|
Abstract:
|
Accurate estimates of historical changes in sea surface temperatures (SSTs) and their uncertainties are crucial for documenting and understanding climate change. A source of uncertainty not yet quantified in historical SST estimates stems from position errors. In this talk, we will introduce a Bayesian framework for quantifying errors in reported positions and their implications on SST estimates. Our model bases on navigational practices that use dead reckoning to infer ship locations and periodically update with celestial corrections. We apply the model to 943 late-19th-century U.S. meteorological ships that reported positions at 2-hourly and 0.01° precision in both longitude and latitude. Running the model allows for accounting systematic position errors arising from locations inferred by dead-reckoning not being updated until celestial corrections. Moreover, random position error (1 s.d.) is quantified to have a posterior median of 0.18° in longitude and 0.15° in latitude, which translates up to 1°C SST uncertainties in western boundaries of ocean basins. Our results highlight the importance of accounting for position errors in SST reconstructions. Combining these results with the recent identification of statistically-constrained systematic offsets among groups of SSTs, we call for more usage of sophisticated statistical methods in understanding past climate change.
|
