University of Idaho

Departments of Geography and Biological & Agricultural Engineering

(208) 885-7333

lauriea@iron.mines.uidaho.edu

http://www.ucgis.org/oregon/papers/ames.htm

Each year thousands of hectares are burnt in agricultural areas. Wheat stubble and Kentucky bluegrass burning particularly impact Eastern Washington and Northern Idaho. Reports on the amount of hectares burnt vary widely depending upon the agency or group issuing the report. With the increased awareness of air quality, this issue is a concern due to the health risk involved.

To accurately assess the amount of hectares burnt, it is proposed that satellite imagery be used. Satellite imagery provides objectively derived estimates that are need by all agencies and groups affected by agricultural burning. Similar techniques are being explored in determining the location and amount of land burnt by forest fires. These techniques focus on heat detection of live fires, using infrared bands from AVHRR satellite imagery and are not a fine enough resolution for small acreage. Burn acreage detection on farmland requires a small pixel size. Fields are typically 16.2 to 48.6 hectares. Therefore Landsat TM 30 meters pixels is preferred.

This is an ongoing study and these are preliminary results. Although we have had some success on a subset of the study area (Latah County) the work is not complete for the entire area.

Each year thousands of hectares are burnt on the Palouse and other agricultural areas. Reports on the amount of hectares burnt vary widely depending upon the agency or group issuing the report. With the increased awareness of air quality, this issue is a concern due to the health risk involved. Concerned citizens formed a group called, Save Our Summers (SOS), this group states that "hundreds of thousands of acres"¹ of wheat stubble in Eastern Washington are burnt each year. Their figures for the amount of field burning state that "(b)etween 1963 and 1195, burning in Eastern Washington expanded from 16,000 to 60,000 acres."² "Acreage has grown dramatically in the last few years. There (was) three times as much burning last year (1996) as the year before."³ (Grant Pfeifer, DOE spokesman. Capital Press, October 31, 1997) SOS statistics show "(a)sthma rates in Spokane County are twice the national average"⁴ they believe this is related to the increase in field burning.

Wheat and grass grower association and the regulating agencies also put out similar information, however the data is different from each group. The controversy has escalated and lawsuits are currently in the courts over the amount and practice of field burning. With our information we hope that one issue will be mitigated and the parties can work on the other issue and come up with some solutions to the overall problem of air contaminate. "Field burns, because they produce a great cloud of air contaminants, are highly visible and have the potential for substantial effect on visual range and on a wide variety of measures of general visibility."⁵

Wheat Field Stubble Burning

Satellite imagery can be a useful tool to objectively assess the amount of hectares burnt. Similar techniques are being explored in determining the location and amount of land burnt by forest fires. The MegaFires project in Mediterranean Europe uses satellite images for fire management and estimate post-fire effects (http://www.geogra.alcala.es/megafires/web.htm). These techniques use AVHRR satellite imagery for the majority of their work, but also use Landsat imagery for validation.

The study area includes four Eastern Washington counties; Spokane, Whitman, Garfield, and Columbia, plus three Northern Idaho counties; Kootenai, Benewah, and Latah. These areas are large producers of wheat and a major producer of bluegrass seed. Producing 90% of bluegrass seed according to the Grass Seed Growers Association.

Figure 1: Washington & Idaho Agriculture Burn Counties

Objectives

Develop a procedure and evaluate the viability of using Landsat TM imagery to determine hectares of fields burnt. If we can accurately estimate burnt hectares, we can develop technique to aid agencies in their assessment of the impact of agricultural burning. All interested parties will have objectively derived estimates rather than widely varying estimates. The applications of this technique are far reaching in agriculture, forestry and for air quality issues.

The issue of field burning has come to the forefront in the news the last few years. Current estimates of how many acres are burnt rely on farm reports, permits and visual inspection by aircraft. As the Palouse has grown in population the affects of fielding burning and health issues has increased. The EPA has been asked to evaluate wheat stubble burning and decide on enforcement according to an article in The Idaho Spokesman-Review, Saturday, February 26, 2000 edition. As funding is cut back it becomes a more critical issue of how to enforce and determine agricultural burning each year. Satellite imagery can be used to help mitigate the problems of determining acreage as well as helping to cut the high cost of air flight time.

Interested agencies include the Natural Resource Conservation Service (NRCS), Department of Environmental Quality and the Department of Environment. These agencies, farm cooperatives and associations all need to have accurate information as they make informed decisions on field burning, plant rotation and management practices

This on going study started with one Landsat 7 Enhanced Thematic Mapper (ETM) image taken on September 18, 1999. This image was purchased from EROS Data Center with the help of an Idaho Space Grant Consortium grant. For more information on Landsat 7 and the Enhanced Thematic Mapper (ETM+) visit this website – http://geo.arc.nasa.gov/sge/landsat/l7.html swath width is 185 kilometers, satellite return is sixteen days, it flys at an altitude of 705 kilometers and was launched April 1999.

The study image is path 43 row 27, this area covers North to Lake Pend Oreille and the Spokane-Coeur d’Alene corridor to the Snake River in the South. The Eastern most edge is the Idaho Panhandle and the Western at the beginning of Lake Roosevelt where the Columbia River enters the lake. The eastern section is of most interest to our study. The image data was correct for scan direction and band alignment but without radiometric or geometric correction.

For analysis ERDAS Imagine 8.3.1 and ArcView GIS version 3.1 on Micron NT workstations were used in this study.

After receiving the image a subset selected to help with loading time and to develop our technique. Landsat images are large files. Latah County was chosen as information was easy to obtain for the county.

Figure 2: Latah County, Idaho

An unsupervised classification preformed of 30 classes. The software based on numerical information in the data groups spectral classes. Clustering algorithms, choose natural (statistical) breaks in the data. It then assigns the pixel to one of the 30 classes. Unsupervised classification does not start with a pre-determined idea of where the pixels should be assigned as is done in a supervised classification. The following website has excellent information for those unfamiliar with remote sensing: http://www.ccrs.nrcan.gc.ca/ccrs/eduref/tutorial/indexe.html

Figure 3: Unsupervised Classification, http://www.ccrs.nrcan.gc.ca/ccrs/eduref/tutorial/indexe.html

Examining the spectral properties of the 30 classes allowed a determination of the class which had high thermal emittance (band 6), low vegetation (band 4) and dark or low in the visible range (bands 1, 2 & 3), this class was then our class for burn scar. This class was colored blue to distinguish it from the other classes.

From the 30 classes we selected out five classes based on areas that were of known landcover types. The five classes are: forest, grass, burnt fields, wheat stubble and bare soil. These classes gave a reference image and all classes agreed with ground truth..

Figure 5: Plot of Five Main Classes

The areas that showed as burn fields were matched with records at the NRCS. The results have been encouraging this far. Our preliminary analysis of burned fields, derived from the single ETM image agrees with NRCS records of fields that had been burnt in the period just before our image acquisition. One field that showed as burnt was not in the NRCS records after a field check it was determined that the field had been burnt.

Distinguishing dark soils from burnt ground so that pixels of dark bare soil were not confused with burnt soil and added to the hectare count. Two basalt quarries showed in the unsupervised burn class, to be sure of our technique we felt it necessary to take dark soils into account and rule out any confusion.

A soil coverage was obtained from NRCS and the image was brought into ArcView for analysis. An Avenue script was written to aid in converting the satellite image subsets. In order to work with the image in Arc View and use any type of analysis the image must go through three steps. First, it must be made into a shapefile. Second, the shapefile converted to a grid the burn polygons selected and the selected polygons made into a shapefile again. Third, the shapefile must be converted back into grid of only the burn polygons.

The image was georegistered to the soils coverage within ArcView.

Overlaying the soils and burn coverages lead to the decision that dark soils are not being confused with the burn soils, but further statistics need to be calculated in order to insure the validity of the technique.

Actual burn in hectares for Latah County total to 9181 hectares for the September 18, 1999 image. The other counties are still waiting processing.

As we develop a robust operational technique to be used by future agencies we will need to obtain more images and add the temporal factor to the dimension. By adding an image scene after September 18, 2000 will allow us to identify fields that had been burnt, then tilled. A scene taken after our image date will also give us an idea of what happens to those areas over the return period of the satellite.

By adding a layering of information on soil type and landuse (future) helps us to distinguish and validate areas of agricultural burning.

To better serve the needs of the agencies currently involved in disputes we will use images from previous years as well. This will help to alleviate discussion on whether agricultural field burning has increase over the last five years.

Integrating the image into ArcView gives us more useable image to view and manipulate for display. This will allow agencies to have not only a robust operational technique, but also a visual representation for public use. It is our goal to make this a user-friendly technique enabling us to introduce non-traditional users to the application of remotely sensed data.

V. Acknowledgements

The author would like to thank Myron Molnau for his encouragement and gentle pushes toward greatness. It was Myron’s vision for this project that got me started. The person who most contributed and whose enthusiasm has kept my going is Ken Houska from the Natural Resource Conservation Service located in Moscow, Idaho. Every project needs someone with local knowledge, Ken you have been a fountain of knowledge, thank you.

1Hoffman, Patricia, Wheat Stubble Burning Eastern Washington, Handout from Save Our Summers

2Ibid

3Save Our Summers website: www.nicon.org/sos/sos.html

4Ibid

5Molnau, Myron, "Snake River Airshed Quality From Discretionary Burning" 1992, pg. 2

Electronic Source

MegaFires Project Chuvieco, Emilio

July 13, 1999

Remote Sensing of large wildland fires in the European Mediterranean Basin

http://wwwgeogra.alcala.es/MEGAFiReS/MEGAFiReS.htm

Report

Connor, Tim

NEEF Report: Washington state Officials Colluded with the Wheat Industry to Avoid Regulation of Field Burning and Obstruct Public Involvement in an Important Public Health Controversy

Northwest Environmental Education Foundation Report

January 2000

Journal Article

F. Biard, and Fre'de'ric Baret

"Crop residue Estimation Using Multiband Reflectance"

Remote Sensing of Environment

Vol 59, Pages 530-536

1997

Journal Article

Herisoa Razafimpanilo, Robert Frouin, Sam F. Iacobellis, and Richard C.J. Somerville

"Methodology for Estimating Burned Area from AVHRR Reflectance Data"

Remote Sensing of Environment

Vol 54, No. 3, Pages 273-289

December 1995

Generic

Hoffman, Patricia

Wheat Stubble Burning: Eastern Washington

Save Our Summers-1999

http://www.nicon.org/sos/sos.html

Journal Article

Hugh Eva, and Eric F. lambin

"Remote Sensing of Biomass Burning in Tropical Regions: Sampling Issues and Multisensor Approach"

Remote Sensing of Environment

Vol 64, No. 3, Pages 292-315

June 1998

Journal Article

Mark W. Patterson, and Stephen R. Yool

"Mapping Fire-Induced Vegetation Mortality Using Landsat Thematic Mapper Data: A Comparison of Linear Transformation Techniques"

Remote Sensing of Environment

Vol 65, No. 2, Pages 132-142

1998

Conference Proceedings

Molnau, Myron

"Snake River Airshed Quality From Discretionary Burning"

1992 International Winter Meeting of The American Society of Agricultural Engineers ASAE Conference in Nashville, Tennessee

Air quality; burning; crop residue; grass

Conference Proceedings

Myron Molnau, and Maria E. Piper

"Air Quality in the Snake River Plain Airshed"

1994 Pacific Northwest Section The American Society of Agricultural Engineers and The Canadian Society of Agricultural Engineers

Utah State University, Logan Utah

Journal Article

Paulo M. Barbosa, Jose' M. Pereira, and Jean-Marie Gre'gorie

"Compositing Criteria for Burned Area Assessment, Using Multiemporal Low Resolution Satellite Data"

Remote Sensing of Environment

Vol 65, No. 1, Pages 38-49

July 1998

Electronic Source

Pinnock, Simon

World Fire Web

GOFC: Global Observation of Forest Cover Project

Date accessed: July 13, 1999

http://www.gvm.sai.jrc.it/projects/fire/wfw/wfw.html

Journal Article

Ridley, Matt

"Asthma, Environment, and the Genome"

Natural History

Vol 109, No. 2, Pages: 54-65