Anthony Clark, Tim Brinkley, Greg Laughlin, Climate and Agricultural Risk, Bureau of Rural Sciences, Po.Box E11 Kingston, ACT, 2604, AUSTRALIA

Barrie Lamont, Exceptional Circumstances Section, Rural Support and Adjustment, Department of Agriculture, Fisheries and Forestry, GPO Box 858 Canberra, ACT, 2601, AUSTRALIA

This abstract introduces some of the aspects of climate analysis, drawing from a broader paper that will also be presented at Climanage. The broad paper outlines the Australian Governments Exceptional Circumstances (EC) policy, using the draft Australian - New Zealand standard on Risk Management (AS/NZS 4360:1999) as a broad framework for discussion. Factors such as the policy context, criteria development, the application process, analysis, evaluation, decision making and the treatment of agricultural risks are outlined from a national perspective. This abstract briefly summarises part of the analysis stage, focusing on meteorological drought. Some of the technology and methods used in the analysis are discussed, particularly new tools for characterising climatic features spatially.

1. Introduction

Part of the risk analysis stage of the EC policy involves determining the spatial and temporal dimensions of climatic factors, to determine if an event is rare (occurring once in 20-25 years) in the long-term historical context. Other analyses, which are not discussed in this abstract, are also undertaken including agronomic and economic impact.

Ideally, event frequency, agronomic impact and economic impact analysis of agricultural risk(s) are undertaken in the formal community-State Government application for EC assistance. In the national phase of an EC assessment the National Rural Advisory Council (NRAC) is responsible for final comparison of analysis against the EC criteria. The Commonwealth Minister for Agriculture, in consultation with the Federal Cabinet, is responsible for the final acceptance or rejection of an application. BRS provides an independent analysis of the event frequency and agronomic impact to NRAC.

2.1 Point based analyses

In the case of the rainfall, or meteorological drought, point-based analyses, are performed on a number of observation stations selected from an EC application area. This involves simple presentation of the specific event(s) relative to the long term historical record (typically greater than 80 years). When analysing monthly and seasonal rainfall, the ‘event’ can be ranked statistically and attributed a percentile value, where the 5th percentile provides a guideline threshold of a 1 in 20-25 year event.

In the analysis, a series of moving averages, ranging from 1 month to 36 month moving windows can be applied to rainfall or other data sets. This is used to determine the temporal scale of the drought event (Smith and McKeon 1998), that is establishing to what degree a meteorological event is short to long in duration, relative to other events in recorded history. These types of analyses are performed routinely in a commercial spreadsheet program.

2.2 Broad Scale Spatial Analysis

Other sources of information, particularly useful in broader regional contexts, are maps of both long term and specific events based on gridded rainfall data from the BoM. The grids are computer generated using the Barnes successive correction technique. This technique applies a weighted average to data reported within set grids across Australia (Jones and Weymouth 1997). On most maps, each grid represents a square area with sides of approximately 25 kilometres. The size of the grids is limited by the relative sparsity of rainfall measuring stations in some areas of Australia.

This sort of analysis is useful for determining broad scale climatic events and histories where large topographic features, such as the Great Dividing Range, and mesoscale climatic processes, like the ENSO cycle influence rainfall patterns. Alternatively, maps of rainfall reliability may be produced, meaning maps of the probability of receiving the seasonal average in a predetermined way. These analyses have been made a relatively simple task through the development of the BRS Rainfall Reliability Wizard; see for example:

http://www.brs.gov.au/agrifood/reliability.html

2.3 Regional Scale Spatial Analysis

Experience in applying the 25 km spatial resolution has shown that it can be deficient in some cases, particularly where small regions are defined by the applicants, or when the boundaries bisect one or more of the grid cells. It was also difficult in some cases to explain the interpretation of this output—the concept that the surfaces accurately summarised variability within the 25km grid by smoothing—to local producers, decision makers and agricultural professionals who were not specialist climate scientists. There was need to develop a more flexible and robust modelling framework, providing a means if interpreting climatic events at more localised scales.

The Integrated Toolset is part of a development project funded by the Climate Variability in Agriculture Program (CVAP). One component of the Integrated Toolset has been to enable sophisticated surface fitting and contouring routines as described in Hutchinson (1998 a,b), and known as ANUSPLIN in a desk-top Geographic Information System (GIS). This model framework is currently being coupled with a data warehouse called the SILO patched point data set (BoM, Queensland Department of Natural Resources and CVAP), to provide a near real time operational tool. This system has the following features and capacities, wjile an example analyses is presented in Figure 1:

• The modelling approach uses Generalised Cross Validation (GCV) to find the optimum trade off between smoothness of the final surface and model error. The user can also specify this trade off when required.

• The capacity to perform two-dimensional modelling, where the relationships between the observation sites are used to create a continuous surface.

• The capacity to perform three-dimensional modelling. This uses the relationships between the rainfall observation points and the underlying landscape (elevation), to create a continuous surface. In some regions there is potential to build surfaces at finer ‘local scales’, mapping some climatic features that are smoothed out by 25km model output or two dimensional modelling—for example some rain-shadow effects;

• Both modelling approaches have explicit whole model and spatial error diagnostics. This provides the capacity to ensure that the surfaces are robust throughout the iterations of their development. Additionally, the spatial error diagnostics enhance the interpretation of the model results.

Figure 1. An example of modelled rainfall for the Coom-Monaro region in south-eastern Australia using the Integrated Toolset, incorporating the ANUSPLIN suite of programs.

These analyses are not the final indicator of an exceptional event; for example there have been instances where a 5th percentile rainfall event was apparent at an 18 month time scale, while the important agronomic production season (Spring) within this 18 month period, was at the 60th percentile. In such cases the primary (and rapid) rainfall evaluation indicates that more detailed analysis of the agronomic impact is required, via a simulation study, remote sensing, or if available field data.

Climate information and its analysis is an important part of the EC assessment process, as it provides opportunities to undertake frequency analysis of some agricultural with a long term historical perspective, as well as analysing those risks spatially across broad and sometimes ill defined geographic regions. Information from the climate observation network is one of the few sources of reliable information for analysis within the timeframe of the EC assessment process, providing decision-makers with pertinent information about current seasonal conditions in near real time.

While climate analysis is highly valued, there are many other parts of the EC process that are important. The policy context, criteria development, the application process, the full range of analyses, evaluation, decision making and the treatment of agricultural risks should also be considered. Many of these aspects are outlined in a broader information paper to be presented at Climanage.

John Sims, Prachi Jain, Phillip Rosewarne, Jim Walcott and Craig Pearson, BRS

David Poulter, Exceptional Circumstances Section, Rural Support and Adjustment, AFFA

Power, S.B., T. Casey, C Folland, A Colman, and V Mehta, 1999: Inter-decadal modulation of the impact of ENSO on Australia. Climate Dynamics, 15, 319-324.

Smith, D.M.S and McKeon, G. (1998) Assessing the historical frequency of drought events on grazing properties in the rangelands, Agricultural Systems 57(3), 271-299.

Jones, D. and Weymouth, G. (1997). An Australian monthly rainfall dataset. Technical Report 70, Bureau of Meteorology, Melbourne, Australia, 19 pp.

Hutchinson MF. 1998a. Interpolation of rainfall data with thin plate smoothing splines: I two dimensional smoothing of data with short range correlation. Journal of Geographic Information and Decision Analysis, 2(2), 152-167. http://publish.uwo.ca/~jmalczew/gida_4.htm

Hutchinson MF. 1998b. Interpolation of rainfall data with thin plate smoothing splines: II analysis of topographic dependence. Journal of Geographic Information and Decision Analysis 2(2), 168-185. http://publish.uwo.ca/~jmalczew/gida_4.htm