The forests of western North America are affected by root diseases caused by several endemic fungi. These have both important economical and ecological impacts. Phellinus weirii (laminated root rot) is particularly important in coastal Douglas fir forests. Forest managers would like to know the location of pockets of Phellinus weirii infected trees for the purpose of salvage, remedial activities and inventory. Airborne multi-spectral imagers, coupled with automated detection of damaged trees have potential to provide a cost-effective survey method. Two sets of Compact Airborne Spectrographic Imager (CASI) airborne multi-spectral imagery were acquired at 60 cm resolution over the same Douglas fir dominated site in coastal British Columbia, Canada. They were acquired in successive years and radiometric corrections for the effects of illumination and view angle applied. Trees of varying levels of root rot symptoms were assessed in the field and manually delineated on the imagery. Spectral properties of these trees were related to levels of damage symptoms. There was considerable overlap of the spectral signatures of the different damage levels, especially healthy through moderate. The range of reflectances for healthy trees was large. The near-infrared and red bands and band ratio involving those two bands proved most related to root rot damage. A blue band was also useful, as were ratios of the near-infrared or red bands to the blue band. Classification of these trees using the best combination of four spectral bands indicated average class accuracies in the order of 55–60% for healthy, light-healthy, light, moderate, severe, 100% needle loss, snag and shadowed snag classes. There was important confusion among the moderate through to healthy class. However, these classes are a finer categorization than is necessary for most applications. Accuracy for broader classes was much better (e.g. average class accuracies were 82% if a tolerance of ±1 class was permitted, ranging from 50–100% for individual classes). An automated tree isolation method was applied to the data. This automated tree isolation was good for the 1995 data but suffered from splitting of large trees into several segments on the 1996 data. All but one of the ground reference trees had associated automatically isolated tree crowns. Classification of the isolations corresponding well to ground reference trees was similar to accuracies for the manually delineated trees, but poorer if ground reference trees without a good matched isolation are considered an error (42% and in the order of 60% with a ±1 class tolerance). The overall distribution of root rot damaged trees as indicated by the automated tree isolation and classification was spot checked throughout the site. There was a generally good correspondence, with concentrations of moderate and severe damage trees being associated with areas of root disease. Concentrations of predominantly light damage trees were not a reliable indication of root disease, and forest regions where the main symptoms of root disease are light will be difficult to survey. Some damage zones occurred that seemed to be related to poor health but not specifically related to root disease. As well, isolated trees with similar characteristics as laminated root rot infected trees do appear on the imagery in scattered locations unrelated to root disease activity. It is felt that these false alarms can be largely mitigated by identifying the characteristic pattern of root damaged trees (i.e. stressed trees around a centre, the centre often being a hole or gap in the canopy). High resolution multi-spectral imagery combined with automated procedures seems viable for detecting laminated root rot centres when severe symptoms are present.