Helen Macdonald¹, Carine G. R Costa², Rafael Santana^1,3, Joao M. A. C.  Souza, Joe O’Callaghan

¹National Institute of Water and Atmospheric Research, , New Zealand, ²Metocean Solutions, Part Of Metservice New Zealand, , New Zealand, ³University of Auckland, , New Zealand

In the Aotearoa New Zealand (NZ) oceanic regions, the number of locally observed oceanic variables is small, creating a barrier to developing data assimilative models. Here we use 4D-var within the Regional Ocean Modelling system configured for the northeast shelf region of the NZ north island to investigate what can be done with small amounts of data. We performed 2 sets of experiments: 1) investigating the minimum amount of data needed for a coastal data assimilating model and, 2) Investigating the effect of concentrating the available observations resources to produce short forecasts.  In the first set of experiments, hindcast model which assimilates locally collected mooring data was compared to hindcast models with only assimilate global products (e.g. satellite SST).  It was found that assimilating subsurface temperature and salinity data does correct the model’s representation of the thermocline but has a negative effect on the model’s velocity field. However, having shorter assimilation windows can overcome this negative effect. In the second section we implement a proof-of-concept forecast system with data assimilation in the northeast shelf of New Zealand. Here we deploy an automatous underwater glider to temporarily concentrate data in the focus region. Results show that the assimilation of glider data improves the predicted temperature (RMSE 1.8°C when assimilating glider data versus 2.83°C when not assimilating glider data). The results suggest that when data availability is low, the ideal model/data setup could change depending on the intended application. Using movable gliders can produce enough data in one location for a short period of time and for some applications. However, consistent long-term observations of surface and subsurface density and velocity are advised if NZ is to increase the quality of longer-running data assimilating models needed for a large range of applications.

Biography:

Helen Macdonald is a scientist at the National Institute of Water and Atmospheric Research (NIWA) in Wellington, New Zealand. She specialises in regional and coastal ocean modelling, usually resolving the 1-20 km scale. She has worked on a variety of ocean modelling projects including lower food web dynamics (NPZD-type models), marine connectivity, western boundary currents, data assimilation and regions of freshwater input.