Using local and historical data to enhance understanding of spatial and temporal rainfall patterns

Abstract

Farmers face uncertainty in their businesses from many factors, but rainfall is a key determinant of both the nature of the production system and variation in financial returns. Currently, various weather forecasting services are available from the Australian Bureau of Meteorology (BoM) based on about 7000 stations covering all of Australia. Seasonal Climate Forecasts are seen as another tool that can help to improve farm productivity. It is well known that many farmers keep their own rainfall records, and likely that the farmers have a high degree of confidence in their own records. Australian Bureau of Statistics figures indicate that there were possibly 7000 grain related ‘agricultural businesses’ in NSW alone in 2009/10 indicating that there is the potential to increase data density by up to an order of magnitude. This project is part of a broader study to improve rainfall predictions for grain farmers using data collected locally to the users (crowd sourcing). The data is collected directly on farm, and from other sources which may be available. The focus is on the historical data, its collection and analysis, in terms of discerning patterns in time and space which may help provide a local framework, within which coarser scale forecasts can be interpreted and understood. Data will be stored on secure database systems at the University of Sydney. Results indicate that farm data does provide more local detail, temporally and spatially. Deficit and surplus analysis demonstrates the predictive capacity of the local temporal data, despite limited data precluding the definition of ideal criteria and parameters for predictive ‘similar year’ selection. The spatial data demonstrates quantifiable site specific differences from institutional data. Testing across more climate types may allow these differences to be defined within and across regions. Tests for an indicator time period show that farm rainfall in the early part of the growing season (April and May) may indeed be indicative of seasonal condtions, while more data is needed to confirm this. The use of southern oscillation life cycle information to select appropriate years considerably improved the relationships revealed, with a doubling of relationship strength across all climatic types, although the strength of the relationships differed across the climatic types, and the strongest relationships were split between the months of April and May. More extensive analysis, with more data across more BoM districts (and therefore climate classes) will be required to confirm this conclusion, but it appears that farm rainfall records and SOI information can provide an indicator time period to help farmers interpret, refine and utilise seasonal forecasts.