surrounding the other ET estimation methods and the small uncertainty associated with the parameters measured at the lysimeter sites it is likely that the lysimeters provide the best estimates of AET.

5.2. Implications of the drainage results 5.4. Applicability of the data We have treated irrigation as rainfall to obtain water for water balances. That implies that to model recharge over a wider area, irrigation measurements, or good assumptions about application rates, will have to be made. The local catchment authority gets irrigation water use data from farmers that may be useful for such modelling. Within site variability of annual drainage from individual lysimeters is large, ranging from +36 to −35% of the mean annual drainage for a site over the whole study. The variability is primarily attributed to soil variability, but there could be some variability due to differences in vegetation cover. The most credible water bal-

While this study covers a range in climates and soil types in Canterbury, there still remains a question as to the applicability of the results to the wider Canterbury region. A number of issues need to be considered: • Irrigation efficiency: Centre pivot irrigators that have the potential to apply water more efficiently than travelling rotary irrigators. However, unless variable rate irrigation is employed, centre pivot irrigators tend to apply too much water near the pivot and increase drainage there. For long pivots the instanta-

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

31

Fig. 15. Precipitation (rain or rain + irrigation) vs drainage from irrigated and non-irrigated lysimeter studies in Canterbury. Dots = this study, irrigated shallow stony soils; Squares = White et al. (2003) shallow stony soils: Triangles = deep silt loam soils, both studies.

neous application rates at the ends of the pivots can be greater than the soil infiltration rate causing excess water to pond with resulting increased drainage. • Irrigation management: Only about 20% of Canterbury irrigators use soil water content to schedule irrigation leaving the rest to schedule irrigation with less reliable guidance (Dr A Davoren, HydroServices, Pers. Comm.). Unreliability of water supply encourages farmers to keep their soils topped up with water, resulting in little to no space for rain water storage, and increasing the potential for drainage. Reliable rainfall forecasts can inform irrigation scheduling practice and reduce drainage losses. • Soil physical quality: If soils have poor structure they can have low infiltration rates that result in water running off in some areas and ponding in others leading to drainage. • Topography: If irrigation rates exceed infiltration rates, then water will flow to topographically low spots where it is likely to cause excess drainage.

This study contains data from sites where irrigation is based on soil water status and rain forecasts, and where there is excess irrigation due to unreliability of water supply. Thus the study has results from sites with a range of conditions that may reflect a quasi-average Canterbury situation, and so could be used to extend the data beyond the farms on which the measurements have been taken. The implications of this study are that, given good irrigation management, most of the groundwater recharge on the Canterbury Plains will be where there are shallow stony soils and be proportional to the annual precipitation. Irrigation applications could be further optimised using predictions of precipitation and evaporation. Drainage could be minimised by reducing irrigation in the autumn and thus increasing the potential for storing precipitation in the soil and delaying the onset of drainage. The results of this study could be used for soil water balance modelling, testing drainage assumptions in models such as OverseerTM (Wheeler et al., 2003) and Irricalc (Bright, 2009), regional drainage/groundwater recharge estimation, and promoting good irrigation practice.

6. Conclusion Three sets of drainage lysimeters, each comprising of three individual drainage lysimeters, a ground level rain gauge and an adjacent weather station, have been deployed in different soils and with different climates on the Canterbury Plains, where the land use was irrigated dairy pasture. Drainage from a deep silt loam of 153 mm/year was 18% of precipitation (rainfall + irrigation). Average drainage from two sites with shallow stony soils of 331 mm/year was 33% of precipitation. While the fraction of precipitation that drained was similar to that found from similar studies in Canterbury of dryland sites, the quantity of drainage was more than from the dryland sites because total precipitation was greater because of the irrigation. Annual precipitation (rainfall plus irrigation) appeared to have more influence on drainage than the quantity of irrigation applied and shallow stony soils had more drainage than deep silt loam soils. In common with non-irrigated sites, a greater proportion of total precipitation occurred as drainage from irrigated sites in winter, when rainfall exceeds evaporative demand and soil water stores fill. Ideal irrigation management occurs when the following conditions apply: 1. Irrigation is applied when measured soil water content falls to 50–60% of plant available water in the root zone. 2. The application does not increase the soil water content above 80% of available water. 3. Irrigation is postponed when rainfall is forecast. 4. When irrigation application rates are similar to predicted PET rates. 5. When variable rate irrigation is used if there are significant variations in the water holding capacity of the soils within the irrigator footprint. The paper has examples of where the first three conditions have been applied and there has been no drainage. There are also occasional examples where there have been lapses in irrigation management resulting in drainage where irrigation has been

32

M.J. Duncan et al. / Agricultural Water Management 166 (2016) 17–32

applied when soil water contents are too high, where there has been irrigation when it has been raining or just before large rainfalls and where application rates have consistently exceeded PET. The recommendations for good irrigation management apply to all soils, but because shallow stony soils have smaller soil water stores, more care is required and these sorts of soils require more frequent irrigation with lower amounts than soils with higher water holding capacities. Acknowledgements Special thanks to farmers Craige Mackenzie, Willy Leferink and Graham Sutton for agreeing to the installation of the lysimeters and weather stations on their farms. This research was funded by Ministry of Business, Innovation and Employment, NZ, through contract C01 × 1006 Waterscape and Environment Canterbury. References Arauzo, M., Martinez-Bastida, J.J., Valladolid, M., Diez, J.A., 2010. Field evaluation of gee passive capillary lysimeters for monitoring drainage in non-gravelly and gravelly alluvial soils: a useful tool to estimate nitrogen leaching from agriculture. Agric. Water Manage. 97, 465–474, http://dx.doi.org/10.1016/j. agwat.2009.11.006. Barringer, J.R.F., Porteous, A., Salinger, M.J., Trangmar, B.B., 1995. Estimating spatial patterns of wilting point deficit using a water balance model and a geographic information system. J. Hydrol. (NZ) 34 (1), 42–59. Bethune, M.G., Selle, B., Wang, Q.J., 2008. Understanding and predicting deep percolation under surface irrigation. Water Resour. Res. 44, http://dx.doi.org/ 10.1029/2007wr006380. Bright, J.C., 2009. Estimation of seasonal irrigation water use–Method development. Aqualinc Report C08000/1, prepared for Irrigation New Zealand. Bryla, D.R., Trout, T.J., Ayars, J.E., 2010. Weighing lysimeters for developing crop coefficients and efficient irrigation practices for vegetable crops. Hortscience 45, 1597–1604. Campbell, D.I., 1989. Energy balance and transpiration from tussock grassland in New Zealand. Bound.-Layer Meteorol. 46, 133–152. de Vries, J.J., Simmers, I., 2002. Groundwater recharge: an overview of processes and challenges. Hydrogeol. J. 10, 5–17, http://dx.doi.org/10.1007/s10040-0010171-7. Di, H.J., Cameron, K.C., 2002. Nitrate leaching in temperate agroecosystems: sources, factors and mitigating strategies. Nutri. Cycl. Agroecosyst. 46, 237–256. Evett, S.R., Mazahrih, N.T., Jitan, M.A., Sawalha, M.H., Colaizzi, P.D., Ayars, J.E., 2009. A weighing lysmeter for crop water use determination in the Jordan Valley, Jordan. Trans. Asabe 52, 155–169. Gee, G.W., Zhang, Z.F., Ward, L., 2003. A modified vadose zone fluxmeter with solution collection capability. Vadose Zone J. 2 (4), 627–632, http://dx.doi.org/ 10.2136/vz/2003.6270. Goss, M.J., Ehlers, W., 2009. The role of lysimeters in the development of our understanding of soil water and nutrient dynamics in ecosystems. Soil Use Manage. 25, 213–223, http://dx.doi.org/10.1111/j.1475-2743.2009.00230.x. Graham, S.L., Kochendorfer, J., McMillian, A.M.S., Duncan, M.J., Srinivasan, M.S., Hertzog, G., 2015. Evapotranspiration in irrigated grassland under contrasting management. Agric. For. Meteorol. (In review). Greve, A., Andersen, M.S., Acworth, R.I., 2010. Investigations of soil cracking and preferential flow in a weighing lysimeter filled with cracking clay soil. J. Hydrol. 393, 105–113, http://dx.doi.org/10.1016/j.jhydrol.2010.03.007. Habib, E., Krajewski, W.F., Kruger, A., 2001. Sampling errors of tipping-bucket rain gauge measurements. J. Hydrol. Eng. 6 (2), 159–166. Kear, B.S., Gibbs, H.S., Miller, R.B., 1967. Soils of the plains and downlands of Canterbury and North Otago, New Zealand. N. Z. Soil Bur. Bull. 14, 93.

Kendy, E., Gerard-Marchant, P., Walter, M.T., Zhang, Y.Q., Liu, C.M., Steenhuis, T.S., 2003. A soil-water-balance approach to quantify groundwater recharge from irrigated cropland in the North China Plain. Hydrol. Process. 17, 2011–2031, http://dx.doi.org/10.1002/hyp.1240. Kim, Y., Jabro, J.D., Evans, R.G., 2011. Wireless lysimeters for real-time online soil water monitoring. Irrig. Sci. 29, 423–430, http://dx.doi.org/10.1007/s00271010-0249-x. Li, H., Zheng, L., Lei, Y., Li, C., Liu, Z., Zhang, S., 2008. Estimation of water consumption and crop water productivity of winter wheat in North China Plain using remote sensing technology. Agric. Water Manage. 95, 1271–1278, http://dx.doi.org/10.1016/j.agwat.2008.05.003. Marek, T., Piccinni, G., Schneider, A., Howell, T., Jett, M., Dusek, D., 2006. Weighing lysimeters for the determination of crop water requirements and crop coefficients. Appl. Eng. Agric. 22, 851–856. Martin, E.C., Loudon, T.L., Ritchie, J.T., Werner, A., 1994. Use of drainage lysimeters to evaluate nitrogen and irrigation management strategies to minimize nitrate leaching in maize production. Trans. ASAE 37, 79–83. Monaghan, R.M., Hedley, M.J., Di, H.J., McDowell, R.W., Cameron, K.C., Ledgard, S.F., 2014. Nutrient management in New Zealand pastures-recent developments and future issues. N. Z. J. Agric. Res. 50 (2), 181–201, http://dx.doi.org/10.1080/ 00288230709510290. Penman, H.L., 1963. Vegetation and hydrology. (Technical communication No. 53, commonwealth bureau of soils, harpenden) commonwealth bureaux, Farnham Royal. Q. J. R. Meteorol. Soc. 89 (382), 565–566. Rajanayaka, C., Donaggio, J., McEwan, H., 2010. Update of water allocation data and estimate of actual water use of consented takes -2009-10. Aqualinc Report No H10002/3, 118p. Rodell, M., Velicogna, I., Famiglietti, J.S., 2009. Satellite-based estimates of groundwater depletion in India. Nature 460, 999–1002, http://dx.doi.org/10. 1038/nature08238. Scanlon, B.R., Healy, R.W., Cook, P.G., 2002. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol. J. 10, 18–39, http://dx.doi.org/ 10.1007/s10040-001-0176-2. Scanlon, B.R., Keese, K.E., Flint, A.L., Flint, L.E., Gaye, C.B., Edmunds, W.M., Simmers, I., 2006. Global synthesis of groundwater recharge in semiarid and arid regions. Hydrol. Process. 20, 3335–3370, http://dx.doi.org/10.1002/hyp.6335. Schoen, R., Gaudet, J.P., Bariac, T., 1999. Preferential flow and solute transport in a large lysimeter, under controlled boundary conditions. J. Hydrol. 215, 70–81, http://dx.doi.org/10.1016/s0022-1694(98) 00262-5. Sevruk, B., Lapin, M., 1993. Precipitation measurement and quality control. In: Proceedings international Symposium on Precipitation and Evaporation, 1. Slovak Hydrometeorological Institute, Bratislava, Slovakia. Tait, A.B., Woods, R.A., 2006. Spatial interpolation of daily potential evapotranspiration for New Zealand. J. Hydrometeorol. 8, 430–438. Thompson, C.S., 2011. HIRDS. V3: High Intensity Rainfall Design System–The method underpinning the development of regional frequency analysis of extreme rainfalls for New Zealand. Revision 1 March 2011. 27p. http://www. niwa.co.nz/data/assets/pdf file/0018/104481/niwa Hirdsv3 method-rev1.pdf. Thorpe, H.R., Scott, D.M., 1999. An evaluation of four soil moisture models for estimating natural groundwater recharge. J. Hydrol. (NZ) 38 (2), 179–209. Wheeler, D.M., Ledgard, S.F., De Klein, C.A.M., Monaghan, R.M., Carey, P.L., McDowell, R.W., Johns, K.L., 2003. OVERSEER® nutrient budgets?moving towards on-farm resource accounting. Proceedings of the New Zealand Grassland Association 65, 191–194. White, P.A., Hong, Y.S., Murray, D.L., Scott, D.M., Thorpe, H.R., 2003. Evaluation of regional models of rainfall recharge to groundwater by comparison with lysimeter measurements, Canterbury, New Zealand. J. Hydrol. (NZ) 42 (1), 39–64. Yang, J.F., Li, B.Q., Liu, S.P., 2000. A large weighing lysimeter for evapotranspiration and soilwater-groundwater exchange studies. Hydrol. Process. 14 (10), 1887–1897, 10.1002/1099-1085(200007) 14<1887:aid-hyp69>3.3.co;2-2. Young, M.H., Wierenga, P.J., Mancino, C.F., 1996. Large weighing lysimeters for water use and deep percolation studies. Soil Sci. 161, 491–501, http://dx.doi. org/10.1097/00010694-199608000-00004. Zhang, X.Y., Pei, D., Hu, C.S., 2003. Conserving groundwater for irrigation in the North China Plain. Irrig. Sci. 21, 159–166, http://dx.doi.org/10.1007/s00271002-0059-x.