Experimental design
The Winchmore Irrigation Research Station is in the centre of the Canterbury plains, the largest area of flat land in New Zealand (43.787° S, 171.795° E; Fig. 1). It is at an altitude of 160 m above sea level, a mean annual temperature of 12 °C, and has an annual rainfall of 745 mm (range 491–949 mm)20. The soil is a Lismore stony silt loam classified as an Orthic Brown soil in the New Zealand soil classification and as an Udic Ustochrept in USDA soil classification21. Flood irrigation, known locally as border-check/dyke irrigation, was installed at the site in 1947. However, the two long-term trials, hereafter known as the fertiliser and irrigation trials, were established in 1952 and 1949, respectively.
Location of Winchmore within the Canterbury region (coloured green) and the layout of the long-term fertiliser and irrigation trials over time.
Full details of the setup of the fertiliser and irrigation trials between 1949–1951, including the political rationale for the trial, its statistical design, cultivation dates, sowing rates of perennial ryegrass (Lolium spp) and white clover (Trifolium repens) and initial fertiliser and irrigation treatments are available elsewhere20.
The fertiliser trial has 20 border check irrigation bays divided into five treatments each with four replicates set out in a randomised block design (Fig. 1). From 1952/53 to 1957/1958 treatments were: nil (no P applied), 188, 376 (annually and split P applications), and 564 kg SPP ha−1. All P applications occurred annually in autumn except for the 376 kg SSP ha−1 treatment which had two treatments divided into an annual autumn application and split applications in between autumn and spring. From 1958/59 to 1979–80 the nil and 188 and 376 (split autumn and spring application) SSP treatments were unaltered, while P applications were stopped to the annual 376 and 564 SSP treatments. In 1972, 4.4 t/ha of lime (caclium carbonate) was applied to all treatments22. From 1980 onwards the nil, and 188 SSP treatments and the 376 SSP treatment, now receiving winter fertiliser applications, were joined by a treatment applying 250 SSP ha−1 in winter to the previous 376 SSP treatment and a Sechura rock phosphate treatment applying 22 kg P ha−1 in winter to the former 576 SSP treatment.
Each irrigation bay was fenced off, 0.09 ha in size and grazed by separate mobs of sheep at 6, 11, and 17 stock units (with one stock unit equivalent to one ewe at 54 kg live-weight) per replicate for the nil, 188 SSP, and 376 SSP treatments, respectively. This separation prevented carry-over of dung P and other nutrients and contaminants between treatments. No grazing occurred in winter. Flood irrigation was applied when soil moisture content (w w−1) fell below 15% (0–100 mm depth). This occurred on-average 4.3 times per year.
The irrigation trial had 24 irrigation bays (each 0.09 ha in size) which had lime applied to the whole trial in 1948 (5 t ha−1) and 1965 (1.9 t ha−1) to maintain soil pH at 5.5–6.0. From 1951 to 1954 treatments were SSP applied at 250 kg ha−1 in autumn annually and either four replicates of dryland, or five replicates of irrigation applied at one, two, three, six-weekly intervals or at three-weekly intervals in alternate seasons. From 1953/54 to 1956/57 the weekly and two-weekly treatments were replaced by irrigation when soil moisture in the top 100-mm of soil reach 50 and 0% available soil moisture (asm), respectively. In 1958 the irrigation trial was cultivated with a rotary hoe and grubber, 140 kg SSP ha−1 applied and the site re-sown in ryegrass and white clover. From 1958/59–2007 the site had the same blanket application of SSP and four replicates of dryland, while a completely randomised design was used to impose five replicates of four treatments (Fig. 1) that looked at irrigation applied when soil moisture in the top 100-mm of soil reach 10, 15 and 20% (equivalent to 50% asm with 0% asm being wilting point) and irrigation on a 21-day interval. The need for irrigation to the irrigation and fertiliser trials was informed by soil moisture measured weekly by technical staff using a mixture of gravimetric analyses (1950–1985), neutron probe (1985–1990) and time-domain reflectometer (1990-onwards). Irrigation was applied at a rate of 100 mm per event20.
Except for winter, when no grazing occurred, each treatment was rotationally grazed by a separate flock of sheep with 6 and 18 stock units per replicate for the dryland and 20% v/v treatments, respectively.
The irrigation trial finished in October 2007 although the P fertiliser regime continued. All irrigated treatments shifted to the same three weekly schedule as the long-term Fertiliser trial. The dryland treatment remained unirrigated. The Winchmore farm was converted into a commercial irrigated farm operation and sold in 2018. The fertiliser trial was also sold but with a covenant ensuring it continues to operate as per normal except that irrigation from 2018 onwards is now applied by spray irrigation with the aim of ensuring soil moisture is maintained above 90% of field capacity. Since January 2019 there are daily soil moisture meter records from a moisture meter installed into one of the control plots. Soil moisture, rainfall and irrigation are recorded.
The production of the Winchmore trials data records23 involved a three-step process (Fig. 2).
Flowchart of the steps involved in sampling, analysis, collation and curation and data analysis and processing of the databases from the Winchmore Trials. Note that blue and orange boxes are sub tasks associated with each step and resulting outputs, respectively.
Step 1: Soil and pasture sampling
Pasture production was measured from two exclusion cages (3.25 m long × 0.6 m wide) per plot24. Areas within each cage were trimmed to 25 mm above ground level and left for a standard grazing interval for that time of year. Following grazing a lawnmower was used to harvest a 0.40 m wide strip in the middle of each enclosure to 25 mm above ground level. The wet weight was determined, and a sub-sample taken to determine dry matter content with a separate sample manually dissected into grass, clover and weeds. All surplus mown herbage was returned to the plot. Approximately 9–10 cuts were made annually. A composite soil sample of 10 cores (2.5 cm diameter and 7.5 cm deep) was collected from each plot. These were collected four times annually (July, prior to fertiliser application, and October, January and April), using established best practices24,25. In 2009 soil samples were also collected from the 0–75, 75–150, 150–250, 250–500, 500–750, and 750–1000 mm depths on both trials17. During 2018, prior to cultivation, soil on the unirrigated, 10 and 20% soil moisture treatments of the irrigation trial were sampled at 0–150, 150–250, 250–500, 500–750, 750–1000, 1000–1500, and 1500–2000 mm depths. The top 250 mm of these samplings were collected by hand using an auger, but deeper depths were excavated via a mechanical digger. Representative sub-samples were taken from each depth. Annual samplings were crushed, dried and sieved <2 mm for storage and later chemical analysis. The depth samplings were crushed, dried and sieved <6 mm for storage and chemical analysis. The mass of stones was recorded during sieving for the depth sampling.
In February 2002 and August 2007 10 cm diameter and 5 cm deep rings were taken in duplicate of the Irrigation trial for later soil physical analyses.
During 2017 soil on all treatments of the fertiliser trial was sampled using a soil corer to 0–75, 75–150, and 150–300 mm depths at five equally spaced distances centrally located down the irrigation bay. A further series of soil samples were obtained from the fertiliser trial during 2018 from the nil, 188 and 376 kg SSP ha−1 yr−1 treatments at 0–75 and 75–175 mm depths at five equally spaced distances down and five locations across the irrigation bay.
Step 2: Soil and fertiliser analyses
Soil chemical samples were commonly analysed for Truog P (1952–1981)26, Olsen P concentration (1976-onwards)27, pH in water28, exchangeable cations (potassium (K), magnesium (Mg), calcium (Ca)29, sulphate-sulphur (S), and occasionally organic S30, reserve K31, inorganic P32, organic C33, organic matter33 and total C34, Cd35, P36, fluorine (F)37, N38 and uranium (U)11. Stored samples of the fertilisers applied from 1998–2010 were also digested and analysed via inductively coupled plasma-optical emission spectroscopy (ICP-OES)11. Several quality assurance checks were made of these analysis and soil moisture measurements (see Technical Validation section). Soil physical measurements of porosity, bulk density, particle size (proportions of sand, silt and clay), and hydraulic conductivity39 were made for samples taken in 2002 and 2007 from the Irrigation trial. Some assessments were made of soil biological diversity in the Irrigation trial (microbial, fungal and invertebrate communities)6,7.
Step 3: Collation and curation
Data were sourced from several technical reports and published articles (Table 1). All data in reports written prior to 1995 had to be digitised before being entered into Excel spreadsheets. Data were converted via a commercial data capture service who used an XML-based data conversion tool, ImageXP with an estimated accuracy of 99.995%40. In addition to records, archived soil samples, each with a unique identifier outlining the trial, year and season of sampling, sample depth and replicate plot number have been stored in a soil archive based at AgResearch’s Ruakura campus in Hamilton, New Zealand. Soil samples exist for approximately 85% of years sampled for the fertiliser and (up to 2007) irrigation trials. Samples of the fertilisers applied to both trials exist in the archive for 1997, 1998, 1999, 2000, 2001, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018 and 2019.
Step 4: Data analysis and processing
Removal of outliers
We used methods outlined in the literature to check our data for outliers41. We established an acceptable range for daily temperature of −20 to +40 and a daily maximum rainfall of 100 mm as these were consistent with long-term meteorological data collected since 1950. Acceptable ranges for soil moisture, soil Olsen P, K, Mg, Ca and S concentrations were set at 0–100% asm, 1–200 mg Olsen P kg−1, 0.5–12 me Ca 100 g−1, 0.1–5 me K 100 g−1, 0.1–4 me Mg 100 g−1, and 1–120 mg S kg−1. The soil chemical concentrations, for example for Olsen P, corresponded to the maximum that were unlikely to indicate contamination by dung or fertiliser42. An upper range for pasture harvest data was set at 5,000 kg ha−1. Data outside these temperature, moisture, soil Olsen P concentration and pasture harvest ranges were discarded. For soil total Cd, concentrations below a detection limit of 0.5 µg kg−1 were recorded as 0.25 µg kg−1. Checks are continuing for regular soil analysis of K, Mg, Ca or S, while no checks are planned for other soil data owing to their infrequent measurement.
We calculated the mean standard deviation (SD) for soil Olsen P, S, and exchangeable Ca, K, and Mg concentrations in decadal intervals from 1960 and flagged those values that were >± 3SDs away from the mean. These data were inspected manually and excluded if variation was not consistent with other replicate plots. This removed <0.3% of the data.
Consistency
We compared weekly soil moisture across replicates and treatments to determine if irrigation events had occurred as the treatment design called.
Missing values
Our dataset contains some missing values that cause gaps in our data records. These were caused by either no sampling or a loss of data records. Missing values are indicated by blanks in our dataset. In the Fertiliser trial they constituted ~18% of pH and exchangeable cations data for 36 sampling dates and 9% of Olsen P data for 20 sampling dates. However, means were available for most of these dates. In the Irrigation trial, 4% of pH and exchangeable cations data for 6 sampling dates and 5% of Olsen P data for 7 sampling dates. Unlike the Fertiliser trial, means were unavailable for these missing dates.
Future content
The intent of the Winchmore database is for it to be kept live. Updates will be made to the Figshare repository23 as and when new data become available. Additional reports on annual soil chemical analyses and pasture production and botanical composition for the Fertiliser trial are available at www.fertiliser.org.nz.
Source: Ecology - nature.com
