Roads are used by humans for travel and by wildlife to cross to the other side, leading to wildlife vehicle conflict (WVC), a problem that results in high costs to wildlife and society. The California Roadkill Observation System (CROS) was launched in August 2009 as a proof of concept for a US-state scale WVC-reporting system.  

CROS includes 212,860 observations collected through multiple platforms. The largest number of observations of a species (n=49,438, 23% of total) is for mule deer (Odocoileus hemionus). The remaining observations are for 501 other bird, mammal, and herpetofauna species and for 36 species groups, such as “squirrel”, where the exact species was not known. Over 92% of the observations were recorded after the first web-instance was created in 2009, with the other 8% coming from historic records maintained by individuals and agencies.

WVC observations have come from a combination of sources, including amateur observers; state, federal, and local agencies; private entities; and state highway patrol officers. We have found that observers are spatially accurate (<13 m, median error), and taxonomically correct (>97% accurate species identification rate), with no difference among professional and amateur observers. Because every roadkill observation is of a species at a certain time and place, CROS observation data have been used to model habitat suitability and species distributions. For many species in CA, as much as half of available observation data are roadkill observations from our system.

Every year, we use these data to locate high-density and statistically-significant clusters of WVC and to estimate the economic costs of WVC using fiscal equivalents for crashes from FHWA and compare that to the costs to build fencing and/or other mitigation actions. CROS data have been used on dozens of wildlife planning projects throughout the state, including on highways 1, 5, 8, 10, 12, 16, 17, 20, 50, 58, 62, 84, 97, 101, 152, 156, 395, 580, and 680. 

Over the last 15 years developing and using this system we have learned several important lessons: 1) consistent and generalizable data formatting is critical for data sharing; 2) people can and will share roadkill observations through phone and desktop platforms; 3) many entities expect to receive our data even in the absence of sharing their own; 4) continuous funding for long-term monitoring of roadkill is virtually impossible; 5) long-term investment in this type of system results in very large datasets where sheer size and longevity have qualities all their own. For example, we can now estimate absolute rates of roadkill of several species in the state, at fine scales. We are working on including these data through a new transportation data-sharing portal, funded by the Wildlife Conservation Network.