The Use of Landsat for Monitoring the Amount and Method of Forest Harvesting

The importance of remote sensing of forests in developed states and developing states alike cannot be understated. Forests are a limited resource and one that was hard to quantify due to its vast cover. Upon its discovery for practical use, forest remote sensing has been used to measure and better understand the mechanisms for not only policy, “what should we do”, but in more other cases, “why is this happening”. Because of both its longevity and cost-effectiveness, Landsat is one of the best tools available to observe changes in forests worldwide, and its 30-metre resolution is a small tradeoff compared with aerial photography when considering large-scale changes to property parcels and greater. For this project, I will be demonstrating the use of Landsat in understanding forest policy by demonstrating one method of measuring changes to forests.
Figure 2a.

Brian Bancroft, Undergraduate Student, University of Ottawa

All data processing for this project was done by a combination of ArcGIS versions 9.3 and 10, All image sets were extracted, and placed into color composites using Landsat bands [4,3,2] which makes use of near-infrared reflection to indicate vegetation, and [3,2,1] the standard RGB composite similar to what is observed through our eyes. Using 30-metre digital elevation models (DEM) that were used to assess regions that were likely to have landslides. An assessment of step regions near rivers and elsewhere found that no regions that appeared to have landslides during this study in any of the datasets since 1972. For all image sets, observation showed that within the study area, there was no region that was harvested during the exposure of one dataset that appeared re-harvested within a following dataset. While for longer study periods in the future this may be an important consideration, an assumption was made that once a region was harvested, it would not be re-harvested. Topology rules were created using ArcGIS that enforced this assertion so that no region would be counted twice. Polygons were drawn for each dataset which indicated a region of harvest, and each set was used to calculate the total area harvested by year.

British Columbia regularly produces publications such as “The State of British Columbia’s forests”, a government digest which seeks to explain the current situation, why it matters and policy in light of the current situation. One claim by this publication is that the average area of clear cut shrank by ⅓ between 1989 and 2006 (p145, BC Ministry of Forests, 2010). For this survey, ratio of size between the mean area of polygons drawn over clear cuts on the 1972 and 1992 rasters is 1/2.81. This shows that the size of the average clear-cut in this study region is smaller than what the mean clear-cut is in the rest of the province.

Table 1: A summary of total harvests within the study region

Year of Data
Figure 2b.

Area Harvested (m2)

Area Harvested (km2)

Rate of harvesting between measurements (km2 a-1)

1972 1992 1999 2007

184449625.1 354230087.9 20810568.51 355460726.4

184.4 N/A 354.2 20.8 355.5 17.7 3.0 44.4

Figure 1: Extent of Study area, as seen using Google Maps. Prince George is around 80km north of this region along the Fraser River that extends on the eastern side of this map.

Figures (2a) and (2b) show the RGB composite of the 1992 data before and after polygons are applied over the harvested regions.

Description of Data
All satellite images for this project were downloaded using the United States Geological Survey’s (USGS) Global Land Survey (GLS) provided at no charge by the USGS Earth Explorer( GLS is a “a collection of high resolution satellite imagery provided in a standardized, orthorectified format, covering the entire land surface of the world except Antarctica” (USGS, 2011). Images are collected and ideal outputs are centered around certain dates in 1975, 1990, 2000, 2005 and 2010 based on the center date of the ideal images. (University of Maryland. 2011). For this project, image sets were chosen around October 1972, August 1992, September 1999 and September 2007. GLS 1975 uses Landsat 1-4 imagery and uses four spectral bands, with spectral ranges of 0.45-0.52μm (blue-green), 0.52-0.60μm (green), 0.63-0.69μm (red), and 0.76-0.90μm (near infrared) at a 60m resolution. GLS 1995 uses the Thematic Mapper sensor suite through Landsat 5 and 7, which adds the 1.55-1.75μm (mid-infrared), 10.40-12.50μm (thermal infrared) and 2.08-2.35μm (mid-infrared) bands, all at a 30m resolution. The ETM+ sensor suite used for Landsat 5 and 7 used in GLS 2000 and 2005 have the same bands as the TM suite at the same resolution among other features. The chosen region for this exercise is a square patch of forest that lays south-south-west of Prince George directly west of the Fraser River with Red Rock as the upper bound and Hixon as the southern bound. The area of the chosen region is 1750 km2. The region was chosen for its low hill gradient, proximity to water, the appearance of a nearly homogeneous composition and minimal development for the purpose of simplicity due to foreseen time constraints.

Harvested Regions by Year
1972, October 1992, August 1999, September 2007, September
0 5 10 20 Kilometers

While the methods used for this project were accurate, they were not timeeffective. Hand-drawn polygons are useful for recording change to features such as glacier extents, but they are not for plotting hundreds of small clear cuts in a region as large as British Columbia. A more efficient method of conducting a study similar to this would be to use an unsupervised classification system on a remote sensing suite such as PCI Geomatica, except in situations where only a handful of regions were clear-cut. Regardless of efficiency, these methods can be used to see change over time, and hold regional governments or private landholders account to the laws and best practices of logging.

Figure 3: All harvested regions in the study area.

Within the area of study, 8.90*104 ± 400 ha of forest land was harvested between 1972 and 2007, which makes up 50.6% of the study area. Assuming harvesting has been stable between when composite sets were taken, the rate of harvest was highest between 1999 and 2007, and lowest between 1992 and 1999 Sizes of harvest polygons have fluctuated over time. Up to 1972, most harvesting took place over large continuous regions in the east of the study area while in the northwest and the south center, some smaller harvests were observed. Between 1972 and 1992, this pattern changed as small blocks of harvesting appeared in the central and western regions of the study area, while existing large clear cuts were expanded. Between 1992 and 1999, this pattern slowed down, and very few small-size clear cuts took place, while some other clear cuts were expanded. Between 1999 and 2007, the style of harvesting was once again in large swathes which connected existing regions of clear cut in the center and the west of the study area, while in the east only small discontinuous harvests took place. Within the time of study, small settlements appeared, including the development of Baldy Hughes and Punchaw. As both these establishments were deforested well before the study, no polygon was created to show deforestation with exception of the direct surrounding regions, which were deforested in 1992.

References Change Detection and Landscape Structure Using Remote Sensing. S.E. Franklin, M.B. Lavigne, M.A. Wulder and G.B. Stenhouse. 2002. The Forestry Chronicle Detecting Landscape Changes in the Interior of British Columbia from 1975 to 1992 using Satellite Imagery. 1998. D.L. Sachs, P. Sollins and W.B. Cohen. Canada Journal of forest research A Land Use and Land Cover Classification System For Use With Remote Sensor Data. J.R. Anderson et al. 1976. US Department of Interior, Remote Sensing and GIS at Farm Property Level: Demography and Deforestation in the Brazilian Amazon. S. D. McCracken et al. 1999. Photogrammetric Engineering & Remote Sensing The State of British Columbia’s Forests, Third Edition. 2010. BC Ministry of Forests, Mines, and Lands.