Accuracy via Magnetic Light Sensing
Using Magnetic and Light sensing to Improve Geo-Positioning Accuracy
With the the unavailability of GPS underwater, marine animal tags often use light sensing to estimate the position of tagged animals. But, due the difficulty of determining the true length of day (sunrise to sunset) underwater, such data is subject to large latitude errors. To address this problem, SeaTag devices integrate a three-axis magnetometer that senses the intensity of the earth's magnetic field. With the intensity exhibiting a distinct north-south gradient in much of the world, latitude estimates can be significantly improved. The attached paper discusses the methodology and field test results.
Light + Magnetic = Improved Position Estimates
Understanding the SeaTag Position Estimation Method
Natural-signal geo-positioning tags have traditionally relied just on light signals to obtain longitude and latitude. But, light based latitude estimates can be quite uncertain. Length of a day may vary only slightly with latitude, and the exact moment of the sun crossing the horizon is difficult to determine for an underwater tag with much confidence. For this reason, SeaTag devices estimate position based on both light and magnetic field intensity readings. The fundamental method, first proposed by Dr. Pete Klimley in the 1990’s, is straightforward. First, longitude is determined by light measurements. Next, the software runs along that line of longitude until a match with the measured total magnetic field intensity is found. This cross-hair of the light based longitude and magnetic based latitude defines the position of the tag. Using 16 data sets, this paper takes a brief look at the accuracy & reliability of the method. The data was taken at nine sites including Galapagos, Ecuador, Finland, Germany and five sites in the U.S. It spans 82 observation days and includes static land observations and buoy tests from 1.5m to 88m depth.
Longitude Estimation Method & Accuracy
SeaTag determines the noon time by finding the half-way point between first & last light, not the moments of sunrise or sunset. This method has the advantage of being independent of light attenuating factors such as tag fouling, cloud cover, water clarity and depth as long as these factors are reasonably similar between sunrise and sunset. It also makes use of the steep rise and fall of light intensity before sunrise and after sunset which lessens the impact of any asymmetries. The observed median longitude estimation error was 20 nautical miles, with results generally robust. The two outliers (15 and 16) point to a limit of the method. These were the two deep test sites in Lake Tahoe, at depths of 73m and 88m. The observed maximum illumination was about 0.1% and 0.02% of the surface value, forcing the detection point to higher surface light levels (later at dawn / earlier at dusk) where the light gradient is less reliable. The next buoy up was at 44m, with 0.9% of surface light and a longitude error of 19 nautical miles. From the Lake Tahoe buoy string (six tags from 3m to 88m depth), we can estimate the impact of dawn/dusk condition asymmetries. The day length was measured 39 minutes less at 44m with 0.9% of surface light as compared to 3m, where light levels may have been 50% of surface. A tag that is experiencing this approx. 50:1 light attenuation ratio at dawn vs. dusk should see a longitude error of about 113 miles.
Latitude Estimation Method & Accuracy
SeaTag takes magnetic readings every four minutes, with the daily average used to estimate the latitude. The median latitude error of the 16 data sets was 44 nautical miles. Once more, the outliers point to the limits of the method. #11 was an observation in a small backyard in Germany, with metal lawn furniture etc. in close proximity. Manmade magnetic anomalies probably account for the large error. Natural anomalies tend to max. out around 400nT, which is roughly 40-80 nautical miles equivalent (5-10 nT/NM gradient). #1 and #4 were two tags in the Sierra Nevada in winter – in cold below the bounds of the temperature compensation curve used by the software. #15 and #16 actually had fairly accurate magnetic readings that should place them with 24-29 miles of the true latitude. The culprit here is the northwest-southeast run of the magnetic field intensity lines in California. These two datasets had poor longitude estimates, which caused the large latitude error. Latitude accuracy depends both on the magnetic field gradient in an area, and the direction of the magnetic field lines.
Position Estimate Confidence
On a smaller scale, the method is subject to errors of a magnitude as shown. On a larger scale, the speed of the day/night transition puts a bound on longitude errors, while errors in magnetic observations bound latitude errors. The maximum magnetic measurement error across the 16 above sets plus another 30 magnetic-only sets was 1700 nT. With gradients around 3-10 nT/n.m. in much of the oceans, upper error bounds are on the order of hundreds of miles.