Dynamical downscaling of General Circulation Model (GCM) data for any region has been made possible due to a set of physics options and model dynamics within the Weather Research and Forecasting (WRF) model. This study evaluated the performance of an ensemble of physics options in simulating rainfall during wet and dry seasons of Lao PDR. The model evaluation criteria focused on identifying the optimum physics options for a range of scenarios. No single combination of physics options performed well in all scenarios reflecting the importance of using different parameterizations according to the geographic location and the intended application of the results. For the dry season, none of the ensemble members performed satisfactorily for the southern region of Lao PDR, while all the ensemble members performed well for the northern and central regions. While almost all the WRF simulations overestimated the rainfall during the wet season, BMJ for cumulus physics performed better in the northern and central regions, and KF performed better in the south region. The YSU scheme performed best as the planetary boundary layer for both wet and dry seasons, while WSM5 for the wet season and Lin for the dry season gave the best model performance as the microphysics option.
Streamflow prediction is a challenging but important task as it affects many processes in the environment. In Australia, the task is particularly complex because the transition from low to high flows can be rapid, and the overall variability is high. Accurate water resource management in Australia requires models that can accommodate these unique characteristics. Complicating matters further, streamflow is not directly measured but inferred from stage height measurements using rating curves. To address the challenges posed by the unique Australian hydrological features mentioned above, the present work proposes a model that: (a) directly predicts stage height rather than streamflow, (b) is sufficiently complex to learn the high variability of stage heights and (c) quantifies epistemic and aleatoric uncertainty by using a Bayesian framework. Our model is a Bayesian Hierarchical Mixture of Experts (BHME) model, with two mixture components. The two components are gamma densities in which the mean and variance are parameterized to depend on upstream stage height and rainfall from the previous day. The mixture weights are also parameterized to depend on the previous day's upstream stage height and rainfall through a logit link function. The method was tested on three gauging stations in the Orara and Richmond rivers. The BHME model provides better fit to the data and better predictive densities compared to a single component model. Notably, the BHME model excels in forecasting less frequent, high stage heights, crucial for managing high flow events. Additionally, transitions between mixture components offer insights into sudden stage height regime changes.
Water resource management is impossible without accurate information about stream discharge. However measuring stream discharge is difficult and therefore stream height measurements, which are easier to obtain, are used as proxies. The conversion of stream height to stream discharge relies on a stage-discharge relationship, which is unique to individual gauging stations and is known as a ‘rating curve’. Accurate assessment of this rating curve and accompanying predictions, is essential for effective water resources management. However, the stage-discharge relationship is often non-stationary, due to erosion or sediment deposition at gauging sites which results from natural processes such as flooding. As a result, the rating curve needs to be re-calibrated regularly. This paper present an approach to estimate the time varying stage-discharge relationships, which we call ‘AdaptRatin’. ‘AdaptRatin' first partitions the time-ordered data into an unknown yet finite number of segments, which are locally stationary. Within each segment, the stage-discharge relationship is modelled non-parametrically by placing a Gaussian process prior over the unknown relationship. We take a Bayesian approach and inference regarding the number and location of locally stationary segments and the corresponding rating curve for each segment is made via the joint posterior distribution of these quantities. We use Reversible Jump Markov Chain Monte Carlo (RJMCMC) to obtain a sample based estimate of this joint posterior. We demonstrate the ability of ‘AdaptRatin’ to successfully capture the changes in the stage-discharge relationship over time and provide reliable estimates of the underlying stage-discharge relationship for each time period for both stationary and non-stationary processes.