Introduction

For many years, humans have altered and changed the land for needs such as agriculture and cultivation, mining, and infrastructure.  In terms of agriculture in particular, many methods have been implemented to increase efficiency, one of which is popularly known as “slash-and-burn.”  This technique is implemented in many forests throughout the world, particularly in tropical and subtropical regions, including the mountains of Northern Thailand in the Chiang Mai province where trees are cut down and the land is burned to clear space for agriculture.  This technique severely degrades the land since soil erosion is significantly heightened because the amount of runoff when it rains is significantly increased.  Although some studies claim that many local farmers have used this method for hundreds of years and know the dynamics of the technique, this can still be a detrimental problem for this area if the technique is malpracticed by the younger generation, or practiced in excess.

A number of studies have undertaken the task to revealing the negative effects of slash-and-burn agriculture on the forests of the Chiang Mai Province.  However, the problem reaches further than just eroding land; slash-and-burn puts carbon emissions into the atmosphere when wood is burned, not to mention effects that spill over into the social sphere.  Traditional ways of life, such as preferred crops, can be lost should slash-and-burn be pushed to a threshold past where any restoration of soil health is insignificant or ineffective.

With this study, it is our hopes that we can clearly present the harms of slash-and-burn agriculture and the ways it is not only causing detrimental health effects for the people and the soil, but also building up an accumulation of unhealthy soil history that can someday reach a point where the soil becomes no longer cultivable.  We wish to use satellite remote sensing to visually identify the extent of the fires, the frequency with which they are set, a time series of this practice from past to present, and make projections for the future.

In this study, we will focus on thermal satellite sensors to find areas where temperatures are notably higher than surrounding areas to identify the fires.  Then, we will compare GIS imagery from a time series to further analyze the effected through time "slash-and-burn" agriculture has on the landscape over a period of five years between 2001 and 2006.

Methods

The geographical extent of this study spanned the Chiang Mai province in Northern Thailand.

To study the practice of slash-and-burn agriculture in Northern Thailand in comparison to changes in vegetation cover and greenness, infrared bands from ETM+ images must be examined to create a Normalized Difference Vegetation Index.  Because slash-and-burn involves fires, We downloaded "MODIS/Terra Thermal Anomalies/Fire 8-Day L3 Global 1km SIN Grid V005" from NASA's Reverb | Echo website to visually diagnose the extent of fires in two different time periods, year 2001 and 2006, and note any significant changes.

In terms of spread, we must first understand the topography of the region.  Because fire spreads uphill, hilly and mountainous regions will facilitate a spread in extent of an existing fire.  To document this, ETM+ data was downloaded from the University of Maryland's Global Land Cover Facility site on Thursday, Nov. 10, 2011.  Our data was processed through the ENVI software to generate NDVI data and create a grayscale model of elevation for a 3-dimensional model.  We then compared two different projections from April 9, 2001 and February 18, 2006, and any difference (increase or decrease) was noted.

Ancillary data included images of the study area taken from Google Earth and FIRMS Firefly.  Google Earth was used as a tool to locate, examine and support our analysis of the study area.  The FIRMS Firefly engine was used as an additional tool to verify and/or dispel MODIS data.


SPECIFIC STEPS FOR ANALYSIS:

Downloading ETM+ data of the study area:

1. Visit the Global Land Cover Facility website from the University of Maryland to access and download map files.
2. On the right column labeled "Download Data", click the link "ESDI".
3. Click the picture labeled “Map Search” to search via a world map.
4. On the left column, choose desired dataset (ie. ETM+, ALI).
5. Search by place by choosing "Place" above the map.
6. In the "Place" search bar, type in place name (ie. "Chiang Mai, Thailand").  Choose "Preview & Download" to choose one from a series of images of that region.

Downloading MODIS data of the study area:

1. Access NASA’s Reverb | Echo website.
2. Under “Spatial Search”, zoom into interested study area (in this case, Northern Thailand); click and drag to set a bounding rectangle.
3. Under “Search Terms”, type “MODIS Fire” to get results that are related to fire and burned area.
4. Under “Temporal Search”, choose the start and end dates of data parcels for the time period of interest. For instance, input 2001-04-01 00:00:00 to 2001-04-30 11:59:59 for data in April 2001. Repeat for other months/years of interest for comparison.
5. Under “Step 2: Select Datasets”, choose the dataset of interest. In this study, “MODIS/Terra+Aqua Burned Area Monthly L3 Global 500m SIN Grid V005” was chosen to analyze all burned areas per month.
6. “Search for Granules” and accept REVERB’s pop-up notification. 
7. Use the "View spatial extent" button of each granule and the map to see its geographical extent.  If you know the spatial extent of your study area, look for files whose third portion of their name match the tile indices of your study area.  For instance, Northern Thailand is in the 27th column and the 7th row, denoted as "h27v27".  For a map of all tiles in the MODIS Sinusoidal Projection, visit this site: http://remotesensing.unh.edu/modis/modis.shtml.  For more information on the nomenclature of each MODIS data file, consult the MCD45 Burned Area User Manual version 2.0.
8. Click the "Add to cart" button for files to be downloaded.  Click "View Items in Cart" when finished.
9. To download the data files, select all and choose "Download selected" in "Text File".
10. Open the Text File and copy-and-paste the URLs beginning with "ftp://" in your browser's URL bar, and hit Enter to save them to your disk.

Opening and analyzing MODIS .hdf files downloaded from "Reverb | Echo" in ENVI:

1. Go to the Power Bar, and choose "File" > "Open External File" > "Generic Formats" > "HDF".
2. Open the .hdf file of interest.  Load "Burned Area" to analyze which areas experienced fire and when.  Load in Grayscale from the "Available Bands List".
3. Go to the toolbar of the display containing the image of "Burned Area" and choose "Enhance" > "[Image] Linear 0-255" to improve the visibility of burned areas.
4. Right click on the image and choose "Cursor Location/Value...".
5. The value following "Data:" is the approximate Julian date of burning.  The legend (as seen in page 6 of the MCD45 Burned Area User Manual version 2.0) is as follows :
• 0 = unburned
• 1-366 = approximate Julian date of burning
• 900 = snow or high aerosol
• 9998 = water bodies (internal)
• 9999 = water bodies (seas and oceans)
• 10000 = not enough data to perform inversion throughout the period
• Julian dating is used by scientific and astronomy communities to denote a date in the calendar year.  For instance, a Julian date of 45 denotes the 45th day of the year, or February 14.  See this website for a useful converter: http://aa.usno.navy.mil/data/docs/JulianDate.php.
• In the MODIS file name, date is denoted as "AYYYYJJJ", where "YYYY" is year and "JJJ" is Julian date.
6. Hover cursor over the image to find areas that are unburned or burned and when the fire occurred.
7. Match with a spectral image of the study area to assist in finding agricultural areas, or vegetation indices (ie. NDVI) to discover changes in vegetation cover possibly due to fires, or an elevation model to analyze vulnerability of spread (fires spread uphill).

Study Area

This study focuses on the Chiang Mai province in Northern Thailand (provincial boundaries highlighted in blue below).

Figure 1:  Political map of Thailand, with county boundaries and Chiang Mai highlighted in blue.

We examined the area in 2001 and compared it to 2006 to find any changes in vegetation due to fire vulnerability.  The following projections are of actual view of the Chiang Mai region where agriculture is practice in the valley, and farms and roads beginning to reach into the tropical dry forest can be seen from space.

Figure 2:  Satellite image of study area in the Chiang Mai province, April 2001

Figure 3:  Satellite image of study area in the Chiang Mai province, February 2006

A snapshot of the study area taken from Google Earth was used as ancillary data to support our analysis of the region and help us locate our study area.

Figure 4:  Image of study area taken from Google Earth (click to enlarge)

Because fires spread uphill, we also examined elevation data as an element of flammability in addition to human agency.

Figure 5:  3D elevation model of study area

Results

Figure 6:  NDVI image of study area, April 2001

Figure 7:  NDVI image of study area, February 2006

Figure 8:  Difference between 2001 and 2006 NDVIs

The following is a snapshot of the FIRMS Firefly data engine, which uses the MODIS Burned Area product to display which areas were burned by fire on which days of a specific month and year.  Our study area is outlined in red.  The different colors signify different days of the month that a fire burned.

Figure 9:  Snapshot of burned areas in study area, from FIRMS Firefly

Below is a zoomed view of the same study area, also obtained from FIRMS Firefly.

Figure 10:  Zoomed snapshot of burned areas in study area (FIRMS Firefly)

The images from the MODIS Burned Area product shows northern Thailand in April 2001 and February 2006.  White areas are areas of no data, black areas are unburned, and grey areas are burned on a specific Julian day based on the data value.  Some white areas signify water bodies, both internal (lakes) and external (ocean), as Thailand's coast can be seen on the bottom left of both images.

Figure 11:  Burned areas from MODIS data, 2001 (study area boxed in red)

Figure 12:  Burned areas from MODIS data, 2006 (study area boxed in red)

Discussion

From the naked eye, it is difficult to detect a significant change in vegetation by examining the visible image.  As shown above in Figures 6 and 7, no change can be detected.  Thus, the manipulation of the spectral bands of our images is helpful to further understand the change in vegetation in the Chiang Mai region due to slash and burn agriculture.

In ETM+ data, placing bands 3 and 4 in the order of 4, 3, 4 into R, G, B in ENVI shows an infrared image of the vegetation of the regions, based on greenness according to the Normalized Difference Vegetation Index.  By loading images in infrared and comparing them with the NDVI index, it is easier to see changes in greenness (ie. vegetation) between two images, ie. two different times.  In this study, we focused on April 2001 and February 2006, and the changes shown in Figure 8 are apparent.  The differences in greenness between the years 2001 and 2006 are shown in contrasting red and blue colors, red indicating an increase in greenness and blue indicating a decrease in greenness.  From Figure 7, we can deduce that there are notable decreases in greenness from 2001 to 2006 on the eastern portion of the study site.  From 99°15' E to 99°30'E shows a big area of vegetation decrease.  This is a more mountainous region where slash-and-burn agriculture is possibly practiced more intensely in order to clear the terrain for human land use.  Further, slash-and-burn can be practiced heavily in areas that are covered by primary forest (left uncleared) to open it up to agriculture.

Figure 9 and 10 show burned areas that do not necessarily correspond with Figures 11 and 12, but indicate that there were significant detectable fires from the MODIS satellite.  Hence, there is evidence of decrease in greenness and vegetation concentration in the study area, particularly in the eastern portion.  However, there is no significant and certain linkage between slash-and-burn and these decreases in greenness.

Because of difficulties working with MODIS data, future studies using MODIS Burned Area products can significantly improve in accuracy if geographic coordinates are superimposed onto the data points in the MODIS Sinusoidal projection (MODIS data does not come with any).  Knowing exact coordinates is extremely important for pinpointing a specific region and standardizing the geographic extent of all images produced from spatial subsetting in ENVI.

Works Consulted

Kleinman, P., D. Pimentel, and R. B. Bryant. "The Ecological Sustainability of Slash-and-burn Agriculture." Agriculture, Ecosystems & Environment 52.2-3 (1995): 235-49. Print.

Palm, C. A. Slash-and-burn Agriculture: the Search for Alternatives. New York: Columbia UP, 2005. Print. 

Roder, W. "Slash-and-Burn Rice Systems in Transition." Mountain Research and Development 17 (1997): 1-10. Print. 

Schmidt-Vogt, Dietrich. "Defining Degradation: The Impacts of Swidden on Forests in Northern Thailand." Mountain Research and Development 18 (1998): 135-49. Print.