We all know that the risks we face depend not just on who we are but where we are. California is the most earthquake-prone state in the continental United States but, even within California, the hazard is far from uniform. San Diego is at scant danger compared with the San Francisco Bay area, for example. And within the Bay area itself, your fate during an earthquake depends on the type of structure you’re in and whether tectonic convulsions can liquify the ground beneath you.

“Wouldn’t it be nice,” asks Mark Monmonier in the preface to Cartographies of Danger, “to have no-go, no-build, or no-live maps for all kinds of nasty surprises?” He answers his own question with a book full of well-presented maps. Those maps presented the author with an immediate danger-namely, that by the third chapter, the readers would start skimming at high speed. Fortunately, thanks to Monmonier’s lucid and stimulating narrative, I never had such a reaction.

The author makes clear early on that cartography is essential for discussing how physical hazards correlate with geography. Only a small number of the risks we face actually do show such a correlation, however. Part of why the book stays interesting is that, with the author’s subtle help, we realize that some maps that claim to delineate hazards could actually do more harm than good.

Even though some maps appear to provide important information, on closer inspection they provide the reader with nothing useful. Consider “Estimated Potential for Contamination of the Intermediate Aquifer System in Polk County, Florida,” which divides the county into three parts based on a low, moderate, or high degree of hazard. I have no idea what health risks, if any, arise from such contamination; nor am I aware of why it seemed prudent to speak of three discrete categories of risk rather than two or a continuous measure of risk.

Similarly, maps that pinpoint the locations of recent crimes imply that they can tell us where we cannot go safely. But the value of such information is questionable. In large cities, for example, the local police allocate patrols according to these maps. If drug deals are known to occur on a certain street, the police may increase their presence there. This may, however, prompt the dealers to move to another less-policed street. Moreover, sheer counts of the number of crimes at various places may be poor proxies for actual risk levels. Consider New York’s Central Park, where few people venture after dark. If no one enters the park at night and the police therefore record no nocturnal muggings, it does not follow that I can stroll there safely at midnight.

We must also be cautious in interpreting maps that claim to link health risks with geography. Monmonier observes, “In real epidemiology, maps often raise more questions than they answer.” Consider a map on age-adjusted coronary death rates for white males. If the rate is unusually high in northern Florida, for example, that circumstance does not necessarily indicate lethal spinoffs from the Daytona speedway. Rather, it might reflect high-cholesterol diets in the region, in which case a “lean-cuisine” eater who moves to the area would be totally unaffected by the overall statistics. The geographic pattern, in other words, might be the starting point for analysis of a risk that ultimately has nothing to do with geography per se. The associated map might be helpful but viewers had better beware of making spurious inferences that “geography is destiny.”

Readers should also be wary of maps that extrapolate geographic tendencies from rare occurrences in nature. In a “threat rating” map by the National Oceanic and Atmospheric Administration based on 17 years of tornado data, Massachusetts ranks as the most dangerous state with a score of 347 while adjacent Rhode Island is assigned a score of zero. But concluding that Massachusetts residents can escape tornado dangers by crossing the (utterly flat) border to Rhode Island is obviously absurd. What drives the reported pattern is one tragic tornado that struck the Worcester area four decades ago. If the tornado had touched down a few dozen miles further south near Providence, R.I., the threat ratings would have been turned upside down.

By the end of this book, we reach an unexpected conclusion: we must handle maps about hazards with care, for a substantial number are potentially misleading as guides to where we should go. The adage that a picture is worth a thousand words may be true here in a perverse way: before we use a cartography of danger as a basis for restricting our movements, we had better read about what the map purports to show. Of all the reasons to restrict our movement, sheer geography may be among the weakest.