Analysis of the relationship between fish habitat classifications and topological lake units

Analysis of the relationship between fish habitat classifications and topological lake units
Analysis of the relationship between fish
habitat classifications and topological lake
units
T. Frezza and C.K. Minns
Central and Arctic Region
Department of Fisheries and Oceans, Great Lakes
Laboratory for Fisheries and Aquatic Sciences
Burlington, Ontario L7R 4A6
2002
Canadian Manuscript Report of
Fisheries and Aquatic Sciences 2600
Canada
Canadian Manuscript Report of
Fisheries and Aquatic Sciences
Manuscript reports contain scientific and technical information that contributes to existing
knowledge but which deals with national or regional problems. Distribution is restricted to
institutions or individuals located in particular regions of Canada. However, no restriction is
placed on subject matter, and the series reflects the broad interests and policies of the
Department of Fisheries and Oceans, namely, fisheries and aquatic sciences.
Manuscript reports may be cited as full publications. The correct citation appears above
the abstract of each report. Each report is abstracted in Aquatic Sciences and Fisheries
Abstracts and indexed in the Department’s annual index to scientific and technical publications.
Numbers 1-900 in this series were issued as Manuscript Reports (Biological Series) of
the Biological Board of Canada, and subsequent to 1937 when the name of the Board was
changed by Act of Parliament, as Manuscript Reports (Biological Series) of the Fisheries
Research Board of Canada. Numbers 901-1425 were issued as Manuscript Reports of the
Fisheries Research Board of Canada. Numbers 1426-1550 were issued as Department of
Fisheries and the Environment, Fisheries and Marine Service Manuscript Reports. The current
series name was changed with report number 1551.
Manuscript reports are produced regionally but are numbered nationally. Requests for
individual reports will be filled by the issuing establishment listed on the front cover and title page.
Out-of-stock reports will be supplied for a fee by commercial agents.
Rapport manuscrit canadien des
sciences halieutiques et aquatiques
Les rapports manuscrits contiennent des renseignements scientifiques et techniques qui
constituent une contribution aux connaissances actuelles, mais qui traitent de problèmes
nationaux ou régionaux. La distribution en est limitée aux organismes et aux personnes de
régions particulières du Canada. Il n’y a aucune restriction quant au sujet; de fait, la série reflète
la vaste gamme des intérêts et des politiques du ministére des Pêches et des Océans, e’est-àdire les sciences halieutiques et aquatiques.
Les rapports manuscrits peuvent être cités comme des publications complèrwa. Le titre
exact paraît au-dessus du résumés de chaque rapport. Les rapports manuscrits sont résumés
dans la revue Résumés des sciences aquatiques et halieutiques,et ils sont classés dans l’index
annuel des publications scientifiques et techniques du Ministére.
Les numéros 1 à 900 de cette série ont été publiés à titre de manuscrits (série
biologique) de l’Office de biologie du Canada, et aprés le changement de la désignation de cet
organisme par décret du Parlement, en 1937, ont été classés comme manuscrits (série
biologique) de l’Office des recherches sur les pêcheries du Canada. Les numéros 901 à 1425
ont été publiés à titre de rapports manuscrits de l’Office des recherches sur les pêcheries du
Canada. Les numéros 1426 à 1550 sont parus à titre de rapports manuscrits du Service des
pêches et de la mer, ministère des Pêches et de l’Environnement. Le nom actuel de la série a
été établi lors de la parution du numéro 1551.
Les rapports manuscrits sont produits à l’échelon régional, mais numérotés à l’échelon
national. Les demandes de rapports seront satisfaites par l’établissement auteur don’t le nom
figure sur la couverture et la page du titre. Les rapports épuisés seront fournis contre rétribution
par des agents commerciaux.
Canadian Manuscript Report of
Fisheries and Aquatic Sciences 2600
2002
Analysis of the relationship between fish habitat classifications
and topological lake units
by
T. Frezza and C.K. Minns
Central and Arctic Region
Fisheries and Oceans Canada
Burlington, Ontario
 Minister of Supply and Canada 2002
Cat. No. FS97-6/2600 ISSN 0706-6457
Correct citation of this publication:
Frezza, T. and C.K. Minns. 2002. Analysis of the relationship between fish habitat
classifications and topological lake units. Can. MS Rep. Fish. Aquat. Sci. 2600:
vi+15p.
ii
ABSTRACT
The goal of this study was to objectively assess subjective fish habitat classification
systems using topological measurements of lake units such as depth, slope, and a range
of fetch indices from a set of small lakes (<25 ha). This study is in attempt to develop a
predictive modelling tool for assessing fish habitat from measurements easily obtained
from bathymetric maps.
The spatial distribution of both substrate and cover (wood and macrophytes) were
mapped for each of the lakes. At systematically selected sampling points, values were
calculated for depth, slope and a range of fetch indices. Discriminant functions analyses
(DFA) were performed to predict substrate and cover classifications separately in the
littoral zone and in the entire lake.
There was poor success in predicting field classifications of both substrate and
cover using topological units as predictors. For substrate, the more simplified the
classes were, the higher the success of predicting classification. The functions for
predicting substrate in lakes are not transferable between lakes.
The sizes of the study lakes were too small to determine the effects of wind driven
forces. Only a small portion of the lakes are affected by wind driven forces at least some
of the time. In the future, this analysis should be repeated on lakes with larger areas to
determine both the effect of wind driven forces and the influence of lake area on
substrate distribution.
RÉSUMÉ
L’objectif de la présente étude était de faire une évaluation objective des
systèmes subjectifs de classification de l’habitat du poisson en se servant de paramètres
topologiques de lacs individuels, comme la profondeur, la pente et une gamme d’indices
de portée provenant d’une série de petits lacs (< 25 ha) en vue de mettre au point un
outil de modélisation prédictive pour évaluer l’habitat du poisson à partir de mesures
faciles à obtenir de cartes bathymétriques.
Nous avons établi des cartes de la distribution spatiale du substrat et de la
couverture végétale (arbres et macrophytes) pour chaque lac expérimental et nous
avons mesuré la profondeur, la pente et une gamme d’indices de portée à des points
d’échantillonnage choisis systématiquement. Puis nous avons utilisé des analyses
iii
discriminantes (AD) pour prédire séparément dans quelle catégorie d’inscrivait le
substrat et la couverture végétale de la zone littorale et du lac entier.
L’utilisation d’unités topologiques comme variables explicatives n’a pas donné de
bons résultats pour ce qui est de prédire la classification sur le terrain du substrat et de
la couverture végétale. Dans le cas du substrat, plus simplifiées étaient les classes, plus
grand était le succès de la prédiction de la classification. Par contre, les fonctions
servant à prédire le substrat d’un lac ne sont pas transférables à un autre.
Il nous a été impossible d’établir les effets des forces dues aux vents parce que
les lacs expérimentaux étaient trop petits. Ces forces ont une incidence sur une petite
partie des lacs au moins pendant un certain temps. Une nouvelle analyse portant sur
des lacs de plus grande superficie devrait être faite à l’avenir en vue d’établir les effets
de ces forces et l’incidence de la superficie d’un lac sur la distribution du substrat.
iv
TABLE OF CONTENTS
Abstract/Résumé..................................................................................................... iii
Introduction.............................................................................................................
1
Methods...................................................................................................................
2
Study Area........................................................................................................
2
Habitat Assessment.........................................................................................
2
Topological Units..............................................................................................
2
Analysis............................................................................................................
3
Results.....................................................................................................................
4
Discussion...............................................................................................................
5
References...............................................................................................................
8
LIST OF TABLES
Table 1
Definition of substrate types revised from Lester et al. (1998)........
10
Table 2
Definitions and descriptions of macrophyte types (Frezza 2001)....
10
Table 3
Results from a principal components analysis using data from
Little Turkey Lake............................................................................. 11
Table 4
Classification success of a DFA to predict substrate classes in the
nearshore area from five topological units.......................................
Table 5
Classification success of a DFA to predict substrate classes in the
entire lake from five topological units...............................................
Table 6
12
12
Classification matrix from the DFA indicating the success of
predicting substrate classes on Little Turkey Lake at sampling
points using topological variables....................................................
Table 7
13
Classification matrix from the DFA indicating the success of
predicting substrate classes at sampling points from all the lakes
using toplogical variables.................................................................
Table 8
Classification success of a DFA to predict cover classes in the
v
13
nearshore area and the entire lake from five topological lake units.
Table 9
14
The area, mean depth, dynamic ratio, and percnet of area
disturbed some of the time by wind driven forces for each lake......
14
LIST OF FIGURES
Figure 1
Factor scores form the DFA of Little Turkey Lake using habitat
units of bedrock, boulder, and other................................................
vi
15
INTRODUCTION
The relationship between energy inputs into a lake and substrate distribution are
hard to define and quantify; this is especially true in small lakes with effective fetch
values less than 5 km where wind and wave influences are limited (Hakanson and
Jansson 1983; Cyr 1998). Topological lake units such as slope, depth, and fetch indices
are known to have an influence on sediment distribution in lakes. Hakansan and
Jansson (1983) and Franzin (1999) have described an erosion, transport, and deposition
concept, whereby hydraulic processes based on energy from wind-driven waves
determine substrate composition in littoral area of lakes (Franzin 1999).
Cyr (1998) investigated the effects of wave disturbance and slope on sediment
characteristics in the littoral zone of small lakes. She found that the depth of waveinduced sediment resuspension increases with increasing wind exposure and with
decreasing size and density of sediment particles. Direct wave action, and higher wind
energy can resuspend fine sediments (Cyr 1998) and transport larger particles, creating
a gradient in particle size distribution (Sly 1978).
An understanding of the effect of energy inputs into a lake, and therefore the
contributions of topological features on particle distribution, would allow for predictive
modelling of fish habitat features. Traditional fish habitat assessments, involving data
collection in the field, can be unreliable, labour intensive, and costly. Several problems
have been documented concerning the reliability of fish habitat data collection. Error in
assessment can be introduced from several sources such as investigator experience or
training (Roper and Scarnnecchia 1995), interpretation of variables to be measured and
observer subjectivity (Kondolf and Li 1996; Stanfield and Jones 1998). The ability to
predict fish habitat using topological lake units would provide a predictive modelling tool
for fish habitat managers and reduce inherent subjectivity of current assessment
methods.
The goal of this study was to objectively assess subjective fish habitat classification
systems using topological measurements of lake units such as depth, slope, and a range
of fetch indices. Analyses are based upon the erosion, transport, and deposition
concept and applied to the fish habitat distribution of a set of small lakes (<25 ha) in
Northern Ontario. This study is in attempt to develop the predictive modelling tool for
assessing fish habitat from measurements easily obtained from bathymetric maps.
1
METHODS
Study Area
Five small lakes (<0.25 km2) were used for this analysis, Little Turkey, Lower
Batchawana, Upper Batchawana, and Wishart Lakes, all located within the Turkey Lakes
Watershed (TLW) (84o25’W, 47o03’N) approximately 50 km north of Sault Ste. Marie,
and Quinn Lake (84o13’14”W, 46o43’52”N) ~ 50 km SE of TLW. All field sampling on the
lakes took place been July 1 and September 5, 1998.
Habitat Assessment
Maps of the spatial distribution of substrate, macrophyte, and wood coverage were
completed on all of the five lakes using the littoral zone cruise habitat assessment
method (Frezza 2001) (area between shore and the 2 m contour). Substrate was
classified into eight different habitat types by particle size (Table 1) and macrophytes
into seven categories according to growth forms (Table 2). Pieces of coarse woody
material (greater than 10 cm diameter) were counted in approximately every 20-50 m of
shoreline distance from shore to the 2 m contour. In a GIS, the shoreline of each lake
was divided into sections where the wood was counted, and the corresponding wood
count assigned to each section. From a previous habitat assessment completed on
these study lakes using a transect sampling method (Frezza 2001), the average
distance from shore that wood was present was calculated. A buffer zone, using the
average distance wood was found from shore as its width, was placed parallel to the
shoreline segments to create polygons. The density of wood within each polygon was
calculated (number of pieces of wood per metre2 of shoreline). Wood densities within
polygons were grouped into five classes: 0 pieces/m2; 0.01-0.5 pieces/m2; 0.51-1.5
pieces/m2; 1.51-2.5 pieces/m2; and >2.5 pieces/m2. The macrophyte and wood
distribution maps were overlaid to create a distribution of cover with unique polygons of
macrophyte and wood density combinations.
Topological Units
Using a GIS, the littoral area of each lake (from shore to the 2 m depth contour)
was divided into 2 m2 grids and the offshore area (>2 m depth) into 10 m2 grids with
respect to the georeferencing and hence some littoral grids overlapped the shoreline and
some offshore ones the littoral zone.
We systematically selected every 5th grid in the
nearshore area and every grid within the offshore area as a sample location, using the
centroid of each grid as a sample point. The substrate in the offshore zone was
assumed to be mud and hence only depth varied among offshore grids.
2
A range of fetch indices for each sample point was calculated using an Arc-Info
procedure. These included the minimum, maximum, and average fetch, the minimum,
maximum and average effective fetch, and the effective fetch in the strongest wind
direction (292.5o). The strongest wind direction was determined from a time-series of
wind speed and direction observations gathered at TLW.
Bathymetric maps with 0.5 m contour intervals were used to interpolate surface
maps of depth. The depth at each sample point was determined, in addition to the
habitat class at each point. The slope at each grid point was calculated by dividing the
depth by the minimum effective fetch, giving the percent rise from shore. This slope
index is the simplest computable and assumes a lake is conic with respect to each grid
point. The bathymetric survey data was acquired digitally by a unit of the Ontario Ministry
of Natural Resources (OMNR) and provided to the study.
Analysis
To reduce the number of variables cosnsidered, a principal component analysis
(PCA) was performed on the dataset from Little Turkey Lake. Variables that contributed
the least to the principal components variates were eliminated from further analysis
unless their exclusion meant that class of variable was excluded. Variables that were
selected were used in a discriminant function analysis (DFA) to predict the habitat
classification at each sampling point. DFA’s were performed to predict substrate
classification and cover (macrophyte and wood combined) classification separately
within the nearshore area and within the entire lake.
Using the substrate dataset, three different habitat classification schemes were
used for the DFA: 1) all the original substrate classes (Table 1); 2) general size
groupings of fine (sand and muck), medium (rubble and gravel), and large (bedrock and
boulder); 3) large particle sizes separate from all others (bedrock, boulder, and others).
This was done to determine if the success of predicting habitat classification from
topological lake units increased with a simplification of the classification scheme. To
further clarify the results from the DFA on the substrate dataset, approximately 1000
sample points were systematically selected (every Nth record) from each lake to provide
a roughly balanced sample and compiled into one dataset (n=4492). A DFA was
performed to predict classification of large particle size groupings to determine if the
same topological units can be used as predictive tools on all the lakes.
The success of classifying habitat categories using a DFA does not take into
account the possibility of chance agreement between field classifications and predicted
3
classifications. Cohen’s Kappa statistic (Cohen 1960) was used to determine the
success of predicting habitat classification, for both substrate and cover, corrected for
chance agreement. Kappa values can be interpreted as percentages or proportion of
agreement. A value of 0 indicates no agreement between predicted classification and
field classification whereas a value of 1 indicates perfect or 100% agreement.
Hakanson (1982) found that distribution of lake bottoms is governed by an energy,
slope, and form factor which can be expressed as a dynamic ratio (DR):
DR =
a
D
where a is the lake area (km2), and D is the mean depth (m). The DR can be used as a
tool to determine which lakes are susceptible to sediment disturbance by wind-driven
waves (Bachmann et al. 2000). We calculated the DR for each lake and determined
what proportion of the lakes are susceptible to sediment disturbance at least some of the
time from the empirical equation provided by Bachmann et al. (2000):
Percent of the lakebed disturbed some of the time = 12.4 + 109 x DR
The dynamic ratio (DR) was plotted against the percent classification success of each of
the lakes (substrate and cover) to determine any trends.
RESULTS
The number of variables used was reduced from nine to five based on the
contribution of each variable to the principle component variates (Table 3) thereby
reducing the number of fetch variables. Depth and slope were retained as they had
previously been shown to be useful variables. The variables selected for use in the DFA
were depth, slope, minimum effective fetch, maximum effective fetch, and effective fetch
in the strongest wind direction. Although slope values had a lower contribution to the
principal component variates (Table 3) than other variables, it was not eliminated from
further analyses. Slope is a contributing factor to sediment distribution (Cyr 1988) and
therefore would be a valuable variable to include in the DFA for predicting habitat
classification. All fetch variables (minimum, maximum, and average fetch) were
eliminated from further analyses even though they had a larger contribution to the
principal component variates. Instead, minimum and maximum effective fetch values
were used since they provide a more rounded measure of distance to shore. Average
effective fetch was excluded due to its redundancy with the other effective fetch values.
4
The results of the DFA for predicting substrate classification are summarized in
Table 4 for the littoral area and Table 5 for the entire lake. Classification success was
poor for most habitat classes, with the exception of the large size substrate classes. As
the substrate classes were simplified, the classification success of the DFA increased
(Table 4 and 5). Upper Batchawana Lake was the only lake that had high classification
success when using all of the original substrate classes (from Table 1). When the
presence of chance agreement between the field classifications and the predicted
classifications were removed through the use of the Kappa statistic, overall classification
success was reduced for each of the lakes. The high success rate of predicting
substrate in Upper Batchawana Lake was reduced by more than half, resulting in a low
classification success.
An example of a classification matrix from the DFA is given in Table 6 using Little
Turkey Lake and general size groupings of fine, medium, and large as habitat units. In
general, the success of predicting habitat units from topological lake units was higher for
large particle classes than the medium and smaller particle classes (Table 6). With an
increased simplification of habitat units, the classification success of predicting habitat
units using the topological variables increased, especially for Wishart Lake (Table 4).
This can be seen in Figure 1, an example of the distribution of factor scores from the
DFA of Little Turkey Lake using bedrock, boulder, and other classes. Although there is
overlap between the habitat classes, there are separate groupings for the other classes
and boulders (Figure 1).
The success of predicting substrate classification was lower for all the lakes than
for the individual lakes. There was approximately a 50% classification success for each
of the substrate classes (Table 7). This indicates there is no consistent pattern of
predicting substrate for all of the lakes and there are unique processes occurring within
individual lakes.
Similar to substrate classification, there was, in general, low success of predicting
classification of cover from the five topological units (Table 8). The highest success of
cover classification was for Little Turkey Lake (63%; Table 8). There was a decrease in
classification success after the Kappa analysis, on some lakes by half (Table 8).
The dynamic ratio of all lakes was under 0.80 (Table 9), falling in the range of
values which experience a linear decrease in the lake area disturbed at one time or
another (Bachmann et al. 2000). Bachmann showed that occasional extreme wind
events are what shape much of the sediment distribution and redistribution in larger
5
lakes. These study lakes do not appear to be greatly affected by wind driven forces at
least most of the time. The shallowest lake, Wishart Lake, has the greatest amount of
area (38.8%) that might be disturbed some of the time (Table 9).
DISCUSSION
Slope, depth, and a range of fetch indices, used to predict fish habitat distribution
in small lakes, are representative of wind driven energy inputs into lake systems. Slope
of the lake bottom will influence the amount of sliding of substrate particles, where steep
slopes significantly increase sediment distribution (Sly 1978; Rowan et al. 1992). The
amount of sorting of substrate particles is in part attributed to water depth. Shallow
areas are subjected to wave action causing sediment resuspension and may be areas of
either sediment erosion or transport. Fetch indices provide indirect measures of wind
exposure, or the potential total effect of waves. Exposure, where fetch and wind energy
are integrated, is considered to be a major factor affecting both plant and substrate
distributions in lakes (Keddy 1982) Although all three variables, slope, depth, and fetch,
play a role in fish habitat distribution in lakes, they do not act independently. Since their
contributions are interconnected, in this analysis we have incorporated all the variables
to investigate their combined contributions.
Although topological lake units have been found to affect substrate and vegetation
distribution, we found poor success of predicting fish habitat classification when using
slope, depth, and fetch indices as predictors. When attempting to predict all substrate
classes (Table 1), there was a low success rate, with a classification success of less
than 50% in more than half the lakes (Table 4). As the substrate classes became more
simplified, the ability to predict these classes from the topological features increased.
For two of the lakes, Upper Batchawana and Quinn, the success of classifying substrate
reached over 70% with the most generalized classes of bedrock, boulder, and others
(Table 4). As seen in the example classification matrix from the analysis of Turkey Lake
data (Table 6), the substrate category with the lowest classification success is fine
particles, followed by medium particles, and the large particles had a relatively higher
classification success of 74%. When these classification are grouped into bedrock,
boulder, and other, the habitat classes with poor classification success (fine and
medium) are grouped together as one, therefore lumping the error into one class and
raising the overall success rate of predicting habitat classes using topological units.
Although there were high success rates of predicting the large group classes, a large
6
portion of this success can be attributed to chance agreement (Table 4 and 5). The
success of predicting habitat classification in Upper Batchawana was reduced from
80+% to 40-% for the different categories of substrate classification (all classes; size
groups; large groups).
When the offshore areas were included in the analyses, predicted classification
success increased slightly over the situation when only the nearshore area was
considered. In the offshore area there is only one habitat class, the mud zone. A higher
success rate of predicting the mud zone was expected since depth was a primary
determinant but overall did not increase prediction success for other habitat classes.
In addition to low prediction successes for substrate classification, the topological
criteria for predicting membership in the most generalized classes were not transferable
between the lakes. For all of the lakes individually, there was a relatively high
classification success with prediction of the large group classification (bedrock, boulder,
and others). When samples from all the lakes were combined, the success of predicting
substrate classification from the topological units was only 50% (Table 7) lower than for
any of the lakes individually (Table 4) pointing to lake specific patterns of variable
associations.
Our attempts to predict cover (vegetation and wood combined) classifications in
lakes from topological units had low success rates, similar to substrate classification.
The same pattern was seen where some of the lakes had a high predicted classification
success rate and other lakes a low success rate. Again, when the offshore samples
were added to the analysis, the classification success increased slightly.
Hakanson (1982) developed the dynamic ratio (DR) which summarizes how the
areal distribution of lake bottom dominated by deposition, erosion and transport is
governed by an energy and morphometric factors. Bachmann et al. (2000) used this
dynamic ratio to create a morphometric index that calculates the potential for wind
disturbance of sediments in lakes. This index was applied to the lakes in this study and
we found that only small portions of their lakebeds are disturbed by wind driven forces
some of the time. The lake most affected by wind driven forces was the shallowest of
the lakes, Wishart, where less than 40% of its area is disturbed some of the time. The
area of these study lakes was too small for energy inputs to influence the distribution of
substrate. In smaller lakes there may be other processes, such as glacial geology,
determining distribution of fish habitat. In addition, ice scouring in winter may also be a
factor in these north temperate lakes.
7
The relationship between energy inputs into a lake and its substrate distribution is
hard to define and quantify, particularly in small lakes (Hakanson and Jansson 1983)
similar to the ones used in this study. Other studies in larger lakes have shown that
wind energy can transport larger particles (Sly 1978) and resuspend fine sediments (Cyr
1998). The results from of study lakes indicate a lower bound to potential impact of the
wind energy impacts on substrate distribution. Further, since these study lakes are on
the Canadian Shield and have low sediment inputs, the supply of sediment may also be
an important factor. In the future, it would be advantageous to repeat this analysis on
wider range of lake areas. This would clarify the influence of lake area and how it
constrains substrate distribution from energy inputs.
REFERENCES
Bachmann, R.W., M.V. Hoyer, and D.E. Canfield Jr. 2000. The potential for wave
disturbance in shallow Florida lakes. North American Lake Management Society.
16:281-291.
Cohen, J. 1960. A coefficient of agreement for nominal scales. Educational and
Psychological Measurement. 2091):37-47.
Cyr, H. 1998. Effects of wave disturbance and substrate slope on sediment
characteristics in the littoral zone of small lakes. Canadian Journal of Fisheries
and Aquatic Sciences. 55:967-976.
Franzin, W.G. Assiniboine River Study: Fish occurrence in relation to physical habitat
features. Detailed measurement of physical habitats in relation to scaleindependent processes. In: Randall, R.G., C.K. Minns, J.R.M. Kelso, C. Boston,
L. Carl, K. Clarke, B. Franzin, J. Hume, M. Ridgway, D. Scruton, K. Smokorowski,
and L. Stanfield. Field measurement of the productive capacity of freshwater fish
habitat – proceedings of a scoping workshop. 7, 8 November, 1999. Department
of Fisheries and Oceans.
Frezza, T.L. 2001. A comparison of observer repeatability and precision of sampling
using two different fish habitat assessment methods in lakes: transect sampling
and the littoral zone cruise. Master’s thesis. Trent University.
Hakanson, L. 1982. Lake bottom dynamics and morphometry; the dynamic ratio. Water
Resources Research. 18:1444-1450.
Hakanson, L. and M. Jansson. 1983. Principles of lake sedimentology. Springer-Verlag.
New York.
8
Keddy, P.A. 1982. Quantifying within-lake gradients of wave energy: Interrelationships of
wave energy, substrate particle size and shoreline plants in Axe Lake, Ontario.
Aquatic Botany. 14:41-58.
Kondolf, G.M. and S. Li. 1996. Accuracy and precision of selected stream habitat
estimates. North American Journal of Fisheries Management. 16:340-347.
Lester, N., K. Cornelisse, M. Stirling, and W. Dunlop. 1998. Fish habitat surveys on
Fisheries Assessment Unit Lakes: A review. Ontario Ministry of Natural
Resources.
Roper, B.B., and D.L. Scarnecchia. 1995. Observer repeatability in classifying habitat
types in stream surveys. North American Journal of Fisheries Management.
15:49-53.
Rowan, D.J., J. Kalff, and J.B. Rasmussen.1992. Estimating the mud deposition
boundary depth in lakes from wave theory. Canadian Journal of Fisheries and
Aquatic Sciences. 49:2490-2497.
Sly, P.G. 1978. Sedimentary processes in lakes. In: Lakes: Chemistry, geology, physics.
Editor: Lerman, A. New York. 65-89pp.
Stanfield, L.W. and M.L. Jones. 1998. A comparison of full-station visual and transectbased methods of conducting habitat surveys in support of habitat suitability index
models for Southern Ontario. North American Journal of Fisheries Management.
18:657-675.
9
Table 1. Definition of substrate types revised from Lester et al. (1998).
Substrate types
Definition
Bedrock
Exposed rock, no overburden
Boulder
>25 cm
Rubble
8 - 25 cm
Gravel
0.2 - 8 cm
Sand
<0.2 cm
Silt/Muck/Detritus
Inorganic, soft and decaying organic
Clay
Inorganic without structure
Other
Does not fit into any of the other categories
Table 2. Definitions and descriptions of macrophyte types (Frezza 2001).
Macrophytes
Description
Grass-like
Tall thin plants, slender leaves, no lateral
foliage
Carpet
Creating bottom cover
Cover rooted
Providing surface cover, rooted
Cover, no roots
Providing surface cover, not rooted
Leafy, extending to surface of water
Leafy foliage, extending vertically to the
water’s surface
Leafy, extending no further than half way
through water column
Leafy foliage, extending vertically only half
way up the water column
10
Table 3. Results from a principal components analysis using data from Little Turkey
Lake. The table lists the contributions from each variable to the principal component
variates.
Principal component variates
1
2
Eigenvalue
4.91
2.51
% Variance explained
48.54
35.22
Slope
0.20
-0.06
Depth
0.14
0.90
Min. fetch
0.09
0.96
Max. fetch
0.95
-0.03
Average fetch
0.96
0.19
Min. effective fetch
0.09
0.96
Max. effective fetch
0.98
0.03
Ave. effective fetch
0.96
0.19
Effective fetch 292o
0.48
0.24
11
Table 4. Classification success of a DFA to predict substrate classes in the littoral zone
area from five topological units. Included is the classification success once chance
agreement is removed (Kappa value).
Lake
Classification Success (%) (DFA/Kappa)
All classes
Size groups
Large groups
Little Turkey
34
14
44
16
52
11
Wishart
48
11
56
21
69
9
Lower Batchawana
50
18
60
17
56
19
Upper Batchawana
80
32
83
37
87
41
Quinn
67
9
72
12
73
7
Table 5. Classification success of a DFA to predict substrate classes in the entire (littoral
and offshore) lake from five topological units. Included is the classification success once
chance agreement is removed (Kappa value).
Lake
Classification Success (%) (DF/Kappa)
All classes
Size groups
Large groups
Little Turkey
48
37
56
43
61
44
Wishart
50
22
58
32
71
32
L. Batchawana
52
24
61
26
58
27
U. Batchawana
81
45
84
52
87
56
Quinn
68
32
74
38
74
36
12
Table 6. Classification matrix from the DFA indicating the success of predicting
substrate classes on Little Turkey Lake at sampling points using topological variables.
Substrate classes from field classification are in rows and the predicted classification in
columns. Numbers in the diagonal are successful predictions of habitat classification.
Fine
Medium
Large
% Correct
Fine
1023
755
954
37
Medium
393
663
227
52
Large
34
48
231
74
Total
1450
1466
1412
44
Table 7. Classification matrix from the DFA indicating the success of predicting
substrate classes at sampling points from all the lakes using topological variables.
Substrate classes from field classification are in rows and the predicted classification in
columns. Numbers in the diagonal are successful predictions of habitat classification.
Bedrock
Boulder
Other
% Correct
Bedrock
27
9
16
52
Boulder
42
110
87
46
Other
993
1136
2072
49
Total
1062
1255
2175
49
13
Table 8. Classification successes of a DFA to predict cover classes in the nearshore
zone and the entire lake from five topological units. Included is the classification
success once chance agreement is removed (Kappa value).
Lake
Classification Success (%) (DF/Kappa)
Nearshore
Entire lake
Little Turkey
63
36
65
52
Wishart
50
30
53
37
L. Batchawana
28
18
30
21
U. Batchawana
52
22
41
25
Quinn
18
13
23
20
Table 9. The area, mean depth, dynamic ratio, and percent of area disturbed some of
the time by wind driven forces for each lake.
Lake
Area (km2)
Mean depth (m)
Dynamic ratio
% of lake disturbed
some of the time
Little Turkey
0.21
6.17
0.07
20.6
Wishart
0.19
1.82
0.24
38.8
L. Batchawana
0.05
3.44
0.07
19.6
U. Batchawana
0.06
3.19
0.08
21.0
Quinn
0.07
4.33
0.06
19.0
14
-4
-2
8
7
6
5
4
3
2
1
0
-1 0
-2
-3
Bedrock
Boulder
Other
2
4
Factor 1 scores
Figure 1. Factor scores from the DFA of Little Turkey Lake using habitat units of
bedrock, boulder, and other. The classification success of each habitat unit was 75%
bedrock, 71% boulder, and 54% other. The total classification success was 55%.
15
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement