About the study site

Watergreen-Tourere, in the Flemington district of Central Hawkes Bay, is a 1250 ha property owned by Pete Swinburn and Suzanne Hoyt and Bruce Isles and Danelle Dinsdale. It is a sheep and beef farm with a mix of cultivatable and medium to steep hill country. Average rainfall is 1000 mm and soils range from heavy clay flats through to papa/sandstone hill country.


Forage trials and soil monitoring were undertaken at Tourere.


Forage trials


Key Messages

  • The combination of lime, Molybdenum and Sulphur increased visual clover content. But, suppressant spray and oversowing had no effect on clover content. 
  • Arrowleaf clover is a prolific seeder, but the challenge of managing it (slugs and poor strike of naturally hard seed) means that it is unlikely to have a role in a permanent pasture situation.


Study 1 - Assessing nutrient limitations to legume growth on farm

What was trialled

  • A replicated fertiliser trial was established in autumn 2018 within a north facing 4.5 ha hill country block.
  • There were 7 nutrient treatments: control, lime, Phosphorus (P), Potassium (K), Sulphur (S), and Molybdenum (Mo), as well a combination of lime + Mo + S.
  • In addition, there were sub-treatments involving suppressant spraying and oversowing of sub clover at 10 kg/ha.
  • Plot size was 10 m x 5 m with fertiliser treatments randomised within each plot.
  • In early and mid-spring, visual assessments of clover content were made.
  • In mid-spring plots were mowing, then grazed off.
  • Plots were re-scored for visual clover content in mid-summer.

Key findings

  • Visual clover content was 13.1% in August 2018, 16.2% in October 2018 and 8.3% in January 2019. Clover content varied widely across the plots.
  • The combination of Lime, Molybdenum and Sulphur had a significant effect on the amount of visual clover present at each scoring time.
  • Suppressant spray and oversowing had no effect on clover content. 


Study 2 - Oversowing arrowleaf clovers on dry East Coast hill country

What was trialled

  • The trial was undertaken on a steep dryland block of native grass pasture and all spray and seed applications were carried out by helicopter.
  • Arrowleaf was oversown on uncultivatable hill country (after being sprayed out) and allowed to set seed to attempt to create a seedbank of arrowleaf seeds.
  • The trial area was oversown post the arrowleaf crop with plantain.
  • Arrowleaf growth rates were measured using 3 exclusion cages with measurements every 4-6 weeks.
  • From Year 2, pasture growth rates were also measured in a resident pasture of similar aspect and contour. Seedling counts and amount of seed set was also measured.

Key findings

  • Oversowing was very successful in terms of achieving a one-off high-quality crop of arrowleaf clover yielding 10 t/ha in an ungrazed situation.
  • Despite large quantities of arrowleaf seed set in Year 1 (1380 kg seed/ha), germination in subsequent years was poor probably due to the hard seed (commercially available seed is scarified to increase germination).
  • Beef cows found the plant material post seed set unpalatable and a large amount of trash was left which created a haven for slugs which negatively impacted seedling establishment of arrowleaf and oversown plantain.
  • Arrowleaf clover is a prolific seeder, but the challenge of managing its means that it is unlikely to have a role in a permanent pasture situation. 


Soil Monitoring


What was achieved

  • A wireless sensor network was established that enabled some of the first daily farm scale mapping of soil properties (temperature and moisture) in New Zealand hill country. These maps can be used to drive forage yield models and help inform decision-making on pasture management.

What was trialled

  • The trial investigated the potential for modelling the distribution of soil temperature and moisture in hill country landscapes at high spatial and temporal resolution. The spatial resolution of existing widely-available soil temperature and moisture data is too coarse to provide useful information at farm scale in hill country, and is unable to account for the influence of topography in these landscapes.
  • A wireless sensor network (WSN) was installed at Tourere in July 2020. The WSN consisted of twenty sensors installed in the soil at 30 cm depth. The sensors were distributed across the farm in a way that accounted for topographic variation in elevation, aspect (the direction a hillslope faces) and the potential for water to accumulate (strongly influenced by slope gradient).
  • The sensors were configured to measure soil temperature and soil moisture at hourly intervals and report measurements back to a cloud database via the cellular network. On the farm, LoRa technology was used to communicate between components of the WSN.
  • Statistical models were fit to the soil temperature and moisture data in order to relate those soil properties to other topographic variables including elevation, aspect and slope. The models were used to predict soil temperature and moisture across the farm at 30 m resolution at daily intervals across a generic model year.

Key findings

  • The WSN performed well.
  • As expected, sensor data revealed that north-facing slopes tended to be warmer and drier than south-facing slopes, which reflects the influence of topography. Interestingly, soils on north-facing slopes warmed from the winter minimum temperature through an arbitrary threshold of 15°C about 44 days faster than soils on south-facing slopes (about 90 days versus 134 days in 2020).
  • The soil temperature model performed very well, but the soil moisture model performed relatively poorly. The difference in performance is due in part to differences in predictability between soil temperature and soil moisture, the former varying more smoothly and more regularly over time than the latter.
  • Soil moisture predictions derived from the model should be interpreted with caution, but should be good enough to provide a broad indication of when soils are near field capacity versus when they are near wilting point.
  • It is expected that model performance should improve with a longer time-series of data, and better sensor calibration.