Short URL

Scalable Forest Monitoring

 

Other Remote Sensing areas:

Loading, please wait...

Scale in Remote Sensing? 

In prefacing the first issue of Remote Sensing of Environment, Simonett (1969, p. v) noted that "[the] quickening of science, and resource use, and the demands of society have brought new urgencies to obtain quantitative, timely information about the environment at a variety of scales in space and time."  Nearly 40 years later, an increasingly common question in remote sensing may be formulated as follows: "Which remote sensor-derived scale(s) in space and time can supply timely and practical quantitative information about environmental variable x?".

 
Figure: Experimental instrumentation-based monitoring of forest environments by CAST-affiliated students a) with an in situ spectroradiometer, b) with various portable sensors from the vantage of a 17 m articulating boom, c) with a near-infrared aerial imaging system mounted in a fixed wing aircraft, and d) in situ photograph of an "old growth" post oak (Quercus stellata).

The forest monitoring domain offers an excellent foundation for developing geomatics principles of scale management.  Data collected at a variety of scales (see figure above) tends to get fragmented because of the propensity for choosing a favorite scale a priori.  Foresters naturally prefer aerial photography (at relatively high spatial resolution) because of the ability to identify familiar objects within the instantaneous field of view (IFOV).  Unfortunately, the spectral diversity in aerial photography is extremely limited and computational costs associated with automatically processing and delivering high spatial resolution datasets are high.

CAST and Forest Monitoring  

In collaboration with Fred Stephen of the University of Arkansas Forest Entomology Lab, CAST staff have participated in a project in forest monitoring investigating the underlying cause of oak forest vulnerability to the red oak borer (Enaphalodes rufulus) beetle.  When red oak borer populations spiked in 1999 to levels never before seen, more than 8,000 hectares were estimated to have experienced severe damage (including oak mortality).  A multi-year U.S. Forest Service-funded study conducted by the Department of Entomology, CAST, and Geosciences has begun applying a variety of remote sensing techniques to monitor red oak borer damage and to determine why some areas were affected more than others (see Photo by Chris Steiner, USDA Forest Service, October 3, 2000).  Students from various departments on campus as well as visiting undergraduate researchers have been active in collecting field data related to this remote sensing effort.  More than 80 MB of field data in numeric and text tables have been compiled into a personal database with a number of PhD candidates applying this data in a remote sensing context.  One highlight is that 70 million relatively high density LIDAR points were acquired in 2006 over a 32 x 32 km2 study to further investigate vertical canopy structure; forest structural information content in the LIDAR that may pertain to oak decline is now being studied by CAST-affiliated PhD students in both Entomology and Environmental Dynamics.

A key for more efficient remote sensing-assisted decision support in such examples as forest monitoring to understand the red oak borer is scale management (e.g. tools for optimizing spatial, spectral, temporal, and/or radiometric scales within data-to-decision pathways).  Coordinating expert and machine learning knowledge with an enterprise database (see related figure) is one way to approach the scale management dilemma, thus enabling queries for "shortcuts" in the remote sensing process.