Using police seizure data and hyperspectral imagery to estimate the size of an outdoor cannabis industry

Abstract

In this paper, we take advantage of aerial police detection data and compare them to an analysis of airborne hyperspectral imagery to estimate the size of outdoor cannabis production in a high density cultivation area located in British Columbia, Canada. The results suggest that law enforcement missed at least 75% of the existing cultivation sites during the operation. Based on this ratio, the total number of cultivation sites in the island/coastal region is estimated at 1772 sites, for a total of 9563–14,663 kg of marketable product. Conservative estimates of street value easily put cannabis as the number‐one cash crop in the region under study. Implications for law enforcement agencies interested about the efficiency of their eradication programs are discussed.

Keywords:

• cannabis
• cannabis cultivation
• drug law enforcement
• estimation methods
• risks of detection

Acknowledgments

Both authors had an equal contribution to the paper (the order of authors was randomly allocated). The authors would like to thank Jon Caulkins, Pierre Tremblay, Eric Beauregard, and the anonymous reviewers for the helpful comments provided on an earlier version of this paper. The authors also would like to thank the Canadian Police Research Centre and the Vancouver Island RCMP detachment who kindly provided the data necessary for this study. An additional thank you to Eva Snirer for her help in compiling background information for the study.

Notes

1. Satellites Pour l’Observation de la Terre (Satellites for Earth Observation).

2. Quality images could not be reproduced for the purpose of this paper, but are available from the corresponding author upon request.

3. Because there was no field verification conducted it is not possible to ascertain for sure what the exact error rate is.

4. No specific reasons explain why our data set only contained specific plant data on 132 out of 443 sites, and we assume our subsample is a good approximation of the total site population. The average number of plants found per case for our subsample (126.8 plants) is very similar to the one found for all cases detected in that region for the 1998–2003 period (98–134 plants, see Plecas et al., 2005).