PROGRESS ON A KNOWLEDGE-BASED INTERNET WEATHER FORECASTING SYSTEM 100-day evaluation trial from 25 September, 2002 to 8 January, 2003. Harvey Stern (Bureau.

Slides:



Advertisements
Similar presentations
List of Nominations Connecting User Needs with Weather Research and Forecasts Rebecca E. Morss National Center for Atmospheric Research Boulder, Colorado,
Advertisements

Enhance the Attractiveness of Studies in Science and Technology WP 6: Formal Hinders Kevin Kelly Trinity College Dublin WP 6 Co-ordinator.
“A LPB demonstration project” Celeste Saulo CIMA and Dept. of Atmos. and Ocean Sciences University of Buenos Aires Argentina Christopher Cunningham Center.
Chapter 4 Design Approaches and Methods
ASII-NG: Developments and outlook NWCSAF 2015 Users Workshop.
Sabrina Abesamis, Lily Dove, Ryan Truchelut, and Robert E. Hart Young Scholars Program, Florida State University Background Due to modern enhancements.
1 Using a Knowledge-Based System to Predict Thunderstorms Harvey Stern Australian Bureau of Meteorologyhttp://
Gridded OCF Probabilistic Forecasting For Australia For more information please contact © Commonwealth of Australia 2011 Shaun Cooper.
Validation of the Ensemble Tropical Rainfall Potential (eTRaP) for Landfalling Tropical Cyclones Elizabeth E. Ebert Centre for Australian Weather and Climate.
An Automated Synoptic Typing System Using Archived And Real-time NWP Model Output Robert Dahni Meteorological Systems Central Operations and Systems Branch.
AMERICAN METEOROLOGICAL SOCIETY 2007 ANNUAL MEETING 1 Harvey Stern 23rd Conference on Interactive Information and Processing Systems, San Antonio, Texas,
Large-scale atmospheric circulation characteristics and their relations to local daily precipitation extremes in Hesse, central Germany Anahita Amiri Department.
A knowledge-based system to generate internet weather forecasts. Dr Harvey Stern, Bureau of Meteorology, Australia Dr Harvey Stern, Bureau of Meteorology,
Creating Architectural Descriptions. Outline Standardizing architectural descriptions: The IEEE has published, “Recommended Practice for Architectural.
Verification of Numerical Weather Prediction systems employed by the Australian Bureau of Meteorology over East Antarctica during the summer season.
Ensemble Post-Processing and it’s Potential Benefits for the Operational Forecaster Michael Erickson and Brian A. Colle School of Marine and Atmospheric.
Evaluation of Potential Performance Measures for the Advanced Hydrologic Prediction Service Gary A. Wick NOAA Environmental Technology Laboratory On Rotational.
Dr Mark Cresswell Statistical Forecasting [Part 1] 69EG6517 – Impacts & Models of Climate Change.
CASE Tools And Their Effect On Software Quality Peter Geddis – pxg07u.
Probabilistic forecasts of (severe) thunderstorms for the purpose of issuing a weather alarm Maurice Schmeits, Kees Kok, Daan Vogelezang and Rudolf van.
For the lack of ground data the verification of the TRMM performance could not be checked for the entire catchments, however it has been tested over Bangladesh.
CARPE DIEM Centre for Water Resources Research NUID-UCD Contribution to Area-3 Dusseldorf meeting 26th to 28th May 2003.
Copyright © Cengage Learning. All rights reserved. 8 Tests of Hypotheses Based on a Single Sample.
National Standards and gifted and talented: are they compatible? Presented 10 Nov – 5pm Webinar 4 of a series of 4 Adrienne Carlisle Evaluation.
© Crown copyright Met Office Forecasting Icing for Aviation: Some thoughts for discussion Cyril Morcrette Presented remotely to Technical Infra-structure.
Advances in weather, climate and water forecasting technologies Alasdair Hainsworth Bureau of Meteorology March 2011.
Searching for Practical and Flexible Definitions of Extreme Events under a Changing Climate Guido Franco Public Interest Energy Research (PIER) Program.
Modeling.
GHP and Extremes. GHP SCIENCE ISSUES 1995 How do water and energy processes operate over different land areas? Sub-Issues include: What is the relative.
A Global Climatology of Baroclinically Influenced Tropical Cyclogenesis Authors: Ron McTaggart-Cowan, et. Al. Monthly Weather Review June 2013.
NOAA’s National Weather Service National Digital Forecast Database: Status Update LeRoy Spayd Chief, Meteorological Services Division Unidata Policy Committee.
Where the Research Meets the Road: Climate Science, Uncertainties, and Knowledge Gaps First National Expert and Stakeholder Workshop on Water Infrastructure.
1 CUTTING-EDGE CLIMATE SCIENCE AND SERVICES Geoff Love.
Celeste Saulo and Juan Ruiz CIMA (CONICET/UBA) – DCAO (FCEN –UBA)
Henry Fuelberg Pete Saunders Pendleton, Oregon Research Region Map Types and Lightning Frequencies.
IMPLEMENTATION QUALITY RESEARCH OF PREVENTION PROGRAMS IN CROATIA MIRANDA NOVAK University of Zagreb, Faculty of Education and Rehabilitation Sciences.
Center for Radiative Shock Hydrodynamics Fall 2011 Review Assessment of predictive capability Derek Bingham 1.
CHAPTER 12 Descriptive, Program Evaluation, and Advanced Methods.
AMERICAN METEOROLOGICAL SOCIETY 1 Harvey Stern 22 nd Conference on Interactive Information and Processing Systems, Atlanta, 30 January, Combining.
Project Phoenix – An Update EuMetCAL Workshop Toulouse, August 2008 Jaymie Gadal.
Overview of 2012/2013 winter over South Korea
17 TH BMRC MODELLING WORKSHOP – OCTOBER Harvey Stern 17 th BMRC Modelling Workshop, Bureau of Meteorology, Melbourne, 6 October, Generating.
CBRFC Stakeholder Forum February 24, 2014 Ashley Nielson Kevin Werner NWS Colorado Basin River Forecast Center 1 CBRFC Forecast Verification.
PRICING A FINANCIAL INSTRUMENT TO GUARANTEE THE ACCURACY OF A WEATHER FORECAST Harvey Stern and Shoni S. Dawkins (Bureau of Meteorology, Australia)
F. Prates/Grazzini, Data Assimilation Training Course March Error Tracking F. Prates/ F. Grazzini.
RFC Climate Requirements 2 nd NOAA Climate NWS Dialogue Meeting January 4, 2006 Kevin Werner.
Using Model Climatology to Develop a Confidence Metric Taylor Mandelbaum, School of Marine and Atmospheric Sciences, Stony Brook, NY Brian Colle, School.
Sources of Skill and Error in Long Range Columbia River Streamflow Forecasts: A Comparison of the Role of Hydrologic State Variables and Winter Climate.
CLIMATE, CATCHMENTS AND COMMUNITIES – SEPTEMBER Harvey Stern 20th Conference on Interactive Information and Processing Systems, San Diego, California,
Trials of a 1km Version of the Unified Model for Short Range Forecasting of Convective Events Humphrey Lean, Susan Ballard, Peter Clark, Mark Dixon, Zhihong.
NWS Digital Services 1 CB Operations Committee Lynn Maximuk DSPO Operations Team Eastern Region HPC Day 4-7 Grid Proposal Review, Findings and Recommendations.
INCORPORATING AN ENSEMBLE FORECASTING PROXY INTO A KNOWLEDGE BASED SYSTEM Harvey Stern, Bureau of Meteorology, Australia. Harvey Stern, Bureau of Meteorology,
Extracting probabilistic severe weather guidance from convection-allowing model forecasts Ryan Sobash 4 December 2009 Convection/NWP Seminar Series Ryan.
1 GOES-R AWG Product Validation Tool Development Hydrology Application Team Bob Kuligowski (STAR)
Slide 7.1 Saunders, Lewis and Thornhill, Research Methods for Business Students, 5 th Edition, © Mark Saunders, Philip Lewis and Adrian Thornhill 2009.
Investigate Plan Design Create Evaluate (Test it to objective evaluation at each stage of the design cycle) state – describe - explain the problem some.
Demonstrating Institutional Effectiveness Documenting Using SPOL.
Multi-Area Load Forecasting for System with Large Geographical Area S. Fan, K. Methaprayoon, W. J. Lee Industrial and Commercial Power Systems Technical.
Impact of the representation of the stratosphere on tropospheric weather forecasts Sana Mahmood © Crown copyright 07/0XXX Met Office and the Met Office.
Downscaling and Why Local Predictions are Difficult to Make Preparing Your Coast.
An Investigation of the Skill of Week Two
Mapping Climate Risks in an Interconnected System
Climatology of Inverted Troughs over the Gulf of Maine
Anne Leroy Météo France, Nouméa, Nouvelle-Calédonie Matthew C. Wheeler
Post Processing.
Test Case Test case Describes an input Description and an expected output Description. Test case ID Section 1: Before execution Section 2: After execution.
(Victorian Regional Office)
Adam J. O’Shay and T. N. Krishnamurti FSU Meteorology 8 February 2001
Dr Harvey Stern, Climate Manager, Victoria and
Presentation transcript:

PROGRESS ON A KNOWLEDGE-BASED INTERNET WEATHER FORECASTING SYSTEM 100-day evaluation trial from 25 September, 2002 to 8 January, Harvey Stern (Bureau of Meteorology, Australia)

Location Diagram (1)

Location Diagram (2)

Example (NWP MSL Prognosis)

Example (The System’s Output)

Background An early version of the knowledge-based system was presented to the 18th IIPS Conference (Stern, 2002). The system was developed for the small (227,000 sq km) southeast Australian State of Victoria. It was described as being capable of generating forecasts for public, aviation, marine and media interests, in languages other than English, and for more than 200 localities in Victoria. The large volume of output proposed here (forecasts for than 200 localities) would be difficult (if not impossible) to generate manually However, it would be straightforward (to generate) utilising automated systems.

“Bank” of forecaster experience A major benefit of the knowledge-based system is that it incorporates an extensive "bank" of forecaster experience. Ramage (1993) has proposed an "iterative" approach to "locking in" improvements in forecasting methodology. The automated nature of the system lends itself to Ramage's approach. The system's skill increases as new knowledge is incorporated into its operation. Hence, progress is gradually made towards the realisation of Ramage's dream.

Not “yet another” instrument of forecast guidance The system is not seen as "yet another" instrument of forecast guidance. Rather, its development is seen as a logical step along the path of having the computer replicate the best aspects of the manual side of the forecast process, by systematically "locking in" new knowledge. As Brooks (1995), wrote: "technology, which initially allowed humans to make routine weather forecasts, will soon close that avenue of human endeavour... (and thereby permit) concentration on severe events". The main human-interaction is in utilising forecast verification data (after the event) to iteratively incorporate new additional forecaster knowledge into its algorithm.

A Forecasting Research Tool A simple illustration is now presented of how the system for generating Internet Forecasts might be employed as a research tool to enhance understanding. The first Figure following depicts the MSL pressure (MSLP) distribution for Synoptic Type 41 (strong cyclonic SSW flow).

A Forecasting Research Tool (cont.) The Figure plots the annual march of PoP forecasts for strong cyclonic SSW flow, with 850 hPa temperatures of 0 deg C, & with 700 hPa RHs of 30% & 70%. One might ask why there is such a marked difference between winter & summer responses to the same situation.

A Forecasting Research Tool (cont.) An illustration is now presented of how one may proceed from the enhanced understanding achieved via the preceding analysis, to increase the potential accuracy of the forecasts generated. A study of Synoptic Type 41 cases suggests that the strength of the SSW flow between SE Australia and over waters to the west of that region might be significant. Regression analysis confirms that adding cyclonicity to the other predictors, cyclonicity proves to be the most significant.

A Forecasting Research Tool (cont.) The Figure plots the annual march of PoP forecasts for strong cyclonic SSW flow with RH=30% and the new equation. It reveals just how important cyclonicity is in assessing the likelihood of precipitation.

November 2001 trial The system's forecast performance (during November 2001) was evaluated for the city of Melbourne. The evaluation showed that, although superiority over climatology was achieved, the forecasts (on most measures) proved to be inferior to the official forecasts. The results of the November 2001 trial were analysed, potential improvements in the forecasting process (employed by the system) were identified, and those improvements were "locked in". The system (so modified) was then subjected to another evaluation for the city of Melbourne, this time over 100 days (between late September 2002 and early January 2003).

Overall Accuracy (in 2001)

Modifications The 2001 system operated by producing its predictions from a restricted number of discrete "forecast sets". The set that was chosen (by the system) was largely determined by the particular synoptic pattern suggested by the selected NWP model. The 2002/2003 modification utilises regression analysis to allow predictions to be selected from a continuous array of possible forecasts. The particular form of regression analysis employed is parameter enveloping. Parameter enveloping allows definition of how the various predictors impact upon, or envelope, the influence (on a predictand) of other predictors.

An illustration of parameter enveloping

Overall Accuracy (2002/2003)

RMS Error of Min Temp Forecasts

RMS Error of Max Temp Forecasts

% Correct Rain/No Rain

RMS Error of QPFs

Need for Further Testing The 100-day verification trial suggests that substantial progress, albeit uneven, towards achieving computer replication of the manual forecast process, has been made. Specifically, the impact of the "locked in" improvements upon the skill displayed by the system has been considerable, particularly for Day-1 forecasts. Nevertheless, the performance of the system over the 100- day trial is so impressive, that one must entertain the possibility that it was simply a "fluke". For this reason, further testing needs to be undertaken.

Implications It was considered that, should further testing confirm the 100- day trial results, the implications would be profound. We could see ourselves at the "dawn" of the operational implementation of Ramage's approach to weather forecasting which, through computer replication of the manual process, would allow for: Systematic incorporation of new procedures that may lead to a quantum leap in the accuracy of the forecast products (in a related piece of work, Ryan et al. (2003) have developed a system that archives the subjective inputs to the forecast process and makes them available for statistical analysis); The opportunity to greatly increase the number and variety of such products.

Further Testing So far, further testing is, indeed, confirming the 100-day trial results. A modest post-trial exercise, involving the derivation of forecasts for Day-1 only for the (now) 130 days ended February 1, 2003, yields: A CSS of 61.4 for the Internet Forecasts; and, A CSS of 60.3 for the official forecasts.