Monday, 1 August 2016

Near real-time monitoring of insect induced defoliation in subalpine birch forests with MODIS derived NDVI

Published Date

August 2016, Vol.181:42–53, doi:10.1016/j.rse.2016.03.040

Title

Near real-time monitoring of insect induced defoliation in subalpine birch forests with MODIS derived NDVI

Author

Per-Ola Olsson^a,^,

Johan Lindström^b

Lars Eklundh^a

aDepartment of Physical Geography and Ecosystem Science, Lund University, Sölvegatan 12, S-223 62 Lund, Sweden
bCentre for Mathematical Sciences, Division of Mathematical Statistics, Lund University, Sölvegatan 18, SE-221 00 Lund, Sweden

Received 13 October 2015. Revised 21 March 2016. Accepted 31 March 2016. Available online 16 April 2016.

Highlights

•
Near real-time monitoring of landscape-scale insect defoliation with MODIS NDVI
•
Monitoring of intra-seasonal insect outbreak dynamics in forests
•
Mapping of within-season refoliation after insect induced defoliation

Abstract

Forestry and nature conservation can benefit from rapid on-line information on forest disturbances, such as insect attacks. This type of information would facilitate timely field studies and enable more rapid counter measures, as well as enable studies of the dynamics of an insect outbreak. In this study we developed a method based on MODIS derived NDVI for near real-time monitoring of insect induced forest defoliation in a subalpine birch forest in northern Sweden. The method is based on deviations from a seasonal trajectory of NDVI representing forest conditions without disturbances. A Kalman filter is applied to handle noise and satellite-derived NDVI observations of low quality, and cumulative sums (CUSUM) of the deviations from the seasonal trajectory representing undisturbed forests are used to detect disturbances. An annual offset of the seasonal trajectory is introduced in CUSUM to handle inter-annual variability in the start of season. Evaluation of the method showed that 74% of the defoliation was detected with a misclassification of undisturbed areas of 39% in MODIS pixels with at least 50% birch forest cover. The ability of the method to detect defoliation can be adjusted to fit the purpose of a study; with a higher threshold applied, 100% of the defoliation in the evaluation data was detected with 56% of the undisturbed areas misclassified as defoliated. The method also facilitates studies of the intra-seasonal temporal dynamics of an insect outbreak, which is a major advantage compared to methods that classifies pixels into undisturbed or defoliated for an entire season. Furthermore, the method can be extended to monitor within-season refoliation after an insect outbreak.