WaterSurface ver. 2.2
2019-Nov-1 update
Kentaro Q. Sakamoto
Overview
To analyze the behaviours of diving animals, time-depth recorder is useful devices, which continuously measure the depth by pressure sensor. When analyzing the depth data, one of the first concerns is that the temperature drift of the pressure sensor (i.e. change of depth value in relation with temperature). This artifact is particularly critical for the depth data from diving mammals and birds. Diving mammals and birds repeatedly submerge in water and return to water surface. Thus, to reveal the characteristics of dives, identification of the time points for the depth of 0 m (water surface) is essential.
The scatter plot (Fig. 1) shows that the relationship between temperature and depth data recorded at 1 sec interval by diving seabird. The plot indicates that the bird dived to 15 or 30 m and the ambient temperature during dive was 10-13 degree. When I enlarge the plot around the the depth of 0 m (Fig. 2), there are a lot of data points indicating the water surface. You will see the values of these data points gradually increase with the temperature from 10 to 30 degree, indicating the temperature drift of pressure sensor. In addition, the values of depth ranges from 1.0 to 1.5 m, although theses values represent water surface and should be zero.
This program corrects the artificial error of depth data, causing temperature drift, and sets the proper offset value for depth records.
Agreements
This program was developed with the assumption of the data recorded by D2GT data logger (Little Leonardo), although it may be applicable for other recorders.
This program assumes the depth data recorded by the animals which stay on water surface for a certain amount of time (mammals and birds etc.).
The author takes no responsibility for its use or misuse.
Preparation
Install Ethographer.
Make sure that you have both depth and temperature data recorded simultaneously at the same sampling frequency.
How to run WaterSurface
Run IGOR Pro, and then run Ethographer (you will choose Misc -> Run Ethographer from the IGOR Pro menu). You will see "Ethographer" in the IGOR Pro menu.
Open WaterSurface.ipf file. Alternatively you may copy Watersurface.ipf file to "User Procedure" folder, and include "waterSurface". You will see "WaterSurface" in the IGOR Pro menu.
In the IGOR Pro menu, you will choose WaterSurface -> Run. By following the instruction, choose depth and temperature data.
Diagnosis figure will appear (Fig. 3). Please confirm whether the program properly corrects the depth data. You will confirm following two points. The successful results will appear something like Fig. 3.
Red lines in the top and bottom panels adequately fit the distribution of bar graph.
Red line in the middle panel properly works as regression line for the data points representing water surface.
The program will outputs "D" wave as the corrected depth data.
Algorithm
Calculate the histogram for the depth data. In case of the data from the animals which stay on water surface for a certain amount of period, the mode will corresponds approximately to the depth of 0 m.
Fit gaussian distribution to the histogram (Fig. 3, top), and extract the depth data in the range of mean +- 3SD, which represent water surface.
Perform regression analysis for the extracted data to examine the relationship between depth and temperature (Fig. 3, middle). The regression line would represents the degree of temperature drift of pressure sensor, because these data points should indicate water surface. Then all depth data are corrected according to the regression line.
Update the histogram by using the corrected depth data. Fit the gaussian distribution to the histogram again (Fig. 3, bottom). It is noteworthy that the variance of the distribution become smaller than that shown in the top panel of Fig.3. The data points scattered around 0 m of depth can be interpreted as representing water surface (Fig. 3, bottom). Thus it can be assumed that this noise (variance around zero) was artificially caused by the recorder.
For the data shown in the bottom panel of Fig. 3, convert the value of the depth data in the range of mean +- 3SD into zero.
The value of mean +- 3SD indicates the "Error by Logger" and will be printed in the command window.