Time Residency Plots:  an Analysis of Common Dive Behavior
Statistics and Implications for TDR Studies


Senior Thesis


Patrick W. Robinson


Department of Ecology and Evolutionary Biology
University of California, Santa Cruz
Santa Cruz, CA  95064
Email:  click here


 

Introduction:

It is the goal of ecologists to study the way animals interact with their biotic and abiotic environments.  For many species this can be done via direct extended observation of behavior and associated location.  For example, the foraging ecology of the Sea Otter has been well documented by this procedure (e.g. Nolet et al. 1993).  Unfortunately, most marine mammal species are not as amenable to this method of study.  Their extreme lifestyle, including distant migrations and deep dives, precludes direct observation of their behavior; indirect means have therefore been employed.  Whaling logs (Gray 1882) and reports of entanglement in undersea cables (Heezen 1957) provided the first insight into the diving behavior of marine mammals, but very little useful information could be reconstructed from these disparate records.

In response to the shortcomings of previous studies, devices that could directly measure pressure, and therefore depth, while attached to a diving animal were created.  The first class of these time-depth recorders (TDR’s) was based on analog data collection on film spools (Kooyman 1976).  This method produced data that, although continuous, was difficult and extremely time consuming to assess quantitatively.  Also, calibration of these instruments was likely a major concern.  More recently, digital TDR’s, in conjunction with provided software, have allowed more in-depth analyses of diving behavior.  Digital TDR’s are made in a variety of forms including acoustic telemetry tags, satellite tags, and archival tags (Hooker and Baird 2001).  Despite the differences in the way data is transmitted in these tags, the data collected is almost always in the same format.  However, some tags do save memory by storing only several statistics about each dive (Fedak et al. 2001).  The common format records instantaneous depth and time at pre-defined intervals until the memory is filled or the unit is retrieved; some models provide additional information, including speed, conductivity, light, heart rate, etc., that can be useful in understanding dive behavior.

Many studies of marine animal ecology and physiology have depended upon the use of TDR dive records and, specifically, the use of mean dive depth, calculated as the mean of the maximum depths reached during each dive of a dive record.  Costa and Gales (2000) used mean dive depth to help explain variation in the mass-specific field metabolic rate seen in New Zealand sea lions.  Similar information was used to determine the diurnal dive behavior (Boveng et al. 1996, Lea et al. 2002) and the distinction of four types of foraging bouts (Boyd 1994) in Antarctic fur seals.  Mean dive depth aided in the description of seasonal shifts in dive behavior of the New Zealand fur seal (Mattlin et al. 1998).  Lastly, unique feeding strategies were explored as a possible explanation for intra-specific variation in mean dive depth of narwhals (Laidre et al. 2002).  The above examples are just a sample of relevant literature, but represent the diverse applications of TDR data analysis.

Because mean dive depth has been such a common statistic to report in diving studies, and is often involved in the explanation of physiology, foraging behavior, etc., a discussion of the exact meaning and implications of this statistic, as well as similar statistics, is justified.  Presumably, these basic dive statistics are employed to understand how animals use the vertical axis of the water column; however, it is not immediately clear whether the mean dive depth, based only on the maximum depth of each dive, accurately reflects where an animal spends the majority of its time.  Several studies have gone beyond this simple statistic to create plots of the proportion of dives to various depths (e.g. Mattlin et al. 1998).  A few additional studies have represented dive behavior in a third way:  plots of the proportion of time spent at various depths (Horning and Trillmich 1997, Laidre et al. 2002, Martin et.al. 1994).  It is the goal of this paper to use plots of the proportion of time spent at different depths, or time-residency plots, to determine how well commonly used dive statistics reflect where animals of different species and diving regimes spend their time.  In addition, other patterns found via time-residency plots at the level of the individual, population, and species are explored.

Methods:

Several researchers allowed the use of TDR dive records for this analysis.  Data was obtained from 81 animals, representing six pinniped species (Table 1).  

Table 1
New name Old name Species Common name Sex Year deployed Time between samples Length of diving record (days)
1 Nf5a.hex Neophoca cinerea Australian sea lion F 2002 4 12
2 NF7A.hex Neophoca cinerea Australian sea lion F 2002 4 13
3 NF8.hex Neophoca cinerea Australian sea lion F 2002 4 9
7 C2-tdr.hex Neophoca cinerea Australian sea lion F 1988 10 3
8 C2-2nd.hex Neophoca cinerea Australian sea lion F 1988 10 3
9 Ad-2nd.hex Neophoca cinerea Australian sea lion F 1988 10 3
10 P-tdr.hex Neophoca cinerea Australian sea lion F 1988 10 3
12 Tdr-14.hex Neophoca cinerea Australian sea lion F 1988 10 3
13 Tdr-ad.hex Neophoca cinerea Australian sea lion F 1988 10 3
14 Tdr-139.hex Neophoca cinerea Australian sea lion F 1990 10 4
15 Ad-tdr.hex Neophoca cinerea Australian sea lion F 1990 10 4
16 C2-neo.hex Neophoca cinerea Australian sea lion F 1990 10 3
17 Tdr-102.hex Neophoca cinerea Australian sea lion F 1990 10 3
19 138-2.hex Neophoca cinerea Australian sea lion F 1990 10 3
20 Tdr-141.hex Neophoca cinerea Australian sea lion F 1990 10 2
21 Tdr-51.hex Neophoca cinerea Australian sea lion F 1990 10 5
22 Tdr-c2.hex Neophoca cinerea Australian sea lion F 1991 10 3
23 n99f2 Neophoca cinerea Australian sea lion F 1999 5 15
24 n99f1 Neophoca cinerea Australian sea lion F 1999 5 17
25 n99k1.hex Neophoca cinerea Australian sea lion F 1999 5 14
26 n99k2.hex Neophoca cinerea Australian sea lion F 1999 5 12
27 n99k3.hex Neophoca cinerea Australian sea lion F 1999 5 10
28 n99k4.hex Neophoca cinerea Australian sea lion F 1999 5 18
29 n99k5.wch Neophoca cinerea Australian sea lion F 1999 5 10
30 n99f5.wch Neophoca cinerea Australian sea lion F 1999 5 22
31 n99f6.wch Neophoca cinerea Australian sea lion F 1999 5 16
32 99-373.wch Neophoca cinerea Australian sea lion F 2000 5 8
33 n00sm2.wch Neophoca cinerea Australian sea lion F 2000 5 29
34 n00sm3.wch Neophoca cinerea Australian sea lion F 2000 5 30
35 n00sm5.wch Neophoca cinerea Australian sea lion F 2000 5 30
36 99-036 Arctocephalus gazella Antarctic fur seal F 1999 5 7
38 99-039 Arctocephalus gazella Antarctic fur seal F 1999 5 5
39 99-040 Arctocephalus gazella Antarctic fur seal F 1999 5 4
40 99-041 Arctocephalus gazella Antarctic fur seal F 1999 5 3
41 99-042 Arctocephalus gazella Antarctic fur seal F 1999 5 3
42 99-043 Arctocephalus gazella Antarctic fur seal F 1999 5 5
43 99-044 Arctocephalus gazella Antarctic fur seal F 1999 5 3
44 99-045 Arctocephalus gazella Antarctic fur seal F 1999 5 3
45 99-046 Arctocephalus gazella Antarctic fur seal F 1999 5 3
46 99-047 Arctocephalus gazella Antarctic fur seal F 1999 5 3
47 99-049 Arctocephalus gazella Antarctic fur seal F 1999 5 2
48 99-062 Arctocephalus gazella Antarctic fur seal F 1999 5 8
49 99-067 Arctocephalus gazella Antarctic fur seal F 1999 5 3
50 99-069 Arctocephalus gazella Antarctic fur seal F 1999 5 6
51 99-079 Arctocephalus gazella Antarctic fur seal F 1999 5 14
52 99-081 Arctocephalus gazella Antarctic fur seal F 1999 5 9
54 E 40 Phocarctos hookeri New Zealand sea lion F 1997 5 10
55 E 41 Phocarctos hookeri New Zealand sea lion F 1997 5 11
56 E 42 Phocarctos hookeri New Zealand sea lion F 1997 5 7
57 E 43 Phocarctos hookeri New Zealand sea lion F 1997 5 11
58 E 45 Phocarctos hookeri New Zealand sea lion F 1997 5 8
59 E 46 Phocarctos hookeri New Zealand sea lion F 1997 5 11
60 E 48 Phocarctos hookeri New Zealand sea lion F 1997 5 11
61 E 49 Phocarctos hookeri New Zealand sea lion F 1997 5 11
62 E 52 Phocarctos hookeri New Zealand sea lion F 1997 5 6
63 E 53 Phocarctos hookeri New Zealand sea lion F 1997 5 10
64 E 55 Phocarctos hookeri New Zealand sea lion F 1997 5 3
65 E 56 Phocarctos hookeri New Zealand sea lion F 1997 5 7
66 E 58 Phocarctos hookeri New Zealand sea lion F 1997 5 7
67 blue Mirounga angustirostris Northern Elephant Seal F 1995 30 32
68 c115r Mirounga angustirostris Northern Elephant Seal F 1990 30 27
69 d197 Mirounga angustirostris Northern Elephant Seal F 1990 30 29
70 D318 Mirounga angustirostris Northern Elephant Seal F 1990 30 40
71 Deb Mirounga angustirostris Northern Elephant Seal F 1996 30 17
77 Bopp Mirounga angustirostris Northern Elephant Seal M 1997 30 60
78 c508 Mirounga angustirostris Northern Elephant Seal M 1991 40 97
79 Fuzz Mirounga angustirostris Northern Elephant Seal M 1989 30 80
80 joe Mirounga angustirostris Northern Elephant Seal M 1991 40 95
81 mario Mirounga angustirostris Northern Elephant Seal M 1992 30 74
87 Obi 1 Arctocephalus forsteri New Zealand fur seal F 1991 10 3
88 Obi 2 Arctocephalus forsteri New Zealand fur seal F 1992 10 10
89 Obi 3 Arctocephalus forsteri New Zealand fur seal F 1992 10 9
91 Obi 5 Arctocephalus forsteri New Zealand fur seal F 1993 10 7
92 Obi 6 Arctocephalus forsteri New Zealand fur seal F 1993 10 7
93 Obi 7 Arctocephalus forsteri New Zealand fur seal F 1993 10 11
94 Obi 8 Arctocephalus forsteri New Zealand fur seal F 1993 10 7
95 Obi 9 Arctocephalus forsteri New Zealand fur seal F 1993 10 12
96 Obi 10 Arctocephalus forsteri New Zealand fur seal F 1993 10 10
97 280019 Zalophus californianus California sea lion F 2003 5 12
98 0280016a Zalophus californianus California sea lion F 2003 5 12
99 00-066a Zalophus californianus California sea lion F 2003 5 10

















































































                            
Dive records were chosen to include variation in time, space, and foraging behavior.  Details of animal handling, deployment, and recapture can be found in previous studies based on the same data:  Australian sea lion (Costa and Gales 2003), New Zealand sea lion (Costa and Gales 2000, Crocker et al. 2001), New Zealand fur seal (Mattlin et al. 1998), and the northern elephant seal (Le Boeuf et al. 2000).  California sea lion handling was presumably done in a similar manner.

All data was obtained from several different models of Wildlife Computers TDR’s (Redmond, WA, USA) (Table 1).  These TDR’s differ in the amount of data storage capacity (i.e. number of readings), frequency of readings, and accuracy of readings.  It was assumed that there was no change in animal behavior due to tagging and that the tagged animals showed representative dive behavior (Boyd et al. 1991, Costa and Gentry 1986).  It was also assumed that the depth sensor accuracy was not affected by temperature, normally a cause of relatively small error (Hooker and Baird 2001).

Each dive record was processed by two distinct methods.  The first method, the “dive method,” attempted to replicate the procedure of most other studies of dive behavior.  Dive records were obtained in hexadecimal format (*.hex or *.wch extension) after previous download from the TDR’s by the original researcher.  Zero Offset correction, a DOS program provided by Wildlife Computers, was then used to correct for improper depth sensor calibration (i.e. identification of zero depth) and slight deviations through time of the zero-depth.  Depths less than four meters were considered to be at the surface, as has been done in other studies to account for the accuracy of the TDR unit (Costa and Gales 2003).  Zero Offset Correction was put in the “confirm-off” mode for all analyses to prevent user-bias.  However, visual confirmation and subsequent correction of the zero depth was completed when gross deviations from the presumed surface were made and when the program failed to find a zero depth, requiring user input.  This became particularly important during bouts of dives containing very short surface intervals.  Dive Analysis, another program provided by Wildlife Computers, was then used to analyze the binary output of Zero Offset Correction.  Depths less than four meters were considered to be at the surface and only dives reaching a depth of at least eight meters were analyzed.  This value was chosen because it is similar to the precedent set by other studies (Costa and Gales 2000, Crocker et al. 2001, Hindell and Pemberton 1997) while representing a depth of twice the accuracy of the TDR with the lowest depth-resolution (Hooker and Baird 2001, Lea et al. 2002).  The output of Dive Analysis shows simply a list of maximum depths, one for each dive.  The mean, median, and mode of these maximum dive depths were then calculated.  Next, the maximum dive depths were binned into 4m categories.  This value was chosen for consistency, as all TDR’s used have a depth resolution of at least 4m.  All depth categories were divided by the number of days of data they contained to allow for direct comparison between individuals and species.  Depth category was then plotted against average number of dives per day in a dive-frequency histogram.

The second data analysis method, the “depth method,” attempted to show how animals use the water column, without the need to distinguish individual dives.  Two additional programs provided by Wildlife computers, 3-M and HexDecode, were used to extract the complete dive record in ASCII format at the maximum resolution possible (between 4 and 40 seconds, depending on the TDR model and user-settings).  3-M and HexDecode split the dive records into multiple files, one for each day of diving activity, enabling Microsoft Excel to import the data.  Splitting the data was necessary due to an inherent limitation in the number of rows allowed in an Excel spreadsheet (65,536 rows).  Frequently, a single day of diving records would produce more than 17,000 rows of data.  It should also be noted that depth records over 1000m were mis-interpreted by Excel as percentages, requiring correction for further analysis.  Lastly, all depth readings for each day of diving were binned into 4m categories (Horning and Trillmich 1997).  Similar to the “dive method,” all depth readings shallower than 8m were removed with the intent of eliminating non-foraging behavior:  traveling dives and resting behavior (e.g. Costa and Gales 2003).  Plots of depth category against number of readings, equivalent to time, were created at the level of day, dive record, sex, population (for Australian sea lions), and species.  Division by sexes was necessary for Northern Elephant seals due to the distinct diving behaviors they employ (Le Boeuf et al. 2000).

Results

Dive Statistics:

To assess whether commonly used dive behavior statistics accurately reflect where an animal spends time in the water column, two analyses of the same data set were compared.  The “depth method” was directly compared to the “dive method” first by creating time residency plots, representing the average daily number of depth readings (analogous to time) recorded within each depth category, for each study animal.  Representative time residency plots of an individual animal from each species are presented in figures showing a continuum from a conspicuous peak to complete lack of peak (exponential decay) (figures 1a-7a).  

Figure 1a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of Australian sea lion record. One depth reading equivalent to 10s. (animal 7)



Figure 2a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of New Zealand sea lion record. One depth reading equivalent to 5s. (animal 58)


 
Figure 3a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of California sea lion record. One depth reading equivalent to 5s. (animal 99)


 
Figure 4a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of female northern elephant seal record.  One depth reading equivalent to 30s. (animal 68)


Figure 5a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of male northern elephant seal record.  One depth reading equivalent to 30s. (animal 80)


 
Figure 6a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of Antarctic fur seal record.  One depth reading equivalent to 5s. (animal 40)


 
Figure 7a:  Average number of TDR readings per day vs. Depth category (time-residency plot) showing typical shape of New Zealand fur seal record.  One depth reading equivalent to 10s. (animal 95)


Next, dive frequency plots, representing the average daily number of dives reaching a given maximum depth category, were created for each study animal.  The average, median, and mode of this data for each animal were calculated.  Regressions of modal time-residency, based on the “depth method”, against the average, median, and modal dive frequencies, based on the “dive method”, were completed for each species and each population, in the case of the Australian sea lions (figures 1c-7c) (Systat 10.2, SPSS Inc. Chicago, IL).  Regressions of similar slope and intercept to the dashed line (a 1:1 relationship), combined with statistically significant R2 values, were presumed to reflect high accuracy in the dive statistics for a species.


 
Figure 1c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all Australian sea lions.  Regression of “average:”  R2 = 0.940 (p<0.001). Regression of “median:”  R2 = 0.977 (p<0.001).  Regression of “mode:”  R2 = 0.988 (p<0.001).  Dashed line represents the ideal relationship between variables.


 
Figure 2c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all New Zealand sea lions.  Regression of “average:”  R2 = 0.056 (p=0.437). Regression of “median:”  R2 = 0.005 (p=0.826).  Regression of “mode:”  R2 = 0.971 (p<0.001).  Dashed line represents the ideal relationship between variables.


 
Figure 4c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all female northern elephant seals.  Regression of “average:”  R2 = 0.723 (p=0.068). Regression of “median:”  R2 = 0.692 (p=0.081).  Regression of “mode:”  R2 = 0.404 (p=0.249).  Dashed line represents the ideal relationship between variables.


 
Figure 5c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all male northern elephant seals.  Regression of “average:”  R2 = 0.675 (p=0.088).  Regression of “median:”  R2 = 0.758 (p=0.055).  Regression of “mode:”  R2 = 0.869 (p=0.021).  Dashed line represents the ideal relationship between variables.


 
Figure 6c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all Antarctic fur seals.  Regression of “average:”  R2 = 0.223 (p=0.143). Regression of “median:”  R2 = 0.327 (p=0.066).  Regression of “mode:”  R2 = 0.450 (p=0.024).  Dashed line represents the ideal relationship between variables.


 
Figure 7c:  Summary statistics of dive method vs. modal depth of depth method.  This plot contains values for all New Zealand fur seals.  Regression of “average:”  R2 = 0.186 (p=0.287). Regression of “median:”  R2 = 0.174 (p=0.305).  Regression of “mode:”  R2 = 0.025 (p<0.706).  Dashed line represents the ideal relationship between variables.



All three dive statistics for the Australian sea lion were found to be accurate; the mode had the highest correlation.  Similarly, the male northern elephant seal showed strong correlations, but only the mode was found to be accurate.  Again, only the mode was found to be accurate for both the Antarctic fur seal and New Zealand sea lion.  Interestingly, the average and median of the New Zealand fur seal data were highly inaccurate.  All dive statistics for the female northern elephant seal and New Zealand fur seal were highly inaccurate.  Dive statistics within these two species were very consistent, which could yield a false sense of accuracy if not compared with a time residency plot.  California sea lion data was not included in this dive statistic analysis due to insufficient software.

Time Residency Plots:

The dive statistics analysis provided the opportunity to study the diving behavior of the species via a qualitative examination of time residency plots.  In addition to characteristic time residency plots for individuals (figures 1a-7a), time residency plots averaged over species, were created. (figures 1b-7b and 1d-1f).  


 
Figure 1b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 30 Australian sea lion dive records, standardized to 10s per depth reading. (animals 1-35).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.



 
Figure 2b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 13 New Zealand sea lion dive records. One depth reading equivalent to 5s. (animals 54-66).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 3b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 3 California sea lion dive records. One depth reading equivalent to 5s. (animals 97-99).  Software limitations prevented analysis via the dive method.


 
Figure 4b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 5 female northern elephant seal dive records. One depth reading equivalent to 30s. (animals 67-71) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 5b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 5 male northern elephant seal dive records, standardized to 30s per depth reading. (animals 77-81).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 6b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 16 Antarctic fur seal dive records. One depth reading equivalent to 5s. (animals 36-53). The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 7b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 9 New Zealand fur seal dive records. One depth reading equivalent to 10s.  Plot is truncated due to lack of accurate TDR depth records deeper than 234m (see Costa and Gales 2000). (animals 87-96) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 1d:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of 7 Australian sea lion dive records, all from the same population.  One depth reading equivalent to 10s. (animals 7-13) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 1e:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of 8 Australian sea lion dive records, all from the same population.  One depth reading equivalent to 10s. (animals 14-21) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.

 
Figure 1f:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of 5 Australian sea lion dive records, all from the same population.  One depth reading equivalent to 5s. (animals 25-29) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.

Several interesting patterns were observed on the level of the individual, population, and species (Hooker and Baird 2001).  Many individuals from benthically foraging species (e.g. sea lions and male elephant seals) produced time residency plots containing multiple distinct peaks, indicating relatively large amounts of time were spent at very particular depth categories.  This showed that multiple prey resources may be taken advantage of on a daily basis.  Also, the peaks of an individual’s time residency plot often occurred at the same depth for multiple days, even if interrupted by time spent on land; therefore, preference for a particular depth category is often fixed for several days at a time, indicating a consistent use a prey resource (figure 2d).

 
Figure 2d:  Number of TDR readings per day vs. Depth category (time-residency plot).  Lines represent consecutive days of diving for a New Zealand sea lion.  (animal 57).

At the population level (animals 14-21 and 25-29), individual Australian sea lions produced an interesting pattern:  the time residency peaks were very distinct and usually not shared with other animals.  The sea lions were essentially spending the majority of their time at staggered depths, indicating the possibility of territoriality in foraging sites (figure 1g).

 
Figure 1g:  Average number of TDR readings per day vs. Depth category (time-residency plot).  Each line represents one Australian sea lion dive records; all individuals are from the same population.  One depth reading equivalent to 10s. (animals 14-21).


At the species level, the time residency plots fall into three distinct categories.  First, TDR data from both species of fur seals produced time residency plots with exponential decay and complete lack of distinct peaks (figures 6b and 7b).  

 
Figure 6b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 16 Antarctic fur seal dive records. One depth reading equivalent to 5s. (animals 36-53).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 7b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 9 New Zealand fur seal dive records. One depth reading equivalent to 10s.  Plot is truncated due to lack of accurate TDR depth records deeper than 234m (see Costa and Gales 2000). (animals 87-96).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


These animals are using a large portion of the water column but preferentially spending their time at shallow depths.  Next, the three species of sea lions produced time residency plots with one or more distinct peaks (figures 1b-3b).  

 
Figure 1b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 30 Australian sea lion dive records, standardized to 10s per depth reading. (animals 1-35).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 2b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 13 New Zealand sea lion dive records. One depth reading equivalent to 5s. (animals 54-66) .  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 3b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 3 California sea lion dive records. One depth reading equivalent to 5s. (animals 97-99).  Software limitations prevented analysis via the dive method.


These animals were using a large portion of the water column as well, but focusing their effort in a small number of depth categories.  Lastly, the time residency plot produced from elephant seal TDR records were intermediate in shape (figures 4b and 5b).  

 
Figure 4b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 5 female northern elephant seal dive records. One depth reading equivalent to 30s. (animals 67-71).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 5b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 5 male northern elephant seal dive records, standardized to 30s per depth reading. (animals 77-81).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


These plots contained numerous small peaks, but none were as conspicuous as the sea lions’.  This result was not affected by the relatively longer sampling period of the elephant seals (up to 97 days), as analysis of subsets of the data revealed a similar pattern.

Discussion:

Studies of diving behavior have provided great insight into the ecology of marine vertebrates.  The depth(s) at which these predators forage is presumably directly related to their physiological constraints and ecological demands (Boyd 1997, Shreer et al. 2001).  Diving animals must decide, within their physiological abilities, to dive shallow to exploit dense, but ephemeral, prey patches, or to dive to greater depths, using more energy, to feed on richer food sources that are consistent, but relatively rare. (Costa 1988).  TDR dive records are, therefore, a valuable tool in understanding this ecology.

Dive Statistics:

    One of the inherent problems with using the dive method to obtain basic dive behavior statistics is the need to provide a single depth as a cutoff to distinguish dives from each other and from non-foraging activities, such as travel or rest.  In previous studies, this cutoff depth has ranged from 2m to 50m (Boyd and Croxall 1992, Schreer and Testa 1995).  This variation potentially has very large, and unpredictable, impacts on resulting dive statistics and prevents direct comparison between studies.  Specifically, a deep cutoff value for an epipelagic forager will drastically inflate the mean dive depth.  Conversely, a shallow cutoff for a benthic forager will decrease the mean dive depth.

The vast majority of studies involving marine vertebrate dive behavior provide average dive depth as the sole depth statistic reported (e.g. Lea et al. 2002), despite recommendations that modal dive depth may be of greater use in assessing the use of the water column at the level of the individual and species (Gales and Mattlin 1997, Hooker and Baird 2001, Machlis et al. 1985).  Several other studies report various combinations of average, median, and modal dive depths with limited comment on their respective meanings in relation to dive behavior (e.g. Costa and Gales 2003, Hindell and Pemberton 1997, Mattlin et al. 1998).

Of the three dive behavior statistics studied, the mode was clearly the most accurate with respect to revealing the depth an animal spends the majority of its time; two thirds of the species’ most frequented depths were identified properly.  The more commonly reported statistic, the average dive depth, showed very low accuracy, identifying the appropriate depth in only one species.  It should be noted that average dive depths deemed inaccurate for two species corresponded to smaller peaks in the time residency plot.  This highlights the importance of moving beyond a single number to classify the depths to which an animal or species dives, thus preventing erroneous (average dive depth) or incomplete (modal dive depth) information.  

Time Residency Plots:

One of the concerns about using time residency plots is the inherent increase in time spent in the upper water column due simply to transit to deeper depths.  However, information in this portion of the graph is not lost due to the consistent speed most marine mammals swim at while in transit; gliding allows animals to maintain relatively constant rates of transit (Williams et al. 2000), thus enabling time residency plots to give more useful details about a dive.  For example, non-modal peaks in time residency plots likely represent a non-random use of particular depths, not just shifts in speed due to swimming irregularities.
Time residency plots can show where in the water column an animal is spending most of its time, illuminating possible ecological interactions at a reasonably fine scale in comparison to the gross scale achieved by dive depth frequency plots.  Several previous studies have utilized time residency plots, or slight variations of them.  First, a pair of studies involving the diving behavior of narwhals were the first to address time spent at depth independently of dives (Laidre et al. 2002, Martin et al. 1994).  Despite being divided into unequal depth categories, the plots revealed unique foraging strategies.  Next, a study of diving ontogeny in Galapagos fur seals was the first to produce time residency plots, or “percent of recording time spent at various depths,” of the form seen in this analysis (Horning and Trillmich 1997).  It was concluded that individuals gradually increase their maximum attainable depth and their total time diving with age.  Interestingly, the plot for adult female Galapagos fur seals was very similar to the plots for the Antarctic fur seal and New Zealand fur seal, indicating an epipelagic feeding behavior (figures 6b and 7b).  

 
Figure 6b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 16 Antarctic fur seal dive records. One depth reading equivalent to 5s. (animals 36-53).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


 
Figure 7b:  Average number of TDR readings per day vs. Depth category (time-residency plot).  This plot is an average of all 9 New Zealand fur seal dive records. One depth reading equivalent to 10s.  Plot is truncated due to lack of accurate TDR depth records deeper than 234m (see Costa and Gales 2000). (animals 87-96).  The solid and dashed lines show the grand average and modal dive depth, respectively, produced via the dive method.


Lastly, time residency plots were created from pantropical spotted dolphin dive data to show a consistent use of the water column among individuals of the same population (Baird et al. 2001).  Clearly these conclusions would not have been easily reached without the use of time-residency plots; however, more profound conclusions can be drawn from these plots.

Ecological Implications:

At the population level, Australian sea lions were found to exploit very particular depth ranges.  These peaks were at different depths for different members of these populations (e.g. animals 14-21), representing a stratified distribution of effort at the population level (figure 1g).  

 
Figure 1g:  Average number of TDR readings per day vs. Depth category (time-residency plot).  Each line represents one Australian sea lion dive records; all individuals are from the same population.  One depth reading equivalent to 10s. (animals 14-21).


Because these animals are benthic foragers preying on predictable, non-dense resources (Costa and Gales 2003), dividing up the hunting ground is likely the most efficient method for obtaining suitable amounts of prey.  It is unlikely that this stratification is simply a result of different physiological capacities between animals, as most animals occasionally dove far deeper than their peak depth.  This pattern is in marked contrast with a non central place benthic forager:  the male northern elephant seal.  The time residency plots for these animals indicate a much broader overlap in utilized depth regions.  This is may be a result of the much larger areas available to the elephant seal, resulting in very little competition for prey along a depth gradient.

Lactating Antarctic fur seals provide a unique opportunity to study feeding ecology in relation to the health of a population.  This species feeds largely on dense patches of vertically migrating krill (Croxall et al. 1985).  The abundance of krill is variable between years, directly affecting the success of Antarctic fur seals (Boyd 1996).  A year of low krill density can have variable effects on a population.  Years of extremely low krill density can be manifested in an increased pup mortality rate (Costa et al. 1989).  However, females can compensate for slightly increased foraging costs behaviorally without negative impact to their pups (Boyd 1997).  Therefore, the use of time residency plots, via a multi-year study, could show trends in “feeding stress” and krill populations that would otherwise remain undetected.

Future Analysis:

Time residency plots allow diving behaviors and favored depths to be distinguished at a higher resolution, and with more accuracy, than methods commonly in use.  The field would greatly benefit from a re-analysis of abundant pre-existing TDR data sets.  These analyses have the potential to reveal changes in behavior at a variety of scales as well as show the assortment of behaviors already displayed within a population (Nordoy 1995).  The use of time residency plots in conjunction with new technologies, such as satellite tags (e.g. Le Boeuf et al. 2000) and stomach temperature monitors (Andrews 1998), will provide a greater understanding of the ecological and possible human interactions (Boyd 1997) that affect marine vertebrates.

Acknowledgements:

    I would like to thank Dan Costa for allowing the use of his TDR data and for his consistent help and input, Carey Kuhn for the use of her TDR data, and my parents for their support.
 

References:


Andrews, R.D. (1998) Instrumentation for the remote monitoring of physiological and behavioral variables. Journal of Applied Physiology, 85, 1974-81.
        
Baird, R.W., Ligon, A.D., Hooker, S.K., & Gorgone, A.M. (2001) Subsurface and nighttime behavior of pantropical spotted dolphins in Hawai’i.  Canadian Journal of Zoology, 79, 988-996.
        
Boveng, P.L., Walker, B.G., & Bengtson, J.L. (1996) Variability in antarctic fur seal dive data: Implications for TDR studies. Marine Mammal Science, 12, 543-554.

Boyd, I.L. (1996) Temporal Scales of Foraging in a Marine Predator. Ecology, 77, 426-434.

Boyd, I.L. (1997) The behavioural and physiological ecology of diving. Trends in Ecology & Evolution, 12, 213-217.

Boyd, I.L., Arnould, J.P.Y., Barton, T., & Croxall, J.P. (1994)  Foraging behaviour of Antarctic fur seals during periods of contrasting prey abundance.  Journal of Animal Ecology, 63, 703-713.

Boyd, I.L. & Croxall, J.P. (1992) Diving behaviour of lactating antarctic fur seals. Canadian Journal of Zoology, 70, 919-928.

Boyd, I.L., Lunn, N.J., & Barton, T. (1991) Time Budgets and Foraging Characteristics of Lactating Antarctic Fur Seals. Journal of Animal Ecology, 60, 577-592.

Costa, D.P. 1988. Methods for studying the energetics of freely diving animals. Canadian Journal of Zoology 66:45-52.

Costa, D.P., Croxall, J.P., & Duck, C.D. (1989) Foraging energetics of antarctic fur seals in relation to changes in prey availability. Ecology, 70, 596-606.        

Costa, D.P. & Gales, N.J. (2003)  Energetics of a benthic diver:  seasonal foraging ecology of the Australian sea lion, Neophoca cinerea.  Ecological Monographs, 73(1), 27-43.
        
Costa, D.P. & Gales, N.J. (2000) Foraging energetics and diving behavior of lactating New Zealand sea lions, Phocarctos hookeri. Journal of Experimental Biology, 203, 3655-3665.
        
Costa, D.P., Gentry, R.L., (1986) Reproductive energetics of the northern fur seal.  In Fur Seals:  Maternal Strategies on Land and at Sea.  Edited by R.L. Gentry and G.L. Kooyman.  Princeton University Press, NJ  pp. 79-101.

Crocker, D.E., Gales, N.J., & Costa, D.P. (2001) Swimming speed and foraging strategies of New Zealand sea lions (Phocarctos hookeri). Journal of Zoology (London), 254, 267-277.
        
Croxall, J.P., Everson, I., Kooyman, G.L., Ricketts, C., & Davis, R.W. (1985).  Fur seal diving behaviour in relation to vertical distribution of krill.  Journal of Animal Ecology, 54, 1-8.
        
Fedak, M.A., Lovell, P., & Grant, S.M. (2001) Two approaches to compressing and interpreting time-depth information as collected by time-depth recorders and satellite-linked data recorders. Marine Mammal Science, 17, 94-110.

Gales, N.J. & Mattlin, R.H. (1997) Summer diving behaviour of lactating New Zealand sea lions, Phocarctos hookeri. Canadian Journal of Zoology, 75, 1695-1706.

Gray, D. (1882) Notes on the characteristics and habits of the bottlenose whale (Hyperoodon rostratus). Proceedings of the Zoological Society of London, 1882, 726–731.

Heezen, B.C. (1957) Whales entangled in deep sea cables. Deep-Sea Research, 4, 105–115.

Hindell, M.A. & Pemberton, D. (1997) Successful use of a translocation program to investigate diving behavior in a male Australian fur seal, Arctocephalus pusillus doriferus. Marine Mammal Science, 13, 219-228.

Hooker, S.K. & Baird, R.W. (2001) Diving and ranging behaviour of odontocetes: a methodological review and critique. Mammal Review, 31, 81-105.
            
Horning, M., Trillmich, F. (1997) Ontogeny of Diving Behaviour in the Galapagos Fur Seal. Behaviour 134: 1211-1257.

Kooyman, G.L., Gentry, R.L., & Urquhart, D.L. (1976) Northern fur seal diving behavior: A new approach to its study. Science, 193, 411-412.
        
Laidre, K.L., Heide-Jorgensen, M.P., & Dietz, R. (2002) Diving behaviour of narwhals (Monodon monoceros) at two coastal localities in the Canadian High Arctic. Canadian Journal of Zoology, 80, 624-635.

Le Boeuf, B.J., Crocker, D.E., Costa, D.P., Blackwell, S.B., Webb, P.M., & Houser, D.S. (2000) Foraging ecology of northern elephant seals. Ecological Monographs, 70, 353-382.

Lea, M., Hindell, M., and Guinet, C. (2002) Variability in the diving activity of Antarctic fur seals, Arctocephalus gazella, at Iles Kerguelen.  Polar Biology, 25, 269-279.

Machlis, L., Dodd, P.W.D. & Fentress, J.C. (1985) The pooling fallacy: problems arising when individuals contribute more than one observation to the data set. Zietschrift Fur Tierpsychologie, 68, 201–214.

Martin, A.R., Kingley, M.C.S., & Ramsay, M.A. (1994)  Diving behaviour of narwhals (Monodon monoceros) on their summer grounds.  Canadian Journal of Zoology, 72, 118-125.

Mattlin, R. H., N. J. Gales and D. P. Costa. 1998. Seasonal dive behaviour of lactating New Zealand fur seals (Arctocephalus forsteri). Can. J. Zool. 76(2):350-360.
Nolet, B.A., Wansink, D.E.H., Kruuk, H. (1993)  Diving of otters (Lutra lutra) in a marine habitat: use of depths by a single-prey loader. J. of Animal Ecology 62,  22-32.
Nordoy, E.S., Folkow, L., & Blix, A.S. (1995) Distribution and diving behaviour of crabeater seals (Lobodon carcinophagus) off Queen Maud Land. Polar Biology, 15, 261-268.

Schreer, J.F., Kovacs, K.M., & O'Hara Hines, R.J. (2001) Comparative diving patterns of pinnipeds and seabirds. Ecological Monographs, 71, 137-162.
        
Schreer, J.F. & Testa, J.W. (1995) Statistical classification of diving behavior. Marine Mammal Science, 11, 85-93.
                    
Williams, T.M., Davis, R.W., Fuiman, L.A., Francis, J., Le Boeuf, B.J., Horning, M., Calambokidis, J., & Croll, D.A. (2000) Sink or swim: strategies for cost-efficient diving by marine mammals. Science, 288, 133-136.