Nichol, Caroline

Poulsom, Elizabeth

2021-05-13T16:34:00Z

2021-05-13T16:34:00Z

2020-12-01

Foresters are increasing using remote sensing for landscape scale monitoring of factors affecting forest growth including tree disease and climate change. This study examines use of Sentinel-2 imagery to detect Phytophthora ramorum, on 18 larch sites in Scotland identified from aerial survey, using RGB bands and 14 vegetation indices. At limited sites, P. ramorum infected trees were discernible in RGB and false colour composite imagery. To quantify results, t-test were run to establish statistical significance on differences between mean index values from diseased and healthy regions of interest. On imagery closest by date to the aerial survey, significant differences were observed between diseased and healthy for six indices when grouping sites collectively into diseased and healthy populations. At an individual site basis, 14 indices showed significant differences. Indices were ranked for sensitivity based on number of significant results per index across 12 months. AVI, EVI, IRECI, SAVI and MSI ranked top five, with all showing significant differences between mean diseased and healthy across 12 months at some sites, even during larch senescence. Unsupervised K-means classification of diseased and healthy larch on a subset area using RGB and individual vegetation indices, gave 79% accuracy using SAVI in mid-summer closest to aerial detection. Mid-winter accuracy was poor across all indicators but marginally improved 12 months prior to detection in mid-summer the previous year. The results indicate that Sentinel-2 imagery and derived vegetation indices can differentiate between diseased and healthy larch, particularly at 10m spatial resolution. Predicting P. ramorum onset out with the visible spectrum could potentially be improved by examining the effectiveness of further indices and with validation sets allowing supervised classification. To improve temporal resolution, other systems including LANDSAT-8 or potentially SAR may compliment Sentinel-2 data.

https://hdl.handle.net/1842/37623

http://dx.doi.org/10.7488/era/904

en

The University of Edinburgh

Remote sensing

Sentinel-2

Phytophthora ramorum

larch

disease detection

vegetation indices

Use of Sentinel-2 to detect Phytophthora ramorum on larch trees in Scotland

The use of Sentinel-2 to detect Phytophthora ramorum on larch trees in Scotland

Thesis or Dissertation

Masters

MSc Master of Science