Dive Clustering and Bouts
Philippine Chambault
2023-01-19
Background: Dive clustering to identify dive types and investigation of dive sequences (bouts). Analysis on the high resolution dataset (n=5). Clustering analysis based on k=4 clusters.
1 Clustering analysis
##
## 1 2 3 4
## 61318 41091 47972 99762## # A tibble: 4 × 7
## cluster maxdep meandep dur desc_rate asc_rate bot_dur
## <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1 42.9 27.8 1.01 1.90 1.54 14.4
## 2 2 65.2 45.6 2.29 1.44 1.51 56.9
## 3 3 154. 93.6 3.05 1.83 1.84 13.3
## 4 4 25.0 17.8 1.16 0.866 0.787 12.4Type 1: moderate dives (<50 m) with the longest bottom duration (close to 1 min, U-shape?) and fastest descent and ascent rates.
Type 2: moderate dives (60 m, 2 min).
Type 3: deepest and longest dives with short vertical speed and moderate bottom duration (V-shape?).
Type 4: shallow ans short dives (<30 m and ~1 min).
The 4 clusters have been selected arbitrary because the R package factominR does not work with large datasets like this one, so I could not use an alternative package or function to identify the optimal number of clusters needed, but 4 seems realistic and large enough to capture to different dive types.
2 Cluster proportions
2.1 All (position=fill)
Figure 2.1: Histograms of dive frequency across months and dive type.
Type 4 and 1 (shallow and short dives) are becoming less and less frequent when approaching winter, while long and deep dives (Type 2 and 3) become more frequent over months.
2.2 All (position=dodge)
Figure 2.2: Histograms of dive frequency across months and dive type.
Easier to see the trends with the argument “position=dodge”.
2.3 Per ID (position=fill)
Figure 2.3: Indiviual histograms of dive frequency across months and dive type.
Similar trend across individuals.
2.4 Per id (position=dodge)
Figure 2.4: Indiviual histograms of dive frequency across months and dive type.
3 Dive clusters
3.1 Cluster visualization
Figure 3.1: Clusters identified using K mean
4 clusters seems realistic, but 3 might be enough.
3.2 Max depth
Figure 3.2: Boxplot of the maximum depth according to the dive clusters. Red dots are the means.
3.3 Mean depth
Figure 3.3: Boxplot of the mean depth according to the dive clusters.
3.4 Duration
Figure 3.4: Boxplot of the dive duration according to the dive clusters.
3.5 Descent rate
Figure 3.5: Boxplot of the descent rate according to the dive clusters.
3.6 Ascent rate
Figure 3.6: Boxplot of the ascent rate according to the dive clusters.
3.7 Bottom duration
Figure 3.7: Boxplot of the bottom duration according to the dive clusters.
4 Dive profiles according to dive type
4.1 ID 93100
Figure 4.1: Dive profile of individual 93100 over time and according to dive type.
4.2 ID 22849b
Figure 4.2: Dive profile of individual 22849b over time and according to dive type.
4.3 ID 22850b
Figure 4.3: Dive profile of individual 22849b over time and according to dive type.
4.4 ID 27262
Figure 4.4: Dive profile of individual 22849b over time and according to dive type.
4.5 ID 27262b
Figure 4.5: Dive profile of individual 22849b over time and according to dive type.
5 Zoom in dive profiles
5.1 ID 93100: 30 Jul
Figure 5.1: Dive profile of individual 93100 over time and according to dive type. Grey and beige shades refer to night and day, respectively.
5.2 ID 93100: 4 Oct
Figure 5.2: Dive profile of individual 93100 over time and according to dive type.
5.3 ID 93100: 3 Jan
Figure 5.3: Dive profile of individual 93100 over time and according to dive type.
5.4 ID 22849b: 26 Jul
Figure 5.4: Dive profile of individual 22849b over time and according to dive type.
5.5 ID 22849b: 11 Sep
Figure 5.5: Dive profile of individual 22849b over time and according to dive type.
5.6 ID 22849b: 5 Dec
Figure 5.6: Dive profile of individual 22849b over time and according to dive type.
5.7 ID 27262b: 4 Dec
Figure 5.7: Dive profile of individual 27262b over time and according to dive type.