BIOS0002 Assessment 2: Statistics
KRYG2
17/12/2020

The analysed dataset is catch data from the DFO Maritimes Research Vessel Trawl
Surveys in the Canadian Atlantic Ocean around Nova Scotia from 1970-2010. Over
these 40 years, 231 taxa were observed in a total of 121,804 observations. For each
catch, the abundance of taxa as well as their total biomass were recorded.
Here, I want to test whether species richness, abundance and biomass in the area
change over time, which could be effects of climate change and overfishing and give
an indication of the status of local populations in the area.
If any of these factors have an impact on the local ecosystem, one would expect all
three observed variables (richness, abundance and biomass) to decrease as the
impact of climate change and overfishing increases with time. Therefore, we can pose
the null hypothesis:
N0: Species richness, abundance and biomass are independent of time
And the alternative hypothesis:
N1: Time has an impact on species richness, abundance and biomass
Further, I want to explore how these variable change in the ten most abundant species
in the data set to understand whether trends in these species follow the overall trends
over time.
Fig. 1 (below) shows the number of species identified per catch. Fitting a generalised
linear model using a Poisson distribution to the data showed that species richness
increased over the observed timespan (χ² = 990.16; d.f. = 1, 1687; p < 2*10-16).

Therefore, the null hypothesis is rejected and species richness in catches increases
over time, which is counterintuitive to the expectation that species richness might
decrease over time.
Fig 2. (below) shows how the number of individuals per catch changes over time. Note
that the log-scale for the number of individuals only applies to the figure and the GLM
was fitted using untransformed count data. Using the same model as for the number
of species above, an overall increase in species abundance in catches was observed
(χ² = 593637; d.f. = 1, 1687; p < 2*10-16)

The observation that both species richness and abundance change over time could
be due to changes in the methods used in sampling which were not documented in
the dataset description. Alternatively, an increase in sea temperature could have led
to an increase in species as new species from more southern areas follow temperature
gradients northwards. The available data however does not allow a conclusive
explanation for this trend.


Fig. 3 (below) shows the mean biomass per catch over time. This was analysed using
linear regression analysis on the log-transformed data, which showed a significant
decrease of mean biomass over time (y = 3.068 – 8.411*10-5 x; F = 357.1; d.f. = 1,
1686; p < 2*10-16), rejecting the null hypothesis.

In the light of these results, I isolated the 10 species with the highest overall abundance
in the dataset to investigate whether their occurrence data shows similar trends. Fig.
4 (next page) shows the abundance of these species over time as well as fitted
generalised linear models based on a Poisson distribution. As above, the y-axis has
been log-scaled, but the models were applied to the raw count data, hence the non-
linear appearance of the regression lines. Apart from C. harengus and M. aeglefinus,
these species show an overall decrease in their average catch abundance. This in turn
means that other species are becoming more abundant due to the overall trend of
abundance increase and that the local community composition might be changing
slowly.

Critical reflection:

This dataset shows several flaws which made analysis difficult and may affect the
strength of these conclusions.
Firstly, there is a very sudden increase in the number of species around 1995, for
which no explanation can be found in the dataset description. The DFO Surveys
have at the same time been performed for invertebrates and the sampling methods
description for the Invertebrate dataset mentions a change in trawling gear in 1995.
In addition, the invertebrate sampling has been extended to “deepwater and inshore
strata” in 1995 which subsequently were “not consistently covered”. Whether these
limitations apply to the dataset analysed here (which does not contain invertebrates),
is unclear. (http://ipt.iobis.org/obiscanada/resource?r=dfogfsdbinv)
Further, the sampling times are not consistent. From 1970-1978, the survey was only
done once a year (see Fig.1-3), with a subsequent increase in sampling frequency
from 1978 until 1985, when the surveys are reduced to twice a year. This
inconsistency could affect the accuracy of results, especially for species that are not
permanent residents and undergo regular migrations.
Finally, the dataset does not contain records of environmental variables such as
temperature which could help in quantifying climate change, or fisheries data which
could indicate changes in the degree of exploitation. While climate change is widely
accepted to have increased over time, time itself is not a measure of climate change
and the observed trends can only indirectly and not conclusively be related to climate
change. The most abundant species are mostly commercially important species, but
their decrease in abundance could be due to climate change, overfishing or some
other factor which cannot be tested for in this circumstance.
In conclusion, the dataset shows some interesting trends in the trawl survey data,
most notably an increase in species richness and abundance, a decrease in mean
catch biomass and a decrease in the abundance of some of the most abundant
species. However, any solid conclusions would require additional data and the