1. Models as bio-indicators – why we need good climate records Holger Meinke and many other colleagues CWE, Plant Sciences Group, Wageningen University, Netherlands APSRU, Australia

2. Converting the vagrancies of weather and climate into real options for risk management … … is a long and treacherous task that requires science without disciplinary dominance, good partnerships and lots of patients. Uncertainty = randomness with unknowable probabilities Risk = randomness with knowable probabilities Real Options = values flexible & adaptive mgt in response to uncertainty (in contrast to discounted cash flow analysis) Knight, F.H., 1921

3. Background and Motivation The 20 th century was the century of analysis based on new discoveries and exploring biological systems in ever increasing detail (from discovery of DNA to mapping of the human genome), creating new disciplines in the process The 21 st century is rapidly becoming the century of synthesis with much greater emphasis on holistic approaches and creating new insights at the interfaces of disciplines (transdisciplinary)

4. Analysis Action Climate factors (eg rainfall, MSLP, SST) ImpactsImpacts and adaptation Integration with other issues – policy, risk management Climate risk mgt mainstreamed, increased ‘climate knowledge’

5. OPPORTUNITY CRISIS HARDSHIP Without adaptationWith adaptation Source: IRI Two sides of the same coin VulnerabilityAdaptive Capacity

6. Example of an application Simulation models are ideal diagnostic tool for identifying broadly based seasonal to annual climatic drivers Hence, we used simulated wheat yield as an integrative quality of rainfall over a long period Contrary to traditional climatic analyses we started with impacts to investigate processes (model as a bio- indicator) Potgieter, A.B, Hammer, G.L., Meinke, H., Stone, R.C. and Goddard, L., 2005. Spatial variability in impact on Australian wheat yield reveals three putative types of El Niño. Journal of Climate, 18: 1566-1574.

7. Probability of exceeding longterm median wheat yields in El Niño years for every wheat-producing district in Australia simulated with a process-based model that uses daily temperature and rainfall records as input and has ‘memory’ at a range of time scales. Now conduct a multivariate analysis on these simulated district wheat yields to reveal spatial impact patterns Dual purpose application

8. Multivariate Analysis Footprint 1Footprint 2Footprint 3

9. Average standardised shire wheat yield relative to all years Footprint 1 (6 years) Footprint 2 (9 years) Footprint 3 9 years

10. Average standardised shire rainfall relative to all years Footprint 1 (6 years) Footprint 2 (9 years) Footprint 3 9 years

11. FP1 FP2 FP3

12. SPCZ FP1 FP2 FP3

13. Footprints of El Niño The evolution of the EN events (timing of onset and locations of the major SST and MSLP anomalies) is highly variable but can be clustered into 3 distinct groups. Over Australia MSLP anomalies were shifted southward affecting the latitude of the sub-tropical ridge, which has significant association with rainfall variability. The shift in tilt of the line of separation between warm and cool SST pools from years in FP1 and FP2 to those in FP3 suggests linkage to the location of the South Pacific Convergence Zone, which has significant association with decadal variability in rainfall.

14. Results  The results indicate that climate system drivers that are responsible for variation in EN events and decadal rainfall patterns.  The associations suggest that variability in impact is most likely forced by differences in the temporal evolution and spatial extent of the ocean/atmosphere patterns. Features outside the tropical Pacific are also contributing.  Plausible mechanisms need to be investigated further by linking of crop models to GCM (lead to better targeted predictive agricultural systems).

15. … but NOT to analyse in ever greater detail quantities that decision makers have no control over (e.g. rainfall). We really need them as input into biophysical decision models that provide REAL OPTIONS for decision making. We need good, longterm, (daily) surface records …

16. Towards ‘Real Options’ for policy: Exposure vs Coping Capacity Meinke et al., 2006. Actionable climate knowledge – from analysis to synthesis. Climate Research, 33: 101-110. Biophysical Multiple dimensions high vulnerability moderate vulnerability low vulnerability

17. Local, longterm weather / climate data Sector specific demand for climate- related forecasts and risk assessments Dynamic climate model Supply of climate variables in ever increasing detail What usually happens - A case of market failure

18. Local, longterm weather / climate data Bio-physical model Sector specific demand for climate-related forecasts and risk assessments Spatial- temporal impact on BP quantity Statistical model categorising impacts (PC or pattern analysis) Spatial- temporally coherent impacts Statistical model clustering global climate indicators Global patterns of fundamental climate drivers Dynamic climate model

19. Local, longterm weather / climate data Bio-physical model Adaptive capacity creation via provision of REAL OPTIONS Spatial- temporal impact on BP quantity Statistical model categorising impacts (PC or pattern analysis) Spatial- temporally coherent impacts Statistical model clustering global climate indicators Global patterns of fundamental climate drivers Dynamic climate model

20. Adaptation and preparedness have emerged as THE biggest issues for a post-Kyoto world – not many know how to do it. The excellent SUPPLY of scientific information and insights will remain without impact (diffuse and unfocused) in the absence of clear user DEMAND for climate services. Policy makers AND practitioners need access to relevant information for informed discussions or debates. We need to become more transdisciplinary and problem oriented in our approaches to science – without disciplinary dominance. We need good modelling tools for all sectors, not just climate. Beyond Impact - the business case for Action

22. References Howden et al. 2007. Adapting agriculture to climate change. PNAS, 104(5), 19691–19696 Lo et al. 2007. Probabilistic forecasts of the onset of the North Australian wet season. MWR, 135, 3506-3520. Maia et al. 2007. Inferential, non-parametric statistics to assess quality of probabilistic forecast systems. MWR, 135, 351-362. Meinke et al. 2006. Actionable climate knowledge – from analysis to synthesis. Climate Research, 33: 101-110. Open access athttp://www.int-res.com/articles/cr_oa/c033p101.pdf.http://www.int-res.com/articles/cr_oa/c033p101.pdf Meinke, H. et al. 2007. Climate predictions for better agricultural risk management. Aust. J. Agric. Res., 58, 935-938. Nelson, R. et al. 2007. From rainfall to farm incomes - transforming policy advice for managing climate risk in Australia. Aust. J. Agric. Res., 58, 1004-1012. Potgieter, A.B, Hammer, G.L., Meinke, H., Stone, R.C. and Goddard, L., 2005. Spatial variability in impact on Australian wheat yield reveals three putative types of El Niño. Journal of Climate, 18: 1566-1574.

23. Methods ENSO classification – combined ocean (Niñ0 3.4 SST - Trenberth, 1997) and atmosphere (SOI - Ropelewski & Jones, 1987) classification. Multivariate analysis on simulated shire wheat yields (Potgieter et al., 2002) Align standardised weighted wheat yields with standardised shire rainfall Mapped 3-monthly SST & MSLP (Smith & Reynolds, 2003)

24. Using the SST time series for Niño 3.4, a year was classified as EN if the 5-month running mean was ≥ 0.5 for six or more months between April and December (Trenberth, 1997). Using the SOI time series, a year was classified as EN if the 3- month running mean was ≤ –5.5 for six or more months between April and December (Ropelewski and Jones, 1987). YearSOISSTYearSOISSTYearSOISST 190119351969  1902  19361970 190319371971 190419381972  1905  19391973 19061940  1974 19071941  1975 190819421976 190919431977  191019441978 1911  19451979 19121946  1980 191319471981 1914  19481982  191519491983 191619501984 19171951  1985 191819521986 1919  1953  1987  192019541988 192119551989 192219561990 19231957  1991  192419581992  1925  19591993  192619601994  192719611995 192819621996 192919631997  193019641998 19311965  1999 193219662000 193319672001 193419682002 