This paper presents a concept for the collaborative distributed acquisition and refinement of geo-related information. The underlying idea is to start with a massive amount of moving sensors which can observe and measure a spatial phenomenon with an unknown, possibly low accuracy. Linking these measurements with a limited number of measuring units with higher order accuracy leads to an information and quality augmentation in the mass sensor data. This is achieved by distributed information integration and processing in a local communication range. The approach will be demonstrated with the example where cars measure rainfall indirectly by the wiper frequencies. The a priori unknown relationship between wiper frequency and rainfall is incrementally determined and refined in the sensor network. For this, neighboring information of both stationary rain gauges of higher accuracy and neighboring cars with their associated measurement accuracy are integrated. In this way, the quality of the measurement units can be enhanced. In the paper the concept for the approach is presented, together with first experiments in a simulation environment. Each sensor is described as an individual agent with certain processing and communication possibilities. The movement of cars is based on given traffic models. Experiments with respect to the dependency of car density, station density and achievable accuracies are presented. Finally, extensions of this approach to other applications are outlined.