David Tennant and David Stephens, Agriculture Western Australia, Locked Bag No. 4, Bentley Delivery Centre WA 6983, (dtennant@agric.wa.gov.au; Ph: 08 9368 3287; Fax:  08 9669 3355).

Climate services provided by Agriculture Western Australia has focussed on the development of decision support tools that enable farmers to prepare and respond to climate variability. Modelling of agricultural production has progressed to the point where farmers can use management tools to assist decision making, or utilize the delivery of timely information on crop development and yield potential. Recent work on climate indices has opened up possibilities for forecasting extreme climatic conditions with increased lead-time. Crop monitoring models that can respond to past and future expected rainfall will add value to the information provided in extension and web services.

Since the early 1980s, Agriculture Western Australia (AGWA) has had ongoing involvement in the development of a range of models, management tools and supporting data bases aimed at predicting agricultural performance in relation to incident seasonal conditions. Early applications were based on a locally developed version of the CERES Wheat simulation model (Perry 1983) and the MIDAS whole farm economic model (Morrison and others). This work culminated in the development of a number of models that address the effects of season type/seasonal variability on specific agricultural activities.

Adoption of seasonal forecasts in WA has been limited due to the low forecasting skill and lead-time of indices such as the SOI. AGWA has been a contributing partner to the Indian Ocean Climate Initiative (IOCI) which has been investigating the role of the Indian and Southern Oceans on WA climate (see Sadler, keynote address). Research by Van Loon and Shea (1987) has suggested that the mid-latitudes in the Australian/South Pacific region play an important role on the formation of extreme ENSO events. This has been further investigated with indications of extreme events clearly identified with up to a years lead-time (Stephens and Lamond, 1999, 2000). This paper will review how farm-system modelling and seasonal forecasting work has progressed and how a future combination of these areas of expertise are planned.

Models that have been developed in AGWA have included TACT, MUDAS, PYCAL, NAVAIL, STIN, SPLAT and FLOWERCAL.

TACT (Tactical Decision Making Tool; Robinson and Abrecht 1993) – Modified from CERES-Wheat to generate probabilities of yield and gross margins. Forecasts crop yield using actual rain to date and all historical finishes. Calculates the timing of major phenological events (initiation, flowering).

MUDAS (Kingwell, Morrison and Bathgate 1991) –Generates whole farm estimates of yield, income and profitability for any one of 9 user selectable season types (extension of MIDAS whole farm economic model).

PYCAL (Potential Yield Calculator; Tennant and Tennant 1996) – Allows the user to estimate stored soil water at the start of the growing season or at sowing. The user can input daily rainfall data and track the progress of the season against historical rainfall data expressed as deciles. Can forecast potential yield at any time in the growing season. Calculates water use efficiencies at the end of the season.

N AVAIL (Nitrogen Availability; Burgess, Diggle, Bowden, and Fillery 1992)– Excel based model that allows the user to input rainfall and calculate available nitrogen. Output reflects interactions between mineralisation, root growth, leaching and fertiliser source.

STIN (Stress Index model; Stephens et al. 1989, Stephens 1995) – Combines the CERES-Wheat soil water balance with the FAO (Food and Agriculture Organization) crop monitoring method (Frere and Popov, 1989). A weekly water-balance forms the basis for an accumulated stress index (from waterlogging/insufficient water) which is regressed with farm, shire, State or national yields. Yields are forecast for an average finish, or for finishes associated with selected grouping of years associated with indicators.

SPLAT (Seasonal Protein Likelihoods and Trade-offs; Bowden, Edward and Robinson 1996-1998) - Provides users with estimates of likely outcomes from fertiliser nitrogen and variety management decisions. Gives users the opportunity to trial "what if" scenarios in relation to time of sowing, soil type, variety and fertiliser choices (was generated using the NWHEAT module of APSIM).

FLOWERCAL (Flowering Calculator FLOWERCAL; Tennant and Tennant 2000) –Allows the user to calculate flowering dates for any date of sowing for a selected variety and crop. Data is also provided on the probability of frost after flowering and high temperature events before selected maturity dates.

Most of the above modelling developments have been used largely within WA, but some have been extended to other states. These include:

Following a referral to Van Loon’s work by Lamond (Lamond Weather Services), Stephens (1995) found that 40% of the variability in average Australian wheat yields could be explained by pressure anomalies at Adelaide in the year before the crop was grown. Expanding on this, Stephens and Lamond (1999, 2000) observed that large negative pressure anomalies in south-eastern Australia in winter/spring (El Nino Prediction Index-EPI) preceded the development of strong El Nino events and droughts in the Indonesian/eastern Australian region in the following year.

Continued low pressures in southeastern Australia in winter/spring are helpful in warming SST along the South Pacific Convergence Zone (SPCZ), which in turn contributes to a major sea-saw in pressure in the mid-latitudes over the following 12 months (van Loon and Shea, 1987). At the eastern end of the sea-saw, in the central South Pacific, very low pressures develop and direct westerly winds against the South Pacific high, thereby weakening the trades. Since negative pressures surrounding Rapa Island (27.6° S, 144.3° W) tend to lead a similar change at Tahiti further north, a Mid-Latitude Southern Oscillation Index (MLSOI- difference between southeast Australia-Rapa) tends to lead the SOI in autumn by 1 to 2 months. Once the El Nino is fully developed, pressure anomalies around Rapa Island eventually turn positive again over the cold SST that forms along the SPCZ. This change leads by 6 to 7 months the end of the El Nino along the Equator. The month that SST changes from warm to cold in the NINO3 region determines whether a La Nina develops or whether normal conditions re-appear. This sequence of events provides forecasting tools which have very useful lead times (summarised in Table 1).

The next logical step is to monitor prevailing seasonal conditions in conjunction with seasonal indicators. The monitoring process begins in October of the previous year. Stored soil water at the start of the season is estimated using PYCAL or STIN. PYCAL provides information at the farm or paddock level while STIN does the same at the farm to shire level. As the season progresses, PYCAL updates stored soil water estimates to the date of sowing, provides information on the progress of the season relative to decile rainfall data and provides potential yield forecasts. STIN is used to provide monthly updates of estimates of stored soil water (prior to sowing) and yield forecasts (after sowing) which are based on various assumptions about future rainfall. Average rainfall can be used, or analogues that match a similar broad-scale pattern in the weather. With the increased lead-time from the recently derived indicators these models have great scope for assisting farmers to better prepare for extreme droughts or bumper conditions.

We expect further consolidation of our modelling and decision support capacity to further enhance seasonal management of agriculture across the Australian wheatbelt. The Climate Risk and Yield Information Service is planned to be expanded and be available for selected benchmark locations on the web. As well, greater coverage of output (climate, potential yield and yield forecasts, flowering time forecasts) from the various models should become available. STIN modelled yields are available on the web for the current season. Predicted yield probabilities for future seasons could also be generated for selected analogue years. Differences in mean yields and yield probabilities between different indicator classes (phases) will further highlight the value of long range indicators and the regional differences in rainfall-yield relationships.

The SOI, EPI, MLSOI and Rapa Island pressures are best viewed as complimentary indices of the Southern Oscillation Phenomenon. The greatest benefit from these indicators should be found in eastern Australia where ENSO more regularly impacts on climate. Future research should define when ENSO impacts more on Western Australia.