This work pertains to the research program on Containment Thermal-Hydraulicsat KTH. The objective is to evaluate and improve performance of methods, whichare used to analyze thermal-hydraulics of steam suppression pools in a BWRplant under different abnormal transient and accident conditions. As a passivesafety system, the function of steam pressure suppression pools is paramount tothe containment performance. In the present work, the focus is on apparentlybenign but intricate and potentially risk-significant scenarios in which thermalstratification could significantly impede the pool’s pressure suppression capacity.For the case of small flow rates of steam influx, the steam condenses rapidly inthe pool and the hot condensate rises in a narrow plume above the steam injection plane and spreads into a thin layer at the pool’s free surface. When thesteam flow rate increases significantly, momentum introduced by the steam injection and/or periodic expansion and shrink of large steam bubbles due to directcontact condensation can cause breakdown of the stratified layers and lead tomixing of the pool water. Accurate prediction of the pool thermal-hydraulics insuch scenarios presents a computational challenge. Lumped-parameter modelshave no capability to predict temperature distribution of water pool during thermalstratification development. While high-order-accurate CFD (RANS, LES) methodsare not practical due to excessive computing power needed to calculate 3D highRayleigh-number natural circulation flow in long transients. In the present work, amiddle-ground approach is used, namely CFD-like model of the general purposethermal-hydraulic code GOTHIC. Each cell of 3D GOTHIC grid uses lumped parameter volume type closures for modeling of various heat and mass transferprocesses at subgrid scale. We use GOTHIC to simulate POOLEX/PPOOLEXexperiment, in order to (a) quantify errors due to GOTHIC’s physical models andnumerical schemes, and (b) propose necessary improvements in GOTHIC subgrid scale modeling. The study performed on thermal stratification in a water poolindicates that GOTHIC CFD-like model is fit for reactor applications in complexfluid-physics scenarios that avoids both over-simplification (as in single lumpedparameter model) and over-complication (as in CFD models). However, simulation of direct steam injection into a subcooled pool cannot be predicted reliablywith the existing models. Thus we develop “effective heat source” and “effectivemomentum” approaches, and provide feasibility study for the prediction of thermalstratification and mixing in a BWR pressure suppression pool. The results areencouraging and further activity on the development and implementation of theproposed models in GOTHIC is currently underway.