Interoperability of Landsat and DMC imagery for continuous detection and quantification of nonindustrial forest harvests in the Western Upper Peninsula of Michigan, USA

Document Type

Conference Proceeding

Publication Date



The relationship between human land use and land cover change is critical to sustainable forest management. Land use decisions by small land managers aggregate into substantial land cover changes at landscape and regional scales. Land ownership across large portions of the Upper Great Lakes region is in considerable flux, as large timber industry tracts are split into many smaller non-industrial ownerships, and new owners prioritize amenity and non-timber forest values. Nonindustrial Private Forest (NIPF) owners also transfer their properties to younger generations or other NIPF owners with different management approaches and goals. Survey data on intended harvests and sales are available through the National Woodland Owner Survey (NWOS), run by the USDA Forest Service. However, the disparity between NIPF owner-stated plans to harvest, and what actually occurs, can be substantially different, especially if annual fluctuations in timber prices or general economic fluctuations cause NIPF owners to deviate from their stated management and ownership intentions. This reduces the NWOS' utility. Remote sensing data have considerable value for identifying small scale harvests and, paired with ownership data at the parcel scale, can measure NIPF harvest rates as related to ownership change at a regional scale. Here we focus on the Western Upper Peninsula of Michigan (WUP) and the most recent decade to develop our methodology, using primarily Landsat images from 2003-2013. However, Landsat data series are characterized by gaps in coverage over long temporal and large spatial scales, and so a methodology to combine multiple remote sensing data sources is necessary for regional-scale land use/land cover change research. We filled these gaps by integrating the available Landsat time series with DMC imagery. We then combined these data with GIS overlays of the parcels and stand-level data on removed basal area (BA) during known harvesting events to develop a classification of harvest intensity for the WUP. Images taken during peak growing season were preferred to calculate NDVI and ΔNDVI, and in general for enhancing possible spectral changes. We classified the harvests as clear cut, selective harvesting or thinning using an object-based image analysis. In particular, we defined a clear cut a harvesting event in which ~90-100% BA is removed, commercial harvesting if ~50-80% BA is removed and thinning if ~20-40% BA removal. This work demonstrates that DMC images can effectively fill the Landsat data gap for the detection and quantification of harvesting events. Preliminary results show that the method is capable of identifying harvests down to ~20% BA removal. These results can then be used to monitor the accuracy of the NWOS, and to develop a probability estimate of harvest given either ownership change or changes in market conditions.

Publisher's Statement

© 2013 [Copyright Holder]. Publisher’s version of record: http://adsabs.harvard.edu/abs/2013AGUFMGC43A1029M

Publication Title

American Geophysical Union Annual Meeting