Tracking and Targeting Illegal Deforestation
Decision support challenge:
To communicate priorities on tropical deforestation to the United Nations Office on Drug and Crime (UNODC) for investigating illegal logging, mining, or illicit crop cultivation.
Decision makers or end users:
Field offices in Colombia, Peru, Bolivia, Myanmar and Laos, where UNODC tracks cultivation of illicit crops coca — the raw material used to process cocaine — and opium poppy. The national governments and law enforcement branches of countries cooperating with UNODC.
Research challenges and data skills relating to decision-making process:
Currently, UNODC field offices track illicit crops and environmental crimes through annual surveys of remote sensing imagery combined with targeted aerial surveys and site visits for validation. Machine learning techniques are deployed to expedite initial analyses, but all decisions are made based on information hand-curated by a few experienced technicians. The annual surveys are unresponsive to the need for immediate action, and decision making currently occurs without modeling the impact of particular actions. Developing real-time tools to quickly rank hotspots of land-use change for further investigation will both increase the performance and reduce the costs of flights and field visits. It will also enhance the ability of national governments to prevent illegal deforestation. Using spatial data on illicit crops, illegal logging and mining, bioclimatic variables, and remote sensing imagery, our multidisciplinary research will produce the tools to track and forecast forest loss. More importantly, building interactive visual scenarios will support interactive target prioritization.
Data science skills necessary to improve decision making:
First, data scientists need to effectively communicate with a law-enforcement audience using non-technical language. The UNODC constituents uphold the Single Convention on Narcotic Drugs and must adhere in their decisions to its protocols, amendments and subsequent agreements. Second, it is critical to communicate uncertainty regarding models and make explicit the assumptions underlying different scenarios. This work bridges remote sensing, high-performance computing, and economic geography. The urgent need for improved tools provides excellent motivation and impact for interdisciplinary students.
Assessing improved decision making:
The effectiveness of improved decision making can be assessed by quantifying the performance of flights and visits. Another metric is the change in deforestation rates over time, as decision making shifts from its current format to real-time, interactive computational tools.