Understanding a ‘wood baskets’ current and potential future wood supply often underpin forest investment strategies. In some countries, public information on plantation area, species and tending regimes compiled from grower surveys are published every five years. The granularity of such data is sufficient to highlight broad regional trends, but for reasons of confidentiality omits spatial boundaries, species and regime information. Such information quickly dates, especially in locations where plantation rotations are short, or where harvest levels are high, or unexpected fire losses occur. In steps satellite data… while often considered as providing a snapshot in time with access to the temporal archive, it becomes something more when linked to routines that track and record change.
In this example, we build a forest description – aka - area by age-class for an Australian asset using only satellite imagery. By implementing the LandTrendr algorithm on a time series of pre-processed Landsat imagery, significant changes in canopy that reflect harvest activities can be extracted (Kennedy et al., 2018). In the following example a breakpoint representing clearfell can clearly be visualized, suggesting an establishment year of the current rotation of 2013.
This process is repeated for every pixel within the area of interest, after which each change event is validated and refined using repeat observations. This process can then be augmented by applying machine learning algorithms to different plantation species of the derived units. In the following example the orange colour represents establishment in 2013.
The resulting dataset represents the current estate at a point in time. The framework offers an attractive solution that allows a discrete assessment of plantation resources – essentially Forest Intelligence.... While in this example, plantations are the subject of interest; it is worth noting that the monitoring framework can be adapted to provide insights that capture and describe a range of land uses changes.
Kennedy, R.E., Yang, Z., Gorelick, N., Braaten, J., Cavalcante, L., Cohen, W.B. and Healey, S., 2018. Implementation of the LandTrendr algorithm on google earth engine. Remote Sensing, 10(5), p.691.