Dr Yves Plancherel


Postdoctoral Research Associate

Email: yves.plancherel@earth.ox.ac.uk
TEL: +44 (1865) 272000
FAX: +44 (1865) 272072
Homepage: http://www.earth.ox.ac.uk/~yvesp


Research Profile

Some anthropogenic activities have reached intensity levels, either in magnitude or due to their cumulative effects, where noticeable effects on the environment can be felt on a variety of scales. While local influences of these activities are well-known and can in general be assessed and remediated directly (e.g. pollution of a stream), assessments on broader or longer scales, such as those relevant for climate, require extensive data coverage both in space and time and an accurate description of the relevant natural variability to tease apart typically small signals from large backgrounds. Studies on these scales necessarily require international community-wide efforts over many years to achieve the data coverage and analytical precision needed. I specialize in the synthesis and interpretation of such data, with my main areas of interest being the dynamics of marine elemental cycles and the interaction of the ocean with climate. As a data “opportunist”, however, I am always looking for interesting problems on which to apply my skills and curiosity. A brief description of various projects is given below.  

Anthropogenic carbon in the ocean

Carbon dioxide in the atmosphere has been increasing from about 280 ppm before the onset of the industrial revolution to 393 ppm today (January 2012). The repeated measurements of dissolved inorganic carbon (DIC) in the ocean is one way to infer the temporal and regional evolution of the marine uptake and fate of this excess carbon. However, because the spatial coverage of DIC data in the ocean is concentrated along sections and that these trans-oceanic sections take months to complete, the spatio-temporal data distribution of these data presents an interpretation challenge. This is exacerbated by the fact that large DIC excursions (of similar magnitude to the anthropogenic signal) exist that are due to natural variability only (e.g. gyre wobble, frontal shifts, etc.).

Means to isolate the anthropogenic signal from the variable background in the context of non-ideal data availability are thus necessary. One such approach (the “eMLR” method) proposes to study the temporal evolution of empirical regression models of the DIC data instead of the DIC data directly. Since by definition regression analysis filters noise, provided “noise” equates to “natural variability”, this technique should in principle allow for the quantification of the interior propagation of the anthropogenic signal on time-scales relevant for repeated basin-scale observations (i.e. 5-15 years). Does this method in fact work? By how much did the interior anthropogenic carbon sink change since the last assessment, about 10 years ago?

Just as observational chemists use idealized solutions (e.g. in a beaker, with known amount of materials) to calibrate their techniques, calibrations are also necessary for computational analytical techniques. Models of the ocean circulation and biogeochemistry, which are typically used to provide hind- and forecasts on a variety of scales that are not easily measurable, can also be used as the computational equivalent to the “beaker” to calibrate algorithms.

Using output from a state-of-the-art circulation model coupled with a complex biogeochemical module run with historical heat fluxes and wind stresses, it has been possible to generate a synthetic yet realistic DIC data set with exactly known anthropogenic carbon evolution, providing a means to quantify absolute methodological errors. This synthetic data set was used to constrain and understand the strengths and weaknesses of the eMLR protocol with regard to its ability to recover the anthropogenic carbon signal.

Application of the optimized eMLR protocol globally to real data is currently underway. This involves also the consolidation and cross-calibration of a global data set of DIC from GLODAP, CARINA, PACIFICA and other new CLIVAR and GO-SHIP cruises.

This work is a collaborative effort with R.M. Key and K.B. Rodgers from Princeton University and A.R. Jacobson from the University of Colorado Boulder and NOAA/ESRL.  

 

Ocean circulation and climate

Given the threat of anthropogenically-induced climate change, there is a very intensive and global effort to develop and run climate models. Some of these modeling results, those used as part of the Intergovernmental Panel on Climate Change Assessment Reports (IPCC AR), are typically archived for public usage. The predictions from these models with regards to the temporal evolution of many variables differ substantially from model to model, however, impairing confidence in these models’ predictive ability.

There is widespread, however as yet unproven, belief that models that simulate the modern observed state better also simulate future climate better. Predictions from these models are often given more weight in inter-model climate assessments. Given the many degrees of freedom of climate models, however, a good fit to observations can be achieved for a variety of reasons, not all physically sound. It is in fact no so clear that better modern simulations automatically yield better predictions.

Given that most models are built and run differently (different numerics, forcings, parameterization, resolutions, etc), it is difficult to pinpoint the dominant causes for model discrepancies. A different way forward exists: instead of trying to minimize biases, which can be a futile exercise, an alternative is to accept model discrepancies and treat them as perturbation experiments. It is then possible to look (and find) relationships between model states and model predictions across all models. In this framework, the observed modern state can be treated as another sensitivity experiment for which a prediction does not exist. If inter-model differences behave in a somewhat predictable way and the inter-model spread captures the observational state, theoretical or empirical relationships derived from regression analysis of inter-model results can be used to generate a prediction of how the observed state should behave.

Using this approach on the IPCC AR-4 model set indicated that the Southern Ocean abyssal circulation is expected to decrease by about 25% by 2100 relative to the modern circulation. Further applications are being investigated.


Synthesis and modeling of metals distribution in the ocean

While there exist globally millions of temperature and salinity measurements, tens to hundreds of thousands of measurements for macronutrients (oxygen, DIC, phosphate, nitrate, silicate), only few (relative to the globe) measurements are available to characterize the dynamics of micronutrient and other metals in the ocean. Some of these metals (Fe, Zn, Mn, etc. ) have been shown to participate in and sometimes limit primary production. Others, such as rare earth elements, readily scavenge onto particles and can be used to learn about large-scale particle dynamics. Given the limited number of metal data available, first order questions regarding the sources, sinks and internal dynamics of these metals have yet to be answered.

The GEOTRACES program, an international community effort aiming to characterize the distribution of trace elements in the ocean, has stimulated collection and analysis of metals data. Given the growing number of measurements available, it is becoming possible to exploit this new data stream to investigate metal dynamics on large and regional scales as well as their role as micronutrients.