A newly published mathematical model suggests that the United States government could increase its chances of catching known criminals at ports of entry by adjusting its biometric identification system to compensate for varying fingerprint image quality.

Visitors now provide two index-finger fingerprints upon arrival, which are compared with database images from a watchlist of about 6 million criminals and suspected terrorists.

Lawrence Wein and Manas Baveja explored two alternative identification systems by using published fingerprint data together with a mathematical model that accounts for difficulties in matching poor-quality images.

The authors' model suggests that under existing policies, the probability of the current system matching an illegal visitor's poor-quality prints to the correct database prints is 52.6%. If the requirement for a match under these conditions is allowed to vary with image quality, however, the model predicts a detection probability of 73.3%. If more than two fingerprints can be used to match visitors with poor image quality, the detection probability increases to 94.9%.

The authors' model may thus address situations where individuals have fingerprints of poor image quality--for example, those deliberately altered through surgery, chemicals, or abrasion.

Land Use Impacts Earth's Water Balance

Human modification of Earth's water cycle through deforestation and irrigation may have significant regional impact on weather patterns and food production, researchers report.

While agricultural practices are known to alter liquid water flow across Earth's surface, human-induced alteration of water vapor flow has received considerably less attention.

By comparing current land use patterns with Earth's pre-human vegetation cover, Line Gordon and colleagues estimated global vapor flow changes due to deforest
'"/>

Contact: Leikny Johnson
PNASNews@nas.edu
202-334-1310
Proceedings of the National Academy of Sciences (PNAS)
9-May-2005