The centralized nature of BioBase (biobasemaps.com) cloud technologies coupled with sophisticated, yet low-cost consumer electronics like Lowrance or Simrad depth sounders/chartplotters have created fertile grounds for developing, testing, and verifying algorithms for typing aquatic environments. The more users upload from a greater range of systems, the more refined algorithms can become addressing a wider range of conditions and use cases!
Early in 2014, we released a revision to our EcoSound bottom composition (hardness) algorithm that is more sensitive and robust in a greater range of depths and bottom conditions. Many outside researchers were involved with collecting important “ground truth” information while they logged their BioBase data. This blog not only serves to describe the new Bottom Composition algorithm, but also publish the results and acknowledge the scientists that helped with this effort.
How it works and the outputs produced
The Composition algorithm processes the 200 kHz Broadband downlooking signal and produces a data point for GPS signals (Typically 1 pt every 1-2 seconds). Algorithms estimate the acoustic reflectivity of the bottom (“E2” in acoustic techno-speak). Signals bounce more on a hard bottom than a soft bottom where signal is absorbed. Hardness ratings are consistent across all mapped systems and not relative to a trip (e.g., a trip with muck and silt will show all light tan colors). GPS point data along tracks are sent to an interpolation (kriging) algorithm to predict hardness between sampled areas and create a uniform map
|Figure 1. Continuous, unitless data are created with each GPS coordinate to reflect relative hardness from soft 0-0.25 (light tan), to medium 0.25 – 0.4 to hard 0.4 – 0.5 (dark red).|
How does it compare with actual data? Verification results from independent researchers
Unlike conventional models or software programs that use limited datasets in a narrow range of conditions to calibrate and verify model outputs, BioBase is able to draw from our central database and network of professionals using the system to develop new or improved algorithms.
For revisions to the composition algorithm, Navico technical staff worked with scientists from the University of Florida (Mark Hoyer), USGS in Little Rock AR and Reston VA and (Drs. Reed Green, Nancy Rybicki and Elizabeth Striano), and across the pond with the Center for Ecology and Hydrology (Drs. Ian Winfield, Helen Miller, and Joey van Rijn) evaluating the agreement of their independently collected bottom composition data with companion BioBase hardness datasets. Despite field error in the precise estimation of actual hardness and overlap with simultaneously collected BioBase EcoSound data, we were encouraged by the high agreement of compared data sets. See for yourself below!
|Figure 3. Sediment depth measured by a 3/4 inch all thread pipe pushed to the “point of refusal” as it relates to simultaneously collected Lowrance acoustic data which was processed for bottom hardness with BioBase. Data were collected in Millwood Lake, Ashdown Co., Arkansas USA by Dr. Reed Green USGS Arkansas Water Science Center. Publication of sediment depth patterns on Millwood can be found at: Richards, J.M., and Green, W.R., 2013, Bathymetric map, area/capacity table, and sediment volume estimate for Millwood Lake near Ashdown, Arkansas, 2013: U.S. Geological Survey Scientific Investigations Map 3282, 1 sheet,http://dx.doi.org/10.3133/sim3282.|
Table 1. Agreement between visually estimated substrate hardness while collecting Lowrance/EcoSound composition data from 9 of 23 samples in coastal Back Bay, Virginia Beach VA, USA in 2012. BioBase composition data at the remaining 13 sites were not generated due to depth or vegetation thresholds. Bottoms cannot be typed where vegetation fills > 60% of the water column or in depths less than 2.4 feet from the transducer face. Data were collected by Dr Nancy Rybicki and Elizabeth Striano, USGS – Reston VA as a component of a vegetation assessment study.
|Figure 4A. Hardness data from Windermere (Cumbria, England) as scored by visual estimation from underwater imagery as it relates to hardness from Lowrance and BioBase EcoSound. Data were collected in 2012 by Dr. Ian Winfield and Joey van Rijn and described in a previous blog post. A peer reviewed paper with these data have been published in Ecological Informatics. The biological context and other companion composition data are presented in Miller et al. 2014.|
|Figure 4B. Bottom substrate (and Northern Pike) as viewed from a camera mounted on a Remote Operated Vehicle (ROV) in Windermere. See Figure 5A for the Hardness Score and Corresponding BioBase Hardness data|
Create your own sediment thickness models
The BioBase EcoSound composition algorithm will not predict sediment depth, only whether the bottom is hard or soft based on the E2 “echo” of the acoustic signal. Still, what we show in Figures 2 and 3 are that sediment depth may correspond predictably with bottom hardness as estimated by EcoSound. It is important to note however that there are situations where hard bottoms can appear acoustically soft and where soft bottoms can appear acoustically soft.
Hard bottoms appearing acoustically soft are more well documented than the opposite situation. In seafloor mapping, very rocky bottoms or wavy sand bottoms can scatter so much energy that a second echo is weak (Penrose et al. 2005 and this link). We have also seen evidence in small productive ponds that a layer of detritus (tree leaves, decaying aquatic plants, sticks, tubers and other root structures) may create a moderately strong E2 signal even if the underlying bottom is very soft. This situation needs further research. Please let us know if you are a researcher and interesting in pursuing the topic.
The primary benefit of BioBase EcoSound is to provide a full-system understanding of where hard and soft areas exist (Figure 5). Investigators can follow up with a couple of dozen coring points in areas of interest (e.g., sedimentation deltas) to develop system-specific relationships like those in Figures 2 and 3. A perfect example of this is demonstrated by Schooley and Neeley in their 2017 publication in the Journal of Applied Icthyology (Figure 5)