Ensemble forecast products for user decisions on multi-week
Ensemble forecast products for user decisions on multi-week to seasonal timescales Debbie Hudson, Oscar Alves, Andrew Marshall, Catherine de Burgh-Day, Morwenna Griffiths, Luke Shelley, Harry Hendon, Andrew Watkins Forewarned is forearmed Managing the impacts of extreme climate events Feedback Work Package 4: Extension and training Adoption USERS Adoption Work Package 3: Interfacing to Industry decisions Photo: Jennifer Metcalfe, Econnect Communications Adoption Feedback
Feedback Work Package 2: Extreme forecast products development and delivery Feedback Feedback Research partners BoM Univ. Univ. Melbourne Melbourne Monash Monash Univ. Univ. Univ. Univ. S. S. Queensland Queensland SARDI SARDI DEDJTR DAFQ
DAFQ Birchip Cropping Group Work Package 1: User needs and Forecast system development Rural RDC & other partners Meat Meat and and Livestock Livestock Australia Australia Grains Grains RDC RDC Sugar Sugar Research Research Australia Australia Cotton RDC AgriFutures AgriFutures Australia Australia Dairy
Dairy Australia Australia Wine Australia Australian Australian Pork Pork Work Package 2: Developing and delivering forecasts Work package 2: Extreme forecast products development and delivery Develop a range of heat, cold and rainfall experimental extremes forecast products from ACCESS-S on multi-week to seasonal timescales; Make experimental products available on a research web server for trial and feedback; Deliver operational forecasts for a subset that have sufficient accuracy and utility. The products will be of broad utility across industries (i.e. not highly tailored for a specific industry).
What is an extremes forecast product? Forecasting climate extremes Averaged/accumulated over a period (e.g. week, fortnight, month, season) e.g., extremely hot month, extremely dry season Forecasting extreme weather events (beyond the 7-day forecast) Likelihood of weather events in a given period e.g., probability of heatwaves, frosts, heavy rainfall events Experimental forecast products R&D web portal ACCESS-S 99-member ensemble Forecasts are updated every day Outlooks for multi-week (sub-seasonal) and seasonal timescales Currently available product operationally is: Probability of above median Forecast for JFM 2019 Rainfall Getting more information from the forecast: Tercile probabilities
Forecast for JFM 2019 Rainfall: Probability for most likely tercile category Getting even more information from the forecast: Chance of being in outer deciles (e.g., very dry) Forecast for JFM 2019 Rainfall: Chance of being in decile 1&2 Increased likelihood of having Decile 1&2 (brown colours) (i.e. of being amongst the driest fifth of JFM seasons Climatological expected probability for Decile 1&2 (20%) Heat extremes example Average Tmax for the week Observed: Week 30 Oct 5 Nov Tmax anomaly Highest Tmax experienced in the week Heat extremes example Forecast for 30 Oct - 5 Nov (from 23 Oct, i.e., week 2 of the forecast) Chance of having a decile 9-10 week Chance of having more hot days+nights than usual in the forecast period
"Hot": decile 10 days Tmean > 90th percentile threshold of observed daily Tmean The days do not have to be consecutive OBSERVATIONS Camden (NSW) 34.04S, 150.69E Black line: The observed 90th percentile threshold of daily Tmean Calculated by looking back over the historical period for the particular time of year. 10% of days are hotter than this threshold for the given time of year. Increased hot day risk: "Orange" if some of the forecast ensemble are forecasting temperatures exceeding the historical (climatology) 90th percentile This is the 75th percentile i.e. more than 25% of the ensemble are forecasting temps > historical (climatology) 90th percentile. The chance of "hot days" is
25% (i.e., more than double the normal risk, which is 10%) FORECAST from 23 Oct The "plume" is the forecast White line: middle (median) value of the forecast ensemble Dark grey: interquartile range (between the 25th and 75th percentiles of the forecast ensemble). 50% of the forecast distribution have values in this range. Light grey: interdecile range (between the 10th and 90th percentiles of the forecast ensemble). 80% of the forecast distribution have values in this range. OBSERVATIONS Camden (NSW) 34.04S, 150.69E FORECAST from 23 Oct FORECAST from 26 Oct FORECAST from 28 Oct
FORECAST from 30 Oct Risk = probability x impact Camden (NSW) 34.04S, 150.69E OBSERVATIONS THI used by Dairy Industry Heatwaves affect milk composition and production Example: Nov 2017 Victorian heatwave 2nd hottest November on record 12% decline in milk production from the first to last week A loss of 2,500 litres/average farm From: Richard Eckard, University of Melbourne https://coolcows.dairyaustralia.com.au/ https://dairy.katestone.com.au/
Ensemble forecasts and decision-making Peter Hayman (SARDI) and Barry Mudge (Barry Mudge Consulting) Barry Mudge (1/11/18) Example: Planting vetch for sheep to graze OR lentils to be harvested 800 800 600 600 400 400 200 200 0 0 -200
43% Vetch 47Vetch 55% Lentils 222 MAX $ 1 27 Lentils 87 MAX $ 40 Lentils 222 MAX $ 127 Lentils 87 MAX $ 40 -500 0
ti l s n e L Vetch Driest 1/3 33 Mid 1/3 34 Wetest 1/3 33 Profit ($/Ha) Profit ($/Ha) Driest 1/3 Mid 1/3 Graph 1/3 Profit X Decile Graph Profit X DecileWetest
10 20 80 90 Profit ($/Ha) as a function of growing season rainfall decile (percentile) Compare "downside risk" and "upside benefit" wedges The upside benefit of planting lentils far outweighs the downside risk How does a forecast change the expected outcome? downside risk increases with increased chance of being dry 60% of the time vetch is better BUT downside risk is still outweighed by potential gains if normal/wet season eventuates. For risk neutral farmer, lentils are still better 100 Some key challenges (BoM perspective) Communicating probabilities 2015/16 engagement: User comprehension is linked with user satisfaction i.e. those that answered comprehension questions incorrectly were 3 times as likely to be dissatisfied with the service
Conveying accuracy/skill Forecast quality does not necessarily reflect value Summary skill measure (average over hindcasts). Does not show windows of forecast opportunity Large sampling uncertainty around scores of most interest Hindcast ensemble size (e.g., 11-members) is smaller than real-time (e.g., 99-members) The years included in hindcast period will influence the skill Data for data assimilation is more sparse as go back in time Seamless forecasts integrating with the NWP forecast Why consult with users and have partnerships? No matter how accurate a weather forecast or climate outlook is: If it doesnt provide information users need The sidewalk that was built The sidewalk the user wants If it isnt issued when users are making their critical decisions If it is misinterpreted If it cannot help make a
decision. *The forecast has little real value Canberra 2017 Partnerships with users will become increasingly important in order to unlock the true potential and value of multi-week and seasonal forecasts Thank you Debbie Hudson 03 9669 4796 [email protected]
Ground Control AERSP 401A Ground System's Basic Elements Mission Elements Control the space segment or handle mission data, and includes: Ground Stations Clocks Operations Centers Facility Elements Support infrastructure that includes: Physical plant (buildings, utilities, staff) Maintenance Mission Elements Ground...
The Direct Object answers the question WHO or WHAT after the verb. ... Tú estás comprando la television para mí. Yo voy a dar un beso a todas las chicas feas y bonitas. Yo les canto. ... Let's do some...
GDPR Articles 9 and 10: The GDPR requires special conditions to be met for processing "special categories" of personal data and criminal convictions data. Articles 9 and 10 of the GDPR prohibit the processing of such data unless the special...
Legal Realism: Challenges the Systems. To positivism it challenges their philosophy saying the legal authorities and systems brainwashes the public into accepting it legitimacy. To naturalism it points to the subjectivity or inability to prove moral statements like murder is...
EDINA and University Data Library (EDL) together are a division within Information Services (IS) of the University of Edinburgh. EDINA is a Jisc-funded National Data Centre providing national online resources for education and research.
Ready to download the document? Go ahead and hit continue!