Northeast Thailand has the largest rainfed lowland rice (RLR) ecosystem in the kingdom and is notoriously known for its high rate of poor smallholders. The unstable rice productivity as a consequence of an unfavourable ecological environment (erratic rainfall and infertile soils) interacting with low price of rice drives these poor people to migrate for more profitable employment leaving often their land and its water underused. During the last 15 years, small water resource improvement programs launched by the Thai government under its poverty alleviation agenda had limited success. Labour migration is an adaptive strategy to cope with the uncertainty of rainfall and its distribution. As a consequence, off-farm employment becomes a more and more important source of income. But the relationship between labour migrations and land and water management on the farms is still poorly documented. Therefore, we used the Companion Modelling (ComMod) approach to improve the understanding of this key interaction and to reinforce stakeholders' adaptive capacity to deal with uncertainty linked to water dynamics and labour management in the Lam Dome Yai watershed of Ubon Ratchathani Province. ComMod facilitates dialogue, shared learning, and collective decision-making to strengthen the adaptive management capacity of local communities through integrative collaborative modelling. The cyclic ComMod process is made of iterative loops comprising field investigations, modelling, and participatory simulations relying on the combinations of Role-Playing Games (RPG) and Agent-Based Models (ABM) used with stakeholders. In this case study, 5 ComMod loops were carried out to better understand the problem being examined, stimulate exchange of points of view and enhance the creativity of the participants while lessening the black box effect of computer models. The key processes embedded in the models are based on stakeholders' decision-making driven by human-environment interactions. We take