The 2008 Joint Meeting of the Society for Range Management and the America Forage and Grassland Council. |
Monday, January 28, 2008 - 1:40 PM
Detecting Aspen Woodlands Remotely – A Comparison of Spot 5 and Landsat 7 Etm+
Eva K. Strand and Stephen C. Bunting. Department of Rangeland Ecology and Management, University of Idaho, 975 W. 6th Street, College of Natural Resources, Moscow, ID 83844-1142
Current decline of quaking aspen (Populus tremuloides) is of concern across the intermountain western United States. Mixing of conifers into aspen stands and excessive browsing by herbivores negatively affect aspen establishment and can eventually lead to loss of aspen stands. To engage in pro-active management it is therefore of great interest to develop remote sensing methods for detection of aspen stands at the landscape scale. We compared the spectral separability between aspen, conifers and mountain shrub species and the detection of different sized aspen stands using images from two satellite sensors, namely SPOT 5 (10 m pixels) and Landsat 7 ETM+ (30 m pixels) in the Owyhee Mountains in Idaho. Many aspen stands are here becoming increasingly dominated by western juniper (Juniperus occidentalis) and Douglas-fir (Pseudotsuga menziesii) and it is of particular interest to detect small and heavily encroached aspen stands.
Satisfactory separability (M-statistic >1) was found between aspen and Douglas-fir in all Landsat and SPOT bands with the highest separability in the near and short wave infrared reflectance bands. Aspen and juniper were best separable in the near infrared band. Poor separability (M = 0.35 for Landsat and M = 0.65 for SPOT) was computed between the aspen and mountain shrub classes, which explains the confusion of these two types in cover type classifications. With regards to the stand size, over 90% of aspen stands larger than 1 ha were detected using either sensor. As expected, the SPOT sensor with a finer spatial resolution, was superior in detecting aspen stands in the size range 0.02 – 0.4 ha by a factor of 1.5-2. Stands smaller than 0.01 ha (10x10 m) are difficult to detect with any of these sensors. Overall, the thematic user’s and producer’s accuracy for aspen was similar for the two sensors.
Satisfactory separability (M-statistic >1) was found between aspen and Douglas-fir in all Landsat and SPOT bands with the highest separability in the near and short wave infrared reflectance bands. Aspen and juniper were best separable in the near infrared band. Poor separability (M = 0.35 for Landsat and M = 0.65 for SPOT) was computed between the aspen and mountain shrub classes, which explains the confusion of these two types in cover type classifications. With regards to the stand size, over 90% of aspen stands larger than 1 ha were detected using either sensor. As expected, the SPOT sensor with a finer spatial resolution, was superior in detecting aspen stands in the size range 0.02 – 0.4 ha by a factor of 1.5-2. Stands smaller than 0.01 ha (10x10 m) are difficult to detect with any of these sensors. Overall, the thematic user’s and producer’s accuracy for aspen was similar for the two sensors.