High Resolution Acceleration

studies with SeaTag-MOD

23JUN13

By Marco Flagg

Acceleration studies can provide a variety of information on animal behavior, including for example feeding and hunting patterns, overall levels of activity or micro-motion reconstruction of the diving behavior of marine mammals.

Yet, such studies were generally beyond the realm of pop-up satellite tags due to limitations of power, memory and high data volumes incompatible with recovery via the low-bandwidth ARGOS satellite system.

Working with the research group of Prof. Nelson Ehrhardt, Mark Fitchett and Bruce Pohlot at the Rosenstiel School of Marine & Atmospheric Sciences (U Miami) in support of their billfish studies, Desert Star Systems has implemented support for high-speed / high-resolution acceleration (HRA) studies using the SeaTag-MOD PSAT and archival tag. HRA capability is available with a tag firmware upgrade to V2.20 or later.

A SeaTag-MOD experiencing Zero-G over the Cachagua valley. 

(photo credit: Joshua Reed-Doyle)

The practical implementation of high-resolution acceleration studies on a PSAT tag is made possible in SeaTag-MOD by three principal advances:

  • Drastically reduced power consumption: SeaTag-MOD uses an integrated 3-axis accelerometer chip with it’s own controller and a built-in FIFO (first-in-first-out) data buffer. Operating autonomously and holding up to 32 3- axis acceleration measurements, the chip needs to be serviced only infrequently by the tag’s main processor to offload the accumulated data in bulk.

  • Large memory capacity: Using a micro-SD memory card, standard SeaTag-MOD tags currently come equipped with 2GB of memory and can be outfitted on special request up to the maximum available micro-SD card capacity (currently 32 GB with 16 GB tested). At an acceleration sampling interval of 0.64 seconds, the standard card capacity is enough for about 3 ¾ years of operation. This can be used for high-resolution studies if the tag is physically recovered to offload data.

  • On-tag data compression for ARGOS compatibility: Using firmware based data compression, SeaTag-MOD can compress data volume by large factors. For example, G-force statistics providing the standard deviation of acceleration in 2 minute intervals, and based on original data sampled in 0.64 sec intervals (to capture short animal activity bursts), compress down to 30 ARGOS packets per day of observation.

A look at raw acceleration data: 3-axis measurements and G-Force

Figure 1: Definition of the three axes of SeaTag-MOD

SeaTag-MOD measures acceleration in the three spatial dimensions (X, Y, Z), corresponding to the three axes of the tag as shown in figure 1.

When looking at the data of a single axis only, one will see the combined effect of static acceleration (Earth gravity, 1G = 9.81 m/sec^2) and dynamic acceleration (change of speed in m/s^2) on that axis.

For example, when holding the tag upright, Earth gravity will cause the Z-axis to read +1G. By rotating the tag until upside down, the Zaxis reading will transition through 0G (tag held horizontally) to -1G (tag fully inverted).

Yet, a 1G Z-axis reading is also the result of holding the tag horizontally and throwing it firmly at an acceleration rate of 9.81 m/sec^2, the equivalent of going from zero to 100 km/h (60 mph) in 2.8 seconds. Because of the large effect of earth gravity, individual axis readings are useful to determine animal orientation of a tag is rigidly mounted to the body of the animal.

The effect of Earth gravity however can be easily removed by looking at the total acceleration rather than axis specific acceleration: Total acceleration = √(x^2+y^2+z^2). Total acceleration is always 1G, independent of the orientation of the tag, as illustrated in a tag drop test in figure 2.

In summary, a look at the value of the three individual axes (X, Y, Z) is primarily of use to judge tag (animal) orientation, while the total G-Force is a metric useful to measure animal activity, that is dynamic acceleration.

Figure 2: A SeaTag-MOD is dropped from about 6m, then carried up the stairs, then dropped again. X, Y and Z axis readings vary greatly depending on tag orientation, but total acceleration is centered at 1G Earth gravity and drops to 0G when the tag is in free fall. Acceleration measurement interval: 0.08 sec, total duration: 50 sec.

Use of raw HRA data

SeaTag-MOD can be configured to store the raw, 3-axis acceleration data on its memory card. Four 3-axis acceleration readings are combined into a single 32-byte data packet of type SDPT_HRA_TS. However, with HRA sampling rates in the sub-second region for perhaps most applications, the volume of this data is much too high for satellite transmission. So, SDPT_HRA_TS packets are only available for direct download from the tag following its physical recovery. Use this capability for applications such as detailed motion or behavioral studies of marine mammals with physical tag recovery.

Figure 3: Data log with SDPT_HRA_TS raw acceleration data (orange), and SDPT_MODSN2 sensor scans (green). The headers for both packet types are on top. Each SDPT_MRA_TS packet contains four successive samples. Acceleration data is sampled continuously at the specified rate (in this case 12.5 samples/second) and temporarily stored in the sensor’s 32-sample buffer. The SeaTag-MOD processor offloads this data periodically and transfers it to the storage card. Depending on the amount of acceleration data accumulated in the sensor’s buffer since the last offload, one to eight SDPT_HRA_TS packets may appear in a sequence. Although acceleration data acquisition is continuous and samples are evenly spaced in time, all packets in an offload sequence will be given the same time stamp, which has a 1 sec resolution.

Use of compressed HRA data

Compressed HRA data is recommended for animal activity studies. Compression focuses on total acceleration (G-Force) rather than individual axis data, so that the effect of Earth gravity (i.e. static acceleration) can be subtracted out. The data packet type SDPT_HRA_GS provides 24 data values, each representing the standard deviation of G over a specified interval. Each of the 24 data values is based on one (x;y;z) set of original accelerometer readings, which are generally spaced to capture the shortest movement of interest by the tagged animal. For example, you might select an acceleration measurement rate of 0.64 seconds and 188 measurements per data value in the SDPT_HRA_GS packet, resulting in an activity resolution of 0.64sec*188 = 120sec = 2 minutes. 24 of these values form one packet, and this packet covers 2 * 24 = 48 minutes. The SDPT_HRA_GS packet uses auto-scaling of the signal magnitude to provide the best possible resolution for each packet interval without causing data clipping. Maximum sensitivity is 0.001G with a full-scale value of 0.255G. The other extreme is limited by the acceleration sensor, which registers up to 16G. In a packet containing a data value of 16G, the resolution will be 0.0625G (16G / 256).

Figure 4: G-Force statistics for a car drive. 188 acceleration data readings sampled at 0.64 second intervals are combined to yield a time resolution of two minutes in this graph constructed from ten SDPT_HRA_GS packets.

Figure three demonstrates the data compression and available data. This data was obtained by securing a SeaTag-MOD to the frame of a truck, so it could register the G-forces experienced during a drive. The selected acceleration measurement interval was 0.64 seconds (fast enough to register the impact of most truck steering and potholes), and a compression factor of 188 was used to provide a time resolution of two minutes in the compressed data. Spanning an 8- hour interval, the data volume is as follows:

  • 11250 three-axis acceleration data packets of type SDPT_HRA_TS, each containing four acceleration readings.

  • 10 G-Force data packets of type SDPT_HRA_GS, each containing 24 data values formed from 188 original measurements and representing a 2-minute interval.

Given that 3-150 data packets might be received via ARGOS per day of transmission (depending on energy availability such as solar or battery, tag location etc.), the data load of ten packets per eight hours (30 packet/day) is reasonable.

Acceleration Measurements without HRA Mode

SeaTag-MOD’s HRA mode offers the advantage of high-frequency acceleration measurements with low power consumption. But, while HRA mode power consumption is low, it’s not negligible (see next section). And sometimes, the mode is just not needed and may be switched off. This is because SeaTag-MOD also states a three-axis acceleration measurement in each combined sensor packet of type SDPT_MODSN2. Here are two examples of such use:

Momentary Acceleration Snapshots

By computing the total acceleration sqrt(x^2+y^2+z^2), you can obtain the momentary acceleration of the animal at the time of measurement. 1G means the animal was at rest or at constant speed. Any variation of this number means that the animal experienced dynamic acceleration forces at the moment that the measurement was taken. Even a single measurement every few minutes can potentially be useful to discern periods of resting, hunting, migratory swimming and the like.

Figure 5: Total acceleration from combined sensor packets of type SDPT_MODSN2, acquired without HRA mode in 4-minute intervals. The tag was alternately attached to a truck frame and carried by a hiker for a four-day 4WD trip to Death Valley National park. Periods of rest (1G), highway driving, dirt track driving and hiking are distinguished by characteristic acceleration signatures.

Animal Body Orientation

The static acceleration component of each (x;y;z) measurement signifies the orientation of the tag relative to earth gravity. If the tag is mounted on an animal in at least a semi-rigid manner, these measurements provide the body orientation at the time of the measurement. Some considerations apply:

  • If the tag was subject to dynamic acceleration at the time of the measurement, then an orientation measurement will be unreliable. Samples subject to dynamic acceleration may be removed by computing the magnitude of acceleration sqrt(x^2+y^2+z^2). This number will be close to 1G, i.e. earth gravity only, if the measurement is not significantly impacted by dynamic acceleration. Use thresholding close to 1G to remove samples that are unreliable for body orientation measurement purposes.

  • In many cases, it may not be practical to mount the tag in a precisely pre-determined orientation on an animal, or the tags orientation may shift over time. To compensate for unknown tag orientation or tag shifting, observe the tags average orientation for a reasonable period such as 24h. Assume that this orientation corresponds to the normal animal body position. Obtain momentary animal body position by subtracting the 24h average tag orientation. Edit

HRA Sample Frequency Analysis

The HRA sample frequency is maintained by the timing circuit of the accelerometer sensor. The plot provides the frequency distribution for a dataset of about 1,700,000 measurements with a set acquisition frequency of 12.5 Hertz. Actual frequency was determined by counting the measurements within each 1-minute interval. The average data acquisition frequency for the entire dataset was 12.65Hz.

Figure 6: Actual sampling frequency distribution for a dataset of about 1,700,000 measurements at a set sampling rate of 12.5 Hertz. (Original dataset courtesy Franziska Broell and Chris Taggart, Dalhousie University)

Power Consumption Data

Power consumption in HRA mode without/with storage of the raw data. Use the information to select your tag configuration, including need for and size of a battery pack.

SENSOR SAMPLING INTERVAL 
(mm:ss)
HR ACCELERATION MEASUREMENT INTERVAL (sec) AVERAGED CONSUMPTION (uA) DARKNESS ENDURANCE WITHOUT BATTERY (hours) DARKNESS ENDURANCE WITH ST-MRPS SMALL BATTERY PACK (days) DARKNESS ENDURANCE WITH ST-MRPL LARGE BATTERY PACK (days) DARKNESS ENDURANCE W/ ST-MRPC SOLAR RECHARGEABLE BUFFER BATTERY (days)
00:01 0.08 2370/4641 0.4/0.2 17.6/9.0 36.9/18.9 0.4/0.2
00:02 0.16 1471/2663 0.6/0.3 28.3/15.6 59.5/32.9 0.6/0.3
00:06 0.32 1144/1059 0.7/0.8 36.4/39.3 76.5/82.6 0.8/0.9
00:16 0.64 313/574 2.7/1.5 132.9/72.6 279.1/152.4 2.9/1.6
00:32 1.28 214/330 3.9/2.5 194.4/126.2 408.3/265.0 4.3/2.8
01:04 2.56 131/189 9.6/6.6 318.6/220 669.0/462.0 7.0/7.4
02:08 5 116/124 10.8/10.1 360.0/336.1 756.0/705.8 7.9/7.4
04:00 10 73/90 17.1/13.9 571.4/463.5 1200.0/973.4 12.6/10.2

 Notes:

  1. Averaged consumption and battery endurance are stated without/with storage of the 3-axis HRA data on the memory card. HRA based statistics are always stored on the memory card.

  2. Data based on tag capacitor discharge tests with V2.20 firmware.

  3. Endurance numbers are time (stated in hours or days) to maintain HRA data acquisition operations. Thereafter, some battery or capacitor capacity will be available to maintain the clock and trigger the release.

HRA Power Consumption Impact on ARGOS Archival Transmissions

If storage of the raw (3-axis) HRA data on the memory card is selected, the effective power needed per archival packet transmission by ARGOS will be elevated because the tag must search through large amounts of raw HRA data before finding the next G-Force statistics packet for transmission. The HRA Measurements/Point parameter determines this power overhead, because higher values result in more raw HRA packets to search through before the next pre-computed G-Force statistics packet is encountered in memory. Consider this impact on the ARGOS transmission power budget when deciding if to store the raw HRA data.

The estimated power ‘overhead’, based on the assumption of a 300mA transmit current, is approximately:

  • 3% with 10 measurements/point

  • 14% with 50 measurements/point

  • 28% with 100 measurements/point

  • 56% with 200 measurements/point

  • 83% with 300 measurements/point

  • 138% with 500 measurements/point (max. value)


Notes:

  • At 300mA transmit current

  • Overhead stated as a percentage of standard transmit current

  • Based on 2700uA-sec consumption for a microSD card 512-byte page read & scan

  • Based on 24 data points per G-Force statistics packet, and 4 measurements per raw HRA packet

Configuring SeaTag-MOD for HRA

Starting Requirements
  1. You must use or install firmware V2.20 or later on your SeaTag-MOD

  2. You must use SeaDock V2.2.0.0 or later to configure the tag

The following screen dumps provide an example for a PSAT HRA mission, with opportunistic transmissions of positiononly messages during brief surface intervals of the tagged animal. Use variations of this template, as appropriate for your requirements.

Start Tab

  • Set the general sensor sampling interval to get the desired HRA rate. Due to the accelerometers data buffering operation, the HRA rate is about 24x faster than the sensor sampling interval. For the exact speed, see the Measurement Rate calculator in the High-Res Acceleration box on the Mission Settings tab. Be aware that faster speed means higher power consumption. Check the previous section of this app-note for details.

  • Select the Battery Type you are using. That info is needed by the tag to properly regulate the ARGOS transmit power to the specified level.

  • Some packets still stored on your tag? Clear them using Tag→Erase Logs so that old data will not be transmitted by ARGOS.

ARGOS ID’s Tab

  • Enter all the Decimal ID (Dec_ID) and Hexadecimal ID (Hex_ID) assigned to your program. You can also put in amemo under Notes. All other fields will be auto-populated upon tag configuration, so don’t enter anything in them.

  • Now check all ID again. Any digit wrong or swapped? The Hex_ID are used by the tag to talk to the satellite. A wrong ID means you will not see any data!

  • Now use the Lock Table button, to provide a little security against unintentional modification of the table. You can unlock at any time.

Mission Settings Tab

  • The Mission Type is PSAT with RT Tracking in which the tag collects data and then pops-off for transmission. During the collection phase, brief messages are sent when the opportunity arises.

  • Select the Constant Depth Release function to instruct the tag to pop its release and start archived data transmission when it has seen no significant depth change for the specified number of CDR Days. This may happen because the tag separated and is already floating at the surface, or is sitting on the sea floor. The depth variation to trigger CDR is tag version specific. Select Communications- >Show Raw Communications and click the Calibration Report button. Look for the ‘Constant Depth Release’ statement. Dismiss the comms window after use to avoid slow-downs.

  • Select HRA Active, to enable high-resolution acceleration measurements.

  • Store HRA selects storage of the raw 3-axis HRA data on the tag. Due to high volume, raw HRA data is only available via direct download from the tag, not via satellite transmission. Be aware of the increased power consumption associated with raw HRA data storage (previous section).

  • Measurements/point specifies how many HRA measurements are combined into one data point in the G-Force statistics. The calculator below indicates the resulting time interval per G-Force statistics data point in minutes and seconds. In the example, 282 measurements at 0.64 sec intervals result in a three minute coverage for each data point in the statistics. The Measurement rate is based on the Sensor Sampling Interval on the Start tab.

  • The Mission Start Date defines when observations start, and the Mission End/Release Data when the tag separates from the animal to start transmitting its archived data.

  • Set the ARGOS ID Code, so that each of your tags has a unique ID code. Be sure to notice which tag you deployed on which animal.

  • Select the packet types for Archival Transmission as needed. Check the ARGOS Transmission Load in response to the packet selection and the HRA parameters. A set of six solar-powered ocean tests has shown 17- 76 packets received per day during the first ten days at liberty (i.e. prior to onset of bio fouling), with totals ranging from 361 to 3049 received before receptions became negligible or sporadic. Battery powered packet receptions should range 800-1600 total for a small pack, and 1600-3200 for a large pack. Solar and battery powered receptions should add, for example 1200 from a small battery plus another 1500 solar powered would give you 2700 packet receptions.

  • Priority Transmit Loops and Days specify how often / for how long condensed data types (Daily Summary, Histograms, G-Force Statistics) are transmitted before transmission of the general sensor data (SDPT_MODSN2 packets) starts. Generally, leave as is.

  • Select if you want to send short, depth & Argos Position Only or the same data as selected for archival transmissions as the Real Time Transmissions (opportunistic transmissions) while the tag is still on the animal. The position only packets consume only 1/3 the energy required to transmit a full packet, and can thus be transmitted 3x as much with a given energy budget. For animals where surface intervals will not be common, this is the preferred setting to maximize the chance for getting position fixes while the tag is on-animal.

  • The Transmit Requirements specify the conditions to be met for a transmission. Once a tag pops loose and floats on the surface, transmissions are very uncritical (generally good view of the sky). For opportunistic onanimal real-time transmissions, RF Surface Detection can provide a good way to detect short surface intervals. But, the method currently costs an additional 60uA current load. Select it only if there is a realistic chance of animal surfacing. Near-surface depth makes sure the tag is actually at least near the surface before attempting a transmission. Little downside here, except if the depth sensor should fail.

  • The Transmit Days specifies for how long the tag will transmit post-release. This is to put an upper limit on your ARGOS budget.

  • The Transmit Repeat Period should be set to the Service ARGOS assigned interval for your program.

  • The Transmit Current defines the ARGOS signal strength vs. power consumption. 300mA is currently our recommended standard.

Histogram Options Tab

The histogram tab currently applies to the depth histogram only, the only one now available. However, the same settings will apply to the acceleration histograms once available.

  • The Time Reference should generally be Local Time, for example to capture day and night activities. The exception is deep-ocean animals where light is not available and local noon cannot be determined by the tag.

  • A Start Time of 06:00:00 and a histogram Duration of 12:00:00 will give you a day and a night histogram.

  • No scaling information is specified, as all histograms from now on will be auto-scaled for best fit.

Configuration and Performance Verification

  • Go to File→Tag Configuration Logs Folder… and inspect the final (bottom of file) configuration settings for all tags you will deploy. A line-by-line inspection will help verify everything is set correctly.

  • Once the tags have been readied for deployment (battery pack attached, antifouling coat applied etc.), put them in the sun under a clear sky for a day. Check to see that you are receiving a reasonable number of ARGOS transmissions from all tags.

  • Finally, we recommend a three-step ramp-up to maximize the success of your tag deployments:

    • Conduct some test ‘mini-missions’ including a release in a controlled environment where you can physically recover the tag. Tests in pools, ponds, water tanks, even on captive animals are useful.

    • If logistically possible, do a small-scale / abbreviated duration test with a few animals first. If that is not practical, place more emphasis on the initial controlled-environment tests.

    • Your full-scale tagging operation will ideally amount to a ‘more of same’ of previous testing. You have been through the loop before and already know what risks and data you can expect.


Ultimately, every field science program will have different opportunities and risk / reward trade-offs. Al Dove of the Georgia Aquarium literally took his first batch of tags on a plane the day after they arrived. He then deployed eight out of ten tags on whale sharks off the Yucatan within a week. It worked, and he got ‘LOTS of data’ in his own words. One of those is the twittering Domino: link.

Methodical advance or Hail Mary method, best of luck with your tagging program!