This study develops two models and an algorithm to determine the optimal bund wall height of the paddy field and the trigger depth for re-irrigation, considering the benefits and costs of the transformation from shallow intermittent irrigation treatment to deepwater ponding irrigation. The models and algorithms are composed of: (i) a hydrologic model for ponding depth simulation in a paddy field; (ii) an optimization model for height design and depth determination in deepwater ponding irrigation; (iii) a real-coded genetic algorithm for solving the optimization model. The results show that the optimal bund wall height of the paddy field is 27.7 cm and the optimal discount ratio of reirrigation depth (i.e. trigger depth divided by the upper limit of ponding depth) is 0.008 in Taiwan, with a unit net benefit of US$18 500 ha‾¹. During the first crop-growing season (January to May) and the second (June to October), the benefit of groundwater recharge, flood mitigation, saved irrigation quantity and income of paddy rice of deepwater ponding irrigation is increased by 73, 13, 53 and 25% than those of shallow intermittent irrigation treatment, respectively. Furthermore, the total benefit of the second season is 16% better than that of the first. K E Y W O R D S deepwater ponding irrigation, paddy rice field, optimal bund wall height, trigger depth for reirrigation, genetic algorithm, tabu search Résumé Cette étude développe deux modèles et un algorithme pour déterminer la hau