Chapter 12 . Pattern Analysis 01 Endemie Fish Speeies in Rivers
243
et al. 1996b). Therefore, ANNs appear to eonstitute a powerful alternative in
predietive eeologieal modelling.
With both the Profile method and the Pad method, TSR is found to be the most
important variable in explaining the variability of the endemie species riehness.
This is in agreement with the eonclusions made by Oberdorff (1999). Moreover,
the way that the TSR varies the ESR is also similar in the two methods. For low
values of TSR, ESR does not inerease, then ESR inereases with average values of
TSR and with high TSR values, ESR inereases but more moderately.
The two variables SAD and NPP are not expressed by both methods. If ESR is
predieted only by the TSR the determination eoefficient found will be r2=0.83.
With the Profile method their eurves have a limited range and with the PaD
method their partial derivatives values are almost all near zero. This is in
agreement with the study of Oberdorff (1999), where the total species riehness is
the only variable influencing the endemie species riehness.
The Profile and PaD methods were applied to the total species riehness to test
whieh of the two variables NPP and SAD is the most important, and how they
intervene in the TSR variation. In this ease, the eurves of SAD and NPP are weil
expressed with the Profile method, and the partial derivatives differ from zero
with the PaD method. TSRs inerease with SADs using both methods, but they
give different results for NPPs. With the Profile method TSRs inerease with NPPs
whereas with the Pad method, TSRs inerease only for values of NPP between
1500 and 2250 kg"2y"1.
In eonclusion, these methods show that the endemie species riehness is
dependent only on the total species richness and that total species riehness is
dependent on the surfaee area of the drainage basin and to a lesser extent, the net
primary produetivity.
Disadvantages
The principal disadvantage of the Profile method is the use of a fietitious matrix.
In faet, twelve values were retained for eaeh variable, taken from between their
minimum and maximum values at equal intervals. The plotted eurves for eaeh
variable are then smoothed, and are therefore not precise. For adefinite value of
one independent variable, it is not direet1y possible to find the value of the
dependent variable. The eurves ean only demonstrate how the variables influenee
the output in relation to an inerease in its values; it is a general view. If the
eontribution order of the variable is needed, this method is not able to give it
directly. An analysis for adefinite value ean only be earried out by eomparison of
the eurves' slopes.
Beeause a graph is obtained for eaeh variable with the Pad method, it is less
easy to get a general view of the influenee of all the variables, as with the Profile
method.
Advantages
The Pad method seems more favourable than the other. This method uses the real
values of the observations. Moreover, it permits the direet obtaining of two things;
the first being the evolution of the output aeeording to the inerease in eaeh
independent variable and this with the precise values of the independent variable.
243
et al. 1996b). Therefore, ANNs appear to eonstitute a powerful alternative in
predietive eeologieal modelling.
With both the Profile method and the Pad method, TSR is found to be the most
important variable in explaining the variability of the endemie species riehness.
This is in agreement with the eonclusions made by Oberdorff (1999). Moreover,
the way that the TSR varies the ESR is also similar in the two methods. For low
values of TSR, ESR does not inerease, then ESR inereases with average values of
TSR and with high TSR values, ESR inereases but more moderately.
The two variables SAD and NPP are not expressed by both methods. If ESR is
predieted only by the TSR the determination eoefficient found will be r2=0.83.
With the Profile method their eurves have a limited range and with the PaD
method their partial derivatives values are almost all near zero. This is in
agreement with the study of Oberdorff (1999), where the total species riehness is
the only variable influencing the endemie species riehness.
The Profile and PaD methods were applied to the total species riehness to test
whieh of the two variables NPP and SAD is the most important, and how they
intervene in the TSR variation. In this ease, the eurves of SAD and NPP are weil
expressed with the Profile method, and the partial derivatives differ from zero
with the PaD method. TSRs inerease with SADs using both methods, but they
give different results for NPPs. With the Profile method TSRs inerease with NPPs
whereas with the Pad method, TSRs inerease only for values of NPP between
1500 and 2250 kg"2y"1.
In eonclusion, these methods show that the endemie species riehness is
dependent only on the total species richness and that total species riehness is
dependent on the surfaee area of the drainage basin and to a lesser extent, the net
primary produetivity.
Disadvantages
The principal disadvantage of the Profile method is the use of a fietitious matrix.
In faet, twelve values were retained for eaeh variable, taken from between their
minimum and maximum values at equal intervals. The plotted eurves for eaeh
variable are then smoothed, and are therefore not precise. For adefinite value of
one independent variable, it is not direet1y possible to find the value of the
dependent variable. The eurves ean only demonstrate how the variables influenee
the output in relation to an inerease in its values; it is a general view. If the
eontribution order of the variable is needed, this method is not able to give it
directly. An analysis for adefinite value ean only be earried out by eomparison of
the eurves' slopes.
Beeause a graph is obtained for eaeh variable with the Pad method, it is less
easy to get a general view of the influenee of all the variables, as with the Profile
method.
Advantages
The Pad method seems more favourable than the other. This method uses the real
values of the observations. Moreover, it permits the direet obtaining of two things;
the first being the evolution of the output aeeording to the inerease in eaeh
independent variable and this with the precise values of the independent variable.