1 / 12

Defining alternative climatologies: why, what for and how

Defining alternative climatologies: why, what for and how. overview:. why?. – problems with current climatologies. 2) what for?. – applications. 3) how?. – discrete process convolution models. why?.

conroy
Download Presentation

Defining alternative climatologies: why, what for and how

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Defining alternative climatologies:why, what for and how

  2. overview: • why? – problems with current climatologies 2) what for? – applications 3) how? – discrete process convolution models

  3. why? Traditional climatologies, such as Levitus climatologies of ocean temperature and salinity, use Objective Analysis Techniques. firstguess(lat,lon,dep)=zonal mean For pass=1 to 3 For all lat, lon and dep newguess(lat,lon,dep)= firstguess(lat,lon,dep) + weighted sum within a radius(pass) of differences between observed and firstguess means Next For all lat, lon and dep Run a median smoother and/or another type of spatial smoother on newguess Next For all lat, lon and dep firstguess(lat,lon,dep)=newguess(lat,lon,dep) Next Next smoothednewguess firstguess observations newguess +

  4. why? Although termed an objective method, several subjective choices must be made: a) number of passes b) weighting function c) radii sizes d) type of smoother and associated parameters The whole process is purely mathematical, and corrections to meet physical assumptions (e.g., stability of the water column) must be performed a posteriori. Example of weighting function Radii sizes in Levitus 1º and ¼º climatologies Example of a linear smoother

  5. why? But the most important problem is that sampling errors are not taken into consideration with the Objective Analysis Technique. (a sampling error is an error that derives from not obtaining a representative sample) World Ocean Database – used by Levitus et al.

  6. why? Impacts of sampling errors: In each region, one particular decade has more observations than the others; this decade is not the same for all regions.

  7. why? Impacts of sampling errors: Since temperature and salinity change from one decade to the next, the climatology may be unrealistic, i.e., it may present inexistent gradients (1) and/or lose others that exist (2). (1) (2)

  8. 2) what for? • Alternative climatological fields, specifically designed for the region of interest, will allow: • a better characterization of the hydrodynamic circulation, including upwelling, countercurrents and undercurrents; • more reliable estimates to be coupled with other circulation models (e.g., MM5), as well as biological models; • more reliable time-series of anomalies that will be used to investigate climate change and relationships between properties (e.g., coastal temperature and winds, coastal chlorophyll-a and winds) • freedom for MOHID users to customize the climatology From: The MOHID Portugal/Galicia Regional model – first numerical experiments

  9. 3) how? With a statistical-based approach that allows the data to dictate the amount of spatial smoothing that seems appropriate and can also accommodate the temporal component of the data. This will lead to time averaged spatial field estimates that are not biased towards any particular decade. E.g.: yearly anomaly fields yearly climatology

  10. 3) how? Discrete process convolution models (DPCMs) Example of a 1D standard process convolution with a Gaussian kernel • A DPCM requires: • a kernel • a discount factor • a lattice The discount factor is the component of the DPCM that determines the amount of information lost from one time instant to the next.

  11. 3) how? Discrete process convolution models (DPCMs) Choosing the lattice, the discount factor and the kernel’s shape determines the quality of the fit. These choices are initially subjective but may be optimized with Markov chain Monte Carlo methods. Different kernel shapes for different regions Examples of lattices

  12. summary on “Defining alternative climatologies” why? – current climatologies derive from subjective mathematical procedures that require posterior corrections; also they do not take sampling errors into consideration. what for? – to provide reliable products for many people how? – with discrete process convolution models, coupled with MCMC methods to increase the goodness-of-fit and reduce the subjectivity.

More Related