Offline Database Queries/Discussion

Hi all. Am playing a little with the offline database and the two practice (homework) papers, and have some queries/comments. Thought I'd post here rather than email.

Barbraud, & Weimerskirch (2001): As noted by El, the only statistical analysis with climate data here are survival estimates for the period 1982-1988. However,

1 - Is it correct to associate observations of survival as "Growth and condition"? I guess you could say a dead penguin was in poor condition, but I would more normally associate this term with length-corrected body mass or something like that. In fact, since demography measures population characteristics, and mortality rate is one of these statistics, surely survival rates fit better here? Alternatively, I would suggest a separate observational category of Survival/mortality rates, with the method of estimation (mark-recapture, length-frequency analysis, etc., as additional detail categories).

2 - The slope of the logit-linear relationship between survival and SST (-6.02 +- 1.01, amazingly, no units provided!) does give some rate information, although I'm not sure exactly what this is. Should we get a demographer's input on this? Does anybody keep a tame one on hand?

3 - Although the field "Tools applied to determine change" is empty in the offline database, it seems to me that there was a formal test: (to me it seems) a log-likelihood ratio test of alternative logit-linear models with and without SST as an explanatory variable (although there are some issues with this test, it is backed up by HUGE differences in AIC). Is there a reason that this field is empty, or did I delete it myself in error?

Any comments/suggestions?

- Dave

OK, have been busy with the end of the academic semester. Back to the papers. Looking at Atkinson et al (krill/salps), I have additional questions:

1 - The authors demonstrate a temporal relationship between krill density and sea ice duration (35 years) and latitude (25 years); both are time-series regressions (hence formal statistical analyses relating to climate-change indicators), both are significant, and both indicate a climate-change impact. Which do we report here? The strongest? The longest? What are our rules? Do we need any? Do we need to capture this info in additional fields or notes?

2 - As far as I can see, although the authors test for and find a (temporal) climate relationship between sea ice and krill, and they use a spatio-temporal model to indicate a significant increase in salp densities over time, they do not tie this increase to any direct climate impact (this is actually reported on p 102: "Salps showed no such relationships, despite a negative one being postulated previously off the Antarctic Peninsula. With shorter life cycles than krill and explosive population growth rates, salps can respond to environmental variation over shorter timescales."). Instead, the authors make an argument on the contrasting directions of the temporal trends in krill and salp density to speculate that low krill abundance frees up ecological space for salps; but they do not report a formal test of the hypothesis that densities of these two groups are negatively correlated. In other words, the influence of climate on salps is not demonstrated here, and if it exists, it is indirect, expressed through an ecological feedback, and even this indirect link is not tested. In other words, we should not report salps in the database, according to my understanding.

Any comments/ideas?

- Dave

More questions, I'm afraid. Why is Barbraud, C and H. WeimerskirchBarbraud & Weimerskirch listed as two separate papers (Papers 1 & 2), but the same Observation (first two data fields)? And why is Atkinson et al listed as paper 3? These data fields are not described in the associated text...

- Dave

On to Dulvy et al - here we are looking at a community response, rather than a sp-specific response. How do we code that in terms of the data fields currently available in "Focus of Study"?

- Dave

Again on Dulvy et al. Here we have a lot of temporal patterns, but far more limited response to climate variables. Response per year/decade is available in the database, but is the warranted - surely a temporal change does not indicate a response to climate; only a relationship with a climate variable can demonstrate that?

Given that we work only with relationships to climate variables, we have here an analysis of a community aggregated at various levels. Which do we use - the most aggregated or the least aggregated? And which response to we list - the strongest only, or all of them?

- Dave

Previously Dave Schoeman wrote:

Hi all. Am playing a little with the offline database and the two practice (homework) papers, and have some queries/comments. Thought I'd post here rather than email.

Barbraud, & Weimerskirch (2001): As noted by El, the only statistical analysis with climate data here are survival estimates for the period 1982-1988. However,

 

1 - Is it correct to associate observations of survival as "Growth and condition"? I guess you could say a dead penguin was in poor condition, but I would more normally associate this term with length-corrected body mass or something like that. In fact, since demography measures population characteristics, and mortality rate is one of these statistics, surely survival rates fit better here? Alternatively, I would suggest a separate observational category of Survival/mortality rates, with the method of estimation (mark-recapture, length-frequency analysis, etc., as additional detail categories).

 

Sydeman - I don't think 'growth and condition' is correct.  For seabirds, survival rate of adults is a (the?) key demographic parameter that drives population dynamics.  I would agree it should be listed under 'population characteristics'.  I thought there was a separate category.

 

2 - The slope of the logit-linear relationship between survival and SST (-6.02 +- 1.01, amazingly, no units provided!) does give some rate information, although I'm not sure exactly what this is. Should we get a demographer's input on this? Does anybody keep a tame one on hand?

 

Sydeman - i know one but he is not tame.  Seriously, though, this rate seems strange...units would be -6% survival per 1 degree C...I don't believe it.  Seabird survival is very high, generally over 90%, so this would indicate the population is extinct, which its not.   

 

3 - Although the field "Tools applied to determine change" is empty in the offline database, it seems to me that there was a formal test: (to me it seems) a log-likelihood ratio test of alternative logit-linear models with and without SST as an explanatory variable (although there are some issues with this test, it is backed up by HUGE differences in AIC). Is there a reason that this field is empty, or did I delete it myself in error?

 

Sydeman - Yes, I agree this is an explicit test of different model configurations -- based on AIC, SST is better in the model than not.  

 

Any comments/suggestions?

 

- Dave