What can we say about prediction

1
What can we say about prediction?

• predictions are needed for decision-making
• predictions can be fuzzy (e.g. 30 chance of rain
  this evening)
• prediction has been used as a justification for
  most GLOBEC research
• full understanding is not necessary for prediction

2

• Secondary Production in the Oceans and the
  Response to Climate Change
• Mike Heath, Francois Carlotti, Brad de Young,
  Oyvind Fiksen and Cisco Werner
• Trends over the past 50 years in the ocean
  climate of the North Atlantic (convective depth,
  poleward heat transport, overflow from the Nordic
  Seas) are now well documented.
• Some of these are correlated with the North
  Atlantic Oscillation Index. What is the impact of
  these large scale, low frequency climatic changes
  on the living resources of the North Atlantic?
• From ICES/GLOBEC Newsletter 8 Sept 2001

3
What kinds of prediction?

• theres a 30 chance of rain this evening
• the return time for a heatwave like 2003 is much
  less than 200 years
• high cod recruitment is less likely when NAO is
  high
• how can we include incomplete understanding?

4
What does cod recruitment look like?

• contrary to popular myth, recruitment varies very
  little
• (Table 1 of Brander 2003 shows values for 60
  stocks)
• some of the variability (about 30) is at a
  bigger scale than the stock

5
(No Transcript)
6
Lots of cross-correlations in cod recruitment
r values
p lt .05
p lt .01
r values
Adjusted deg of freedom
7
From Beaugrand et al. 2003 What aspects of
zooplankton state (as monitored by CPR) would you
expect to affect North Sea cod recruitment? Survi
val index is higher when the right size of prey
are abundant at the right time of year. This
could explain why survival increased from 1958 to
1963 and then decreased again after 1985 These
changes are probably due in part to changes in
the NAO
8
These larvae need prey gt0.03
Heres a low NAO high survival year
9
Divide data (SSB, R and NAO) into 3 categories to
overcome scale differences and carry out
frequency analysis. Rothschild B.J. and Mullen
A.J. 1985. The information content of
stock-and-recruitment data and its non-parametric
classification. Journal du Conseil International
pour l'Exploration de la Mer 42 116-124
10
Joint frequency analysis (200 values) and ?2p
that R is independent of NAO lt0.001 at low SSB
lt0.1 at med SSB gt0.5 at high SSB NAO
has a strong effect on recruitment when SSB is
lowEnvironmental effect is not independent of
SSB
11
Based on observed frequencies NAO has a big
effect on R when SSB is low, but not when it is
high At low SSB high recruitment has been about 9
times more frequent when NAO is low
The processes causing the apparent relationship
could include NAO temperature
relationship Stratification plankton
relationship Inflow plankton relationship
Most GCMs predict continued high NAO
12
Environmental variables
Recruitment, growth, mortality of cod
13
Questions

• What kinds of prediction do we need to do the
  job?
• How do we present predictions (including degree
  of confidence) in order to help decision making?
• How do we design the process of providing advice
  so that we get better as we learn more?