To enable coordinated soaring by a flock of small unmanned air vehicles a method for distributed mapping of the wind field is presented. Presently the work focuses on mapping convective activity (i.e. thermals). The map first discretizes the environment into cells and then uses a Kalman filter to estimate the vertical wind speed and associated covariance in each cell. To improve computational tractability, the wind speed in each cell is assumed to be uncorrelated to all other cells; this results in a set of independent scalar Kalman filters. Measurements of wind speed are available at the location of an aircraft; thermal dynamics are modeled using an exponential decay. The resulting map is combined with a behavior-based controller to enable autonomous soaring. The covariance of wind speed is used to drive exploration and a combination of estimated wind speed and covariance is used to drive exploitation. The utility of the approach is demonstrated using Monte Carlo simulations of a persistent presence task: a flock of UAVs flies in a four square kilometer region and attempts to maximize endurance. Only gliding flight is assumed; with one aircraft the use of the map doubles endurance compared with a no-map case. Increasing flock size to two, four, and eight aircraft results in monotonically increasing performance, with almost all of the eight-aircraft flocks able to remain aloft for the full mission duration.