Newsletters

Ovine Observer

Using the Pasture API to support feed forecasting and grazing management for southern Australia

Dean Thomas, CSIRO WA and Peter Beale, Local Land Services NSW

Author correspondence: dean.thomas@csiro.au

Background

Current and forecast pasture production is of high value to livestock farmers for feed budgeting, but these estimates are currently difficult and time-consuming to obtain. New digital technologies being developed can assist by re-purposing data from historic records, seasonal forecasts, or near real-time data from remote or on-farm sensors. CSIRO’s Pasture Application Programming Interface (API) provides a seasonal forecast at a specific location for up to 6 months into the future anywhere in southern Australia. The Pasture API is specified for a particular livestock enterprise and can forecast pasture production, ground cover, supplementary feeding and livestock growth for the scenario. As with any model-based system, the forecasts are only as good as the quality of the underlying models and input data streams. To ensure the Pasture API is operating at its best, ongoing work is bringing together a range of industry partners to test and improve the models and ensure that they can represent the regions where they are used. Further, we will navigate the best way to make the forecasts widely available, so they are as locally relevant as possible.

How Pasture API works

The Pasture API platform operates by directly accessing input data available from a CSIRO database to simulate a livestock enterprise using the GrassGroTM model. The underlying process model uses daily weather conditions from Bureau of Meteorology (BoM) sites as source data to drive water balance in a soil model and this determines the amount of plant available water in the root zone. If the soil characteristics are accurately described, plant available water will closely reflect patterns observed using soil probes (e.g. Figure 1). The plant available soil moisture determined in the model is then used to drive simulations of daily pasture growth and biomass.

Figure 1 Continuous soil water content from a soil probe in east central NSW, for the period January 2015 to January 2018.
Figure 1 Continuous soil water content from a soil probe in east central NSW, for the period January 2015 to January 2018.

By combining the resulting streams of simulation data, a historical context for current and likely upcoming seasonal conditions can be determined. The seasonal outlook uses climatology over a period of usually 20-30 years of weather data for the specified location, determined to represent enough variability in climate to adequately explain future scenarios. The diagrams below illustrate the seasonal outlook process (Figures 2a, b). The ‘spin-up’ period sets the initial state of the farm system. From that point the climatology can provide a good representation of the potential future variability, assuming long term climatic conditions are stable.

Figure 2a Percentile forecasts (10th, 50th and 90th) of total available pasture herbage (kg/ha), following simulated results for the preceding season.
Figure 2a Percentile forecasts (10th, 50th and 90th) of total available pasture herbage (kg/ha), following simulated results for the preceding season.
Figure 2b The same initial starting condition are used to initialize in-season simulations across many years (e.g. 1981 – 2018), which generate the probabilistic forecast scenarios.
Figure 2b The same initial starting conditions are used to initialize in-season simulations across many years (e.g. 1981 – 2018), which generate the probabilistic forecast scenarios.

How forecasts may be used

One of the main use-cases for the Pasture API platform is to predict the likely pasture production in a paddock several months ahead, based on current seasonal conditions combined with information about the level of stocking, type of pasture species and soil characteristics. Once farmers gain confidence in the accuracy and relevance of forecasts at their location, this opens the possibility to use the information in advanced feed budgeting. A wide range of management decisions may be informed by pasture forecasting including adjusting farm stocking rates, pasture fertilizer applications, determining the amount of land to sow crops or retain for pasture, sowing forage crops, purchase of supplementary feed and conserving forage as hay or silage. Forecasts of the likely feed supply up to three months in advance creates the possibility to make decisions early.

Figure 3 below illustrates the pasture production levels forecast (10th, 50th and 90th percentiles) at sites in Western Australia and New South Wales from May to August 2021. The black dashed and solid lines are the pre- and post-forecast tracking lines for the current season for comparison. For both sites, current and forecast conditions are above average. At Kalgan, there is a 90% likelihood of above average pasture available in August whereas at Jones Island the probability of above average pasture conditions in August 50:50 despite the currently good conditions (Figures 3a, b). 

Figure 3a Pasture Tracker – Website interface with Pasture API for Kalgan, WA shows pasture production is tracking above average for the current season
Figure 3a Pasture Tracker – Website interface with Pasture API. Users can create new sites via the dashboard and revisit the application to see how their registered site is tracking. The chart shown displays historic (green shaded), current season (black dashed and solid), and probabilistic forecasts (red, yellow and blue) of total green herbage (kg DM/ha) for Kalgan, in the Great Southern region of Western Australia
Figure 3b Pasture Tracker – Website interface with Pasture API for Jones Island, NSW is tracking above average for the current season
Figure 3b Pasture Tracker – Website interface with Pasture API. Users can create new sites via the dashboard and revisit the application to see how their registered site is tracking. The chart shown displays historic (green shaded), current season (black dashed and solid), and probabilistic forecasts (red, yellow and blue) of total green herbage (kg DM/ha) for Jones Island in Eastern Central region of NSW. 

Research and development activities

Projects are now underway that will further improve the pasture forecasting system and clarify the best delivery pathways so that the forecasts can be as locally relevant as possible. Activities include calibrating pasture plant models based on local field data, identifying priorities of farmers who might use pasture forecasts, and ensuring the other pasture research is ‘digital-ready’ so it is can be easily incorporated within the forecasting platform.

A collaborative project between CSIRO, Local Land Services NSW and NSW DPI to evaluate and improve the GrassGroTM pasture models for several locations in northern NSW will include running validation experiments for the plant models used by Pasture API and comparing these with the results of field experiments. The outcome of the project will enable the use of improved plant models in the Pasture API platform, which will improve pasture growth forecasting for this region.

Stirlings to Coast Farmers in the Great Southern region of WA is looking at how farmers can make the best use of the Pasture API forecasts as part of a new project funded by the Department of Agriculture Water and the Environment (DAWE) National Landcare Program: Smart Farm Small Grants. This work will help to identify and prioritise the type of forecasting information that farmers can use, and how best to make match the local conditions.

The Dryland Legumes Pasture Systems (DLPS) project is developing the next generation of annual pasture legumes for southern Australia. This work is funded by the Australian Government via DAWE, and GRDC, MLA and AWI as part of its Rural R&D for Profit program. Parallel pasture modelling work in the DLPS project will ensure that the research is ready to be incorporated within the Pasture API forecasting platform.

Where to next?

  • Representative livestock system scenarios and calibrated pasture models will improve forecasts

  • The possibility of data-model fusion will allow farm-level sensors to be used to inform models and improve the precision of forecasts

  • Implementing efficient processes for grower groups seeking to publish pasture forecasts on their websites

  • Engagement with end-users so that the applications are built to be agile and locally relevant as possible and to synthesise and report priority information for decision support. 

Further reading

Thomas, D.T., Mitchell, P.J., Zurcher, E.J., Herrmann, N.I., Pasanen J., Sharman C., Henry, D.A. (2019) Pasture API: A digital platform to support grazing management for southern Australia. In: MODSIM 2019; 1-6 December 2019; Canberra, Australia. Modelling and Simulation Society of Australia and New Zealand Inc. (MSSANZ). pp. 393-399.

Thomas, D.T., Flohr, B.M., Monjardino, M., Loi, A., Llewellyn, R.S., Lawes, R.A. and Norman, H.C. (2021) Selecting higher nutritive value annual pasture legumes increases the profitability of sheep production. Agricultural Systems. Submitted.