Wednesday, February 25, 2009

Using Crappy Geographic Methods to Not Find Osama bin Laden

BOULDER, Colorado -- Two researchers at the UCLA Department of Geography say they have found Osama bin Laden's hiding place, somewhere in the Tribal Areas of Pakistan along the Afghanistan border. But will this supposed discovery be of actual use to anyone?

It has long been assumed that bin Laden was holed up somewhere in this area - his last known whereabouts was just over the frontier in the caves of Tora Bora, way back in 2002 - but professors Tom Gillespie and John Agnew have narrowed it down to three possible buildings in the town of Parachinar.

Eeny, meeny, miney, moe ...

The researchers used Geographic Information Systems to crunch data about the local terrain and population, and about bin Laden himself. They then employed two biogeographic models normally used to explain and predict the distribution of plant and animal populations. Distance-decay theory, which states that the farther one moves away from a given location, the less likely one will find the same distribution of species as they did at the first (the species here being what, terrorists? Members of the bin Laden family?) Island biogeography is also a straightforward biology model - the bigger the island and the closer it is to other landforms, the great diversity of species that it will support.

Finally, the researchers used bin Laden's "life history characteristics," also a practice taken from biology. They would need to identify a city and a structure that fulfilled certain requirements, like having enough rooms to house his entourage, and with access to electricity to power his dialysis machine (even though the claims of his serious kidney problems have long since been debunked.) Noah Schachtman at Wired rightly points out one of the biggest problems of this model - "garbage in, garbage out." While perhaps useful as a thought experiment, the authors claim such a high degree of precision that is based on unreliable or false data and assumptions that it makes the whole exercise less than illuminating.

Murtaza Haider in Toronto's Globe and Mail
points out another important shortcoming of the study - the authors have miraculously managed to identify the only majority-Shiite city in the entire tribal region. Ms. Haider is a native of Pakistan's Northwest Frontier Province and finds it highly unlikely that the leader of a global Sunni Islamist terrorist network would choose to hide in the one town where the residents are least likely to provide him with cover. Spatial methods can be incredibly powerfully tools, but applying biological models like these, and relying on aerial photography to crudely identify buildings and landforms will create more bad targets than real intelligence. As Haider worries, and not without justification, this analysis may do more harm than good:
If the gullible UCLA professors are taken seriously by trigger-happy NATO forces, who lack ground intelligence in the tribal areas and cannot tell friend from foe, the Shiites of Parachinar may have to fend off bombs from American drones, while they are fighting for their survival against the Taliban on the ground.
To be helpful, these methods need to be combined with other techniques, like ground-truthing remote sensing data to actually test if the information you have collected tells you what you thought it did. Getting basic facts about bin Laden's biography would also be useful.

John Agnew is a well-respected and widely-read scholar in geography. Recently he has become enamored of spatial methods, but he appears to have little grasp of their nuances and shortcomings. Last year he released a study, along with Gillespie, Jorge Gonzalez and Brian Min, about the effectiveness of the American troop surge in Iraq in 2007. Titled "Baghdad nights: evaluating the US military 'surge' using nighttime light signatures" and published in Envirnment and Planning A (2008, vol. 40, pgs. 2285-2295), the study also relied on remote sensing data, this time employing nighttime light emissions in Baghdad.

Density plot of the occurrence of deaths in Baghdad (Jones Report, 2007).

Comparing light emissions from periods immediately after the US invasion in 2003 through the surge in 2007, they concluded that light emissions had increased until 2006 and then went into a rapid decline, which coincided with the eruption of Iraq's so-called sectarian civil war. Their conclusion is that the surge had no observable effect on violence, but rather that the conclusion of ethnic segregation led to improved security. The areas of decreased light do coincide with areas of violence - people were driven from their homes by ethnic militias, turned off the lights, and never came back.

There is evidence to support this claim, but not from looking at lights from space. It has been widely documented that American forces indirectly facilitated ethnic cleansing in Baghdad by encouraging residents who had been targeted by militias to move to ethnically-homogeneous neighborhoods. Troops have constructed a vast network of blast walls throughout the city to divide Sunnis and Shiites, ostensibly to protect them from inter-ethnic violence. It may be difficult to argue that the increase in American forces had no discernible effect on the security situation, but certainly violence has decreased in part because the militias have already achieved their goals of turning Baghdad into a segregated city. But reaching this conclusion requires analyzing a wide array of data sources, much of it coming from on-the-ground reports and first-person accounts of the violence. There is only so much we can learn about an urban-scale ethnic conflict, or the location of a single man, from space.

John Krygier over at Making Maps: DIY Cartography has commented on the Iraq surge article, and he has been nice enough to post the original.

No comments:

Post a Comment