Estimating the position of a single sound source within the biodiversity monitoring context can be achieved via an estimation of the time difference of arrival (TDOA) between microphone pairs among several single-channel autonomous recording units spatially distributed across a geographic area to be surveyed. When multiple sound sources are simultaneously active, however, each microphone pair estimates multiple TDOAs leading to an ambiguity of which TDOAs from different pairs correspond to the same physical source, i.e., an association problem. Although several methods have been proposed to address this problem, most rely on the availability of a large number of microphone pairs, whereas in biodiversity monitoring, the number of available microphones is limited. In this paper, we propose a method for localizing multiple sound sources based on sub-band TDOA estimations. The approach facilitates the definition of a sub-band feature vector, which characterizes the spectral content of sources, and is used to solve the association problem. This leads to a decomposition of the multiple source localization problem into independent single source localization tasks. Preliminary simulations demonstrate the potential of the method in scenarios with a limited number of microphones and hence its prospect for biodiversity monitoring.