Each rocky shore is an ecosystem where it is critical to understand, protect and preserve biodiver- sity in the context of climate changes. Efficient autonomous observatories can be developed and deployed to collect data with benefits for environment management, thanks to compact electronic devices and new radio systems. Wireless sensor networks (WSN) offer attractive solutions for observation because of the large availability of physical sensor devices, and compact radio link transceivers, for example Modtronix inAIR9b in Fig. 1 [3]. Radio links enable communications between sensors, gateways, and informa- tion systems on permanent or periodic basis [4]. In the case of marine environments such as shores or islands, it is practically difficult to design radio sensor deployments having an accurate cover- age. Coverages are the way to specify where information is accessible, whatever it is either radio signal, biological, or physical influences. Therefore computing coverage is essential for monitoring purposes and can take various forms. The case of marine shores is difficult because of complex topologies, with rocks of various shape and elevations interleaved with the sea. Managing the sea from radio buoys is another element. This paper relates on a work principles and results oriented to exploration of radio link capabil- ities, taking into account sea shore geography topologies. By modeling geography into cell systems, it becomes possible to simulate many physical phenomena such as wave effects, rain effects, pollution, etc. It is also possible to model the natural behavior of WSN radio signal propagation. This define the ability of signals coming from one point to reach another point (Line of Sight, LoS). Computing covers in presence of obstacles is known to be a compute intensive task, in the complexity class of 3D image synthesis because emitted rays must be compared to each other point in an image taking obstacles into account. We explain and discuss a massive parallel execution led on cellular systems representing the geographic zone. Computations were firstly achieved on communicating processes suitable for multicore processors, then they were ported on Graphics Processing Units (GPU) producing performances in the real time order. The tools developed allow to select arbitrary study zones from a specific map browser called QuickMap 1 . As a result, automatic and manual coverage computations were applied to a set of archipelagoes. Practical results are given in terms of performances and functional results section 2.