Un réseau de surveillance efficace est souvent un outil utile pour surveiller et évaluer les différents états d'un écosystème spécifique. Avec les informations fournies par le réseau, les prédictions peuvent être générés en utilisant des analyses spatio-temporelles approfondies, qui peuvent ensuite aider les décideurs et les parties prenantes. Écosystèmes dominés par les humains sont très dynamique et complexe, où la plupart des variables observées ont des interactions non-linéaires mutuelles. En outre, les impacts des activités humaines sur l'écosystème sont souvent rugueux, dans de nombreux cas, ils perturbent l'équilibre écologique. Par conséquent, la surveillance d'un écosystème devient un problème complexe et il n'est pas facile de proposer un réseau de surveillance optimal en utilisant les techniques traditionnelles. Les techniques d'optimisation traditionnelles ne parviennent pas souvent à tenir compte des évolutions de la réalité associée à ces écosystèmes.Un exemple d'une telle situation se trouve à la région du delta du Mékong du Vietnam, où les gestionnaires provinciaux de l'agriculture sont concernés par les invasions régulières des cicadelles brunes (Brown Plant Hoppers - BPH), un ravageur du riz particulièrement active, à cause des maladies qu'ils véhiculent et transmettent à la les rendements de riz. / An efficient surveillance network is an invaluable tool to monitor and assess the different states of a given ecosystem. With the information collected by such a network, predictions can be generated using thorough spatio-temporal analyses, which can then support decision makers and stakeholders. Human dominated ecosystems are highly dynamic and complex , where most of the observed variables have mutual non-linear interactions. In addition, the human activities have a considerable impact on almost all ecosystems they inhabit, where they tend to disrupt the ecological balance in short period of times. The surveillance of such ecosystems by different technical solutions is complex and dynamic where designing an “optimal” surveillance network, i.e., a network that would reflect an almost realtime situation of an ecosystem. Often traditional optimization techniques fail to reflect the evolutions of the reality associated with these ecosystems.An example of such a situation, is the Mekong Delta region of Vietnam, where the provincial agricultural managers are concerned with the regular invasions of Brown Plant Hoppers (BPH), a particularly active rice pest, because of the diseases they carry and transmit to the rice yields. Their biggest concern is having a constantly accurate account of the current distribution of BPH waves, since it is the basis of establishing different prevention strategies. The time frame is short for applying these strategies: at least one week is needed between the moment where a prediction of the density of BPH can be estimated by the experts and warnings are sent to farmers and other end users and the moment where a strategy can be efficiently applied. To improve the efficiency of the system, the Vietnamese government has established a light-trap network [1] [2] that can capture multiple kinds of insects, especially BPH, and which data (the density of insects per trap) is collected and analyzed daily. Maintaining this network in a good state of operation has become an important national program of the Ministry of Agriculture and Rural Development of Vietnam since 2006. Although the current light-trap network is considered as a necessity for supporting the fight against various plant pests, it has three restrictions: (1) it misses detailed accounts on the life cycle of the BPH, (2) the network has remained stable since its initial design and has not, therefore, completely adapted to the huge changes that the ecosystem of the Mekong Delta has undergone in the recent years especially due to limitations in management, and (3) the network itself is very sparsely distributed.