An Automated Operational System for Collating Field and Satellite Data for Grassland Curing Assessment. Presented by: Alex Chen and Danielle Martin

Transcription

1 An Automated Operational System for Collating Field and Satellite Data for Grassland Curing Assessment Presented by: Alex Chen and Danielle Martin

2 Outline Rationale Background Automated Web-Based System (Collate Field Observations) Satellite Remote Sensing Satellite Model - MapVictoria (Collate Satellite Observations) VISCA - Victorian Improved Satellite Curing Algorithm (Combine Field and Satellite Observations) Conclusions 2

3 Rationale As an input for the Grassland Fire Danger Index, our aim is to improve the quality and production of weekly grassland curing maps in Victoria. CFA has developed a web-based automated system, which integrates visually assessed field observations of curing with real-time satellite data. Why? This information is used to determine: Fire Danger Ratings Total Fire Ban Days Fire Suppression Difficulty Fire Preparedness Community Warnings Resource Allocation 3

4 Background Grassland Curing 0% Grassland Curing is defined as the drying out of grass, and is measured as the % of dead grass in a grassland fuel complex (Cheney and Sullivan 1997). Visual Observations of curing are rounded at 10% intervals 100% 4

5 Background Grassland Curing Grassland curing is an essential component of the GFDI It is combined with forecast weather parameters to determine grassfire danger The Fire Danger Ratings (FDRs) are determined by: The Grassland FDI in NW Victoria. The Forest FDI for the rest of Victoria. Discussions are being made to change the determination of FDRs: From the use of these boundaries. To the use of a Fuel Type Layer. 5

6 Background 1980s onwards Visual Curing Observations AVHRR Satellite Maps - NDVI and curing Past Research: (NIR Red) NDVI = (NIR + Red) Modified NDVI and FMC (Paltridge and Barber, 1988) Rouse et al., 1973 FMC and Curing (Barber, 1990) NDVI and Curing (Dilley et al, 2004) 6

7 Background 2000s onwards BCRC Research MODIS Satellite data and Levy rod Curing Observations Four Nation-wide Models (A, B, C, D): Map A NDVI MapB NDVI and Band7/Band6 MapC Normalised NDVI MapD Normalised SAVI Levy Rod Method 7

8 Background 2010s onwards Network of 150+ volunteers reporting visual observations Online Training available to volunteers Hand-Drawn Thiessen Kriging 8

9 Automated Web-Based System 9

10 Automated Web-Based System Motivation Weekly observations reported from 150+ volunteers, and validated by 30+ operations staff. Prior to 2012 all information went through via HQ staff by s and telephone. Curing map produced and distributed by HQ staff Massive manual processing and error-prone Objectives To streamline the operational workflow and automate the entire process from field data collection, validation, analysis and modelling to a final output. 10

11 Automated Web-Based System Progress onwards Prototype automated web-based system implemented and deployed operationally Enhanced system has been developed and will be in use operationally 11

12 Automated Web-Based System System Architecture Online Application System Java EE HTML 5 Role-based login Database Server SQL Server GIS Server ESRI ArcGIS Server 12

13 Automated Web-Based System Operational Workflow Field card Data Entry Web Interface Data Captured directly by Observers or via Calltakers 13

14 Automated Web-Based System Operational Workflow Data Validation Web Interface Data Validated directly by Validators or via Calltakers 14

15 Automated Web-Based System Operational Workflow Data Modelling Web Interface Data Modelled by Admin 15

16 Automated Web-Based System Operational Workflow Data Output by the System 16

17 Satellite Remote Sensing 17

18 Satellite Remote Sensing Background in Remote Sensing MODIS Satellite Observations 500m Resolution 8-day Composite MODIS 18

19 Satellite Remote Sensing Wavelength (nm) Bare Soil Cumulus Cloud Cirrus Cloud Cured Grass Green Grass Water

20 Satellite Remote Sensing Green Grass VISIBLE Chlorophyll Fuel Moisture Content Near Infrared Wavelength (nm) Mid Infrared Green Grass

21 Satellite Remote Sensing Cured Grass VISIBLE Chlorophyll Fuel Moisture Content Near Infrared Wavelength (nm) Mid Infrared Cured Grass Green Grass

22 Satellite Model - MapVictoria 22

23 Satellite Model MapVictoria Development of a New Model A 500m pixel was located at each field site (for 150 sites) Field values were correlated against the satellite values There are 3,900 records from Oct 2005 to Jan 2013 Visual Curing Value Estimated at Field Site MODIS Reflectance Values (of each band) estimated from a 500m pixel 23

24 Satellite Model - MapVictoria MapVictoria The New Curing Model MapVictoria : NDVI = (Band2 Band1) (Band2 + Band1) Rouse et al., 1973 GVMI = Band (Band ) Band (Band ) Ceccato et al.,

25 VISCA (Victorian Improved Satellite Curing Algorithm) 25

26 VISCA (Victorian Improved Satellite Curing Algorithm) From MapVictoria to VISCA MapVictoria With Forest Masked out Then the pixels are filled in 26

27 VISCA (Victorian Improved Satellite Curing Algorithm) From MapVictoria to VISCA MapVictoria is amended by adjusting the satellite values with field values. If a pixel is located at a field site, and if it contains satellite data 2 days old: the control of the field value is 80%. AGE OF SATELLITE DATA: The control of the field value increases with age of the satellite observation (from 80% up to 99% if 20 days old). DISTANCE FROM FIELD SITE: The control of the field value declines with increasing distance from the site (from 80% to 0% if 30 km). Pixels 30 km from any field site comprise satellite values only. ELEVATION: The 80%-to-0% decline for a 30 km distance changes with elevation, so that at high peaks, the field value only controls pixels within a 3 km distance. These thresholds may be altered for the finalised model 27

28 VISCA (Victorian Improved Satellite Curing Algorithm) VISCA 28

29 CONCLUSIONS 29

30 CONCLUSIONS What we have achieved Amount of manpower required to collect and collate field observations has significantly minimised by the developed automated system VISCA, an improved grassland curing input, provides more accurate information for input for Fire Danger Ratings, declaration of the fire danger period, pronouncement of Total Fire Bans, determination of fire suppression difficulty, fire preparedness and community warnings And resource allocation. 30

31 Acknowledgements Thank you. All Volunteer Field Observers (across Victoria) For reporting weekly field observations of curing Dr Edward King, Mr Matt Paget and Dr Glenn Newnham (CSIRO) For Historical Satellite Data Dr Ian Grant, Dr Paul Loto aniu, Mr David Howard (Bureau of Meteorology) For processing of Real-time Satellite Data Mr Tom Sanderson (CFA) For spatial systems / database assistance Online Services (CFA) Fire Incident Reporting Services (CFA) The Grassland Curing Team (CFA) For everything else!! grassland@cfa.vic.gov.au The Grassland Curing Team 31