Surface forcing of the North Atlantic: accuracy and variability

A new methodology to estimate the turbulent air – sea heat and moisture fluxes and<br/>their uncertainty is developed and assessed using Voluntary Observing Ship (VOS)<br/>observations. Whilst important drivers of the global oceanic and atmospheric circulation<br/>these fluxes remain poorly quantified, both in terms of mean value and uncertainty. The<br/>new methodology addresses both of these issues and is extensible to other data sources.<br/>The individual observations are first bias and height adjusted to remove systematic<br/>errors and the impact of changing observing heights. They are then characterised in<br/>terms of random errors using a semi-variogram analysis and a range of variogram<br/>models. The data quality and sampling are then taken into account using optimal<br/>interpolation (OI) to grid the observations, producing daily mean fields and uncertainty<br/>estimates. These are then used to estimate the fluxes and flux uncertainty on both daily<br/>and monthly time scales.<br/>Comparisons of the mean fields and fluxes to the original input data and to<br/>independent buoy observations show the fields not to be significantly biased. The<br/>adjustments applied before gridding and flux calculation are also shown to improve the<br/>agreement with the buoy observations. The uncertainty estimates are assessed using a<br/>series of cross validation experiments and 3-way error analyses to make alternative<br/>estimates of the uncertainty. These alternative estimates are shown to be of the same<br/>order of magnitude as the OI uncertainty estimates and generally to be within 10 – 20%<br/>of the OI estimate. Whilst all three estimates are similar there are some systematic<br/>differences. The OI uncertainty estimates tend to be lower (higher) than the alternative<br/>estimates in high (low) variability regions.<br/>The representation of the variability in the new dataset is examined and shown to be<br/>improved compared to previous VOS based datasets. The adjustments are shown to<br/>have little impact on the temporal trends in temperature and humidity whilst reducing<br/>the wind speed and sensible and latent heat flux trends. These reduced trends are<br/>thought to be more realistic. The wind speed trend after adjustment is more similar to<br/>the trends reported in previous studies using reanalysis model output. However, there<br/>are still some differences in the trends, with the VOS based estimates larger, leading to<br/>uncertainty in trend estimates. The trends in the adjusted latent and sensible heat flux<br/>estimates are similar to those seen in other flux datasets but when compared to changes<br/>in the upper ocean heat content may still be too large. This may be due to the<br/>overestimate of the wind speed trend. Overall the uncertainty in the wind speed trend<br/>gives the largest uncertainty in the flux trends.<br/>Finally, the advances made in developing the new methodology are summarised and<br/>the potential uses of the new dataset identified. Future work and improvements are then<br/>suggested.

Name dc:type.qualificationname

Ph.D.

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

University of Southampton

Year dc:date.issued

2009

Author dc:creator

• Berry, David Inglis