Skip to main content
Licentiate thesis, 2017

Data assimilation in forest inventories at stand level

Ehlers, Sarah

Abstract

Data assimilation (DA) is a potentially interesting method for forestry if new stand level data about forest attributes are made available at short time intervals. DA is a method where an estimate is forecasted by a model and updated when a new measurement is made. A weighted average of the forecast and the measurement is obtained as the new current state, which increases the accuracy of the estimate. In areas like meteorology DA has been successfully applied for a long time. In this case the availability of very frequent satellite data makes it possible to update weather forecasts several times a day and obtain accurate forecasts. Forest inventories in the traditional way, by field campaigns, are expensive and thus provide new data only every 10-20 years. During this long time a lot of changes due to growth, management and disturbances might occur in the forest stands of interest. Thus, old data are discarded when new data are obtained from a new campaign, and the forecasts of the current state are only based on the last measurement. Since many types of remotely sensed data, e.g. from laser scanners, optical satellite sensors, and radars, have become available during recent years, there are now good opportunities to apply DA also in the context of forest inventory. In this thesis I focus on stand level forest inventories. A first theoretical study with simulated data showed that DA has a strong potential to be successfully applied in forestry and increase the accuracy of inventory estimates. However, the second study, the first with empirical data, pointed at problems to obtain equally good results in practice. In the third study, correlated prediction errors were identified as the plausible reason for this. The higher the correlations the less was found to be gained by applying DA. Despite several remaining challenges, the overall conclusion is that DA has a potential to make forest inventories more efficient in the future.

Keywords

Data assimilation, Kalman filter, remote sensing, forest inventory, correlated prediction errors

Published in


ISBN: 978-91-576-9486-7, eISBN: 978-91-576-9487-4
Publisher: Department of Forest Resource Management, Swedish University of Agricultural sciences

Authors' information

Ehlers, Sarah
Swedish University of Agricultural Sciences, Department of Forest Resource Management

UKÄ Subject classification

Forest Science

URI (permanent link to this page)

https://res.slu.se/id/publ/104108