M.L. 2013 Minnesota Aquatic Invasive Species Research Center Subproject Abstract For the Period Ending June 30, 2019 SUBPROJECT TITLE: MAISRC Subproject 21: Early detection of zebra mussels using multibeam sonar SUBPROJECT MANAGER: Jessica Kozarek AFFILIATION: St. Anthony Falls Laboratory, University of Minnesota MAILING ADDRESS: 2 SE 3rd Ave CITY/STATE/ZIP: Minneapolis, MN 55414 PHONE: 612-624-4679 E-MAIL: jkozarek@umn.edu WEBSITE: www.safl.umn.edu www.maisrc.umn.edu FUNDING SOURCE: Environment and Natural Resources Trust Fund (ENRTF) LEGAL CITATION: M.L. 2013, Chp. 52, Sec. 2, Subd. 06a SUBPROJECT BUDGET AMOUNT: $96,549 AMOUNT SPENT: $96,175 AMOUNT REMAINING: $374 Sound bite of Subproject Outcomes and Results This project tested the utility of a swath mapping system (multibeam sonar) to detect the presence/abundance of zebra mussels. Acoustic backscatter data was collected and machine-learning was used to identify what is present in the substrate. Researchers were able to differentiate by mussel type (native vs. invasive) and density. Overall Subproject Outcome and Results Zebra mussels pose a serious threat to Minnesota lake and river ecosystems. However, monitoring zebra mussel populations is challenging because current methods for detecting and counting zebra mussel colonies rely on time consuming and expensive diving surveys, video imaging, or sampling of veligers (larvae), which limits the areas surveyed. Remote sensing techniques have been shown to quickly and efficiently gather spatially extensive information. Using this technology to detect zebra mussels would likely be much more efficient and more effective than traditional methods and could be used for early detection and warning in rivers, lakes and reservoirs and to track changes in zebra mussel density. This project was the first phase of research designed to test the utility of a swath mapping system, multibeam sonar, for detecting the presence and abundance of invasive mussels. Laboratory experiments were conducted to test the feasibility of using multibeam sonar to distinguish zebra mussel containing substrates. Acoustic backscatter data were collected in a two meter deep tank over sand, gravel, and mixed substrate containing high and low densities of zebra mussels and with native mussels using combinations of different sonar settings (frequencies and pulse lengths). Machine-learning was used to differentiate the acoustic backscattering signatures in a data-driven substrate classifier approach. Using these methods, we were able to classify substrate by size and mussel density. Classification errors decreased with more sonar settings. For minimum errors of less than 20%, 8 sonar settings are required, and for minimum errors of 10% or less for all substrates, 12 sonar settings. Each sonar setting corresponds to a separate boat survey of an area with a multibeam sonar in the field. Therefore, the next phase of this research is to further develop and test multibeam sonar monitoring approaches in the field (MAISRC Subproject 21.2: Field validation of mulitbeam sonar zebra mussel detection).

Subproject Results Use and Dissemination Research results from Phase I will be disseminated through a peer-reviewed publication (in preparation) and will inform Phase II field testing starting July 2019 (MAISRC Subproject 21.2: Field validation of mulitbeam sonar zebra mussel detection). During this one-year project, we participated in MAISRC Fellows meetings and presented our project to the public at the annual MAISRC Research & Management Showcase. The Minnesota

- Page 2 of 2 -

Environment and Natural Resources Trust Fund (ENRTF) will be acknowledged through use of the trust fund logo or attribution language on project print and electronic media, publications, signage, and other communications per the ENRTF Acknowledgement Guidelines.

Environment and Natural Resources Trust Fund (ENRTF) M.L. 2013 Minnesota Aquatic Invasive Species Research Center Subproject Work Plan Final Report

Date of Report: August 9, 2019

Final Report

Date of Work Plan Approval: July 31, 2018

Subproject Completion Date: June 30, 2019

Project Completion Date: June 30, 2019

SUBPROJECT TITLE: MAISRC Subproject 21: Early detection of zebra mussels using multibeam sonar Subproject Manager: Jessica Kozarek

Affiliation: St. Anthony Falls Laboratory, University of Minnesota

Mailing Address: 2 Southeast 3rd Ave

City/State/Zip Code: Minneapolis, MN 55414

Telephone Number: (612) 624-4679

Email Address: jkozarek@umn.edu

Web Address: http://www.safl.umn.edu/ https://www.maisrc.umn.edu/ Location: St. Anthony Falls Laboratory, University of Minnesota, Statewide

Total ENRTF Subproject Budget: Subproject Budget: $96,549

Amount Spent: $96,175

Balance: $374

Legal Citation: M.L. 2013, Chp. 52, Sec. 2, Subd. 06a Appropriation Language: $4,350,000 the first year and $4,350,000 the second year are from the trust fund to the Board of Regents of the University of Minnesota to develop and support an aquatic invasive species (AIS) research center at the University of Minnesota that will develop new techniques to control aquatic invasive species including Asian carp, zebra mussels, and plant species. This appropriation is available until June 30, 2019, by which time the project must be completed and final products delivered.

2

I. SUBPROJECT TITLE: MAISRC Subproject 21: Early detection of zebra mussels using multibeam sonar II. SUBPROJECT STATEMENT: Zebra mussels (Dreissena polymorpha) pose a serious threat to water supply and power plant infrastructure, and to Minnesota lake and river ecosystems, including native mussel species (Baker and Hornbach 1997). Current methods for detection and quantification of zebra mussel colonies rely on time consuming and expensive diving surveys, video imaging, or sampling of veligers (larvae) in the water column. Survey sampling design would be made much more efficient given spatially extensive information on the on the presence/absence of zebra mussel beds. Such remote sensing technology would also be useful for early detection and warning in rivers, lakes and reservoirs through routine monitoring, or to follow changes in zebra mussel density (boom or bust cycles). This study will test the utility of swath mapping systems such as multibeam sonar for detecting and quantifying the abundance of invasive mussels at a very large scale. Multibeam sonar can map tens to hundreds of square kilometers of river or lake bed in a single day from a moving vessel. Ostensibly an instrument for bathymetric mapping, each sounding from a multibeam sonar also records the echo from the bed surface, which can be analyzed to provide information about the roughness and composition of the ensonified bed. This echo can be used to reliably distinguish among various substrates (Brown et al., 2011). Acoustics are also increasingly being used to map and monitor shellfish (e.g. Sanchez-Carnero et al. 2014) and submerged vegetation (e.g. Buscombe et al., 2017). There is a strong likelihood that mussels have a distinct acoustic response (echo) compared to their surrounding substrate. If so, this acoustic signature can be readily used to detect and map zebra mussel beds at cm to m resolution in any navigable waterway of sufficient water depth. This study will define the methodology needed to detect, distinguish and quantify mussels from a moving vessel by studying backscattering of sound by mussels and common mussel-supporting substrates. Mussels are soft-bodied invertebrates with hard shells. The acoustics of backscattering by mussels might depend on many physiological and morphological factors such as size, shape, shell thickness/roughness/composition, and the composition of soft tissues. In concert, these factors manifest as differences in scattering due to differences in roughness and hardness. It should therefore be possible to discriminate between different species of mussel (zebra mussels vs. native species) using acoustics alone, or acoustics in combination with measurable environmental variables that govern the spatial distributions of mussels. In lakes and rivers, this methodology will enable the scanning of large areas for the early detection of zebra mussel colonies. In river systems, it could be applied to detect longitudinal changes in zebra mussel populations downstream from a source population to evaluate the role of downstream drift in zebra mussel spread. The first phase of this study, laboratory experiments, is designed as a proof-of-concept to utilize multibeam sonar to distinguish amongst substrate, native and zebra mussels in a controlled setting. We will study the acoustic backscattering properties of zebra mussels (Dreissena polymorpha) and native mussels, Threeridge (Amblema plicata), under controlled laboratory settings. Experiments in self-contained tanks at the St. Anthony Falls Laboratory will be used to determine the acoustic parameters that will maximize the discrimination between mussels and substrates. Following this study, a second research phase is planned to validate and further develop methodology in the field. Field measurements will allow the incorporation of a larger range of variables (mussel density, mixed substrates, water depth, etc.), once methodology has been tested in carefully controlled laboratory conditions.

III. SUBPROJECT STATUS UPDATES: Amendment Request October 9, 2018: An amendment is requested to adjust the subproject budget in order to account for unanticipated delays in acquiring required equipment (hydrophone) and to accommodate delays in setting up subcontracts. This

3

amendment request will have no impact on activity-level or overall budget totals. Additionally, we do not anticipate the changes in contracts and personnel impacting outcomes of the project. A high frequency hydrophone (TC4038-1 Hydrophone 50 kHz to 800 kHz (>1MHz) 2 M Cable and VP2000 voltage amplifier with filters) was included in the original project budget. The original supplier, Teledyne, was unable to deliver a hydrophone during the timeframe anticipated for the calibration of equipment. While searching for alternatives, an additional supplier, Benthowave, was identified and could provide the necessary hydrophone at a decreased cost ($2,550). Amendment Request 1 Equipment/Tools/Supplies – Move $1,006 from Capital Expenditures to Equipment/Tools/Supplies and adjust line item totals to allow for the purchase of a hydrophone at the total cost of $2,550. This $1,006 increase and adjustment of supply costs will bring the Equipment/Tools/Supplies budget to a total of $9,035. Amendment Request 2 Professional/Technical Services and Contracts – Decrease contract amount with Northern Arizona University (reduced travel costs) and move $4,994 from Capital Expenditures and $1,632 from Personnel to increase USGS contract amount to accommodate additional calibration data collection using the hydrophone. Overall budget impact will increase Professional/Technical Services and Contracts by $6,626 to a total cost of $48,771. Amendment Approved: [10/12/2018] Subproject Status as of January 31, 2019: We successfully completed the planned lab experiments over 4 weeks in September 2018. Using the data, we have developed machine-learning-based substrate classifiers hypothetical situations of abiotic (bare) and biotic (mussel-supporting) substrates. The input into each model is measured backscattering strength of the bed over prescribed combinations of several acoustic frequencies and pulse lengths. The model output is the likelihood of each substrate class. Each model is trained only on distributions of uncalibrated acoustic backscatter measured in the lab over ten unique substrates, namely: 1) sand, 2) mix sand-gravel (MSG); 3) gravel; 4) sand-supported A. plicata; 5) MSG-supported A. plicata; 6) gravel-supported A. plicata; 7) sand-supported D. polymorpha (low density); 8) sand-supported D. polymorpha (high density); 9) gravel-supported D. polymorpha (low density); and 10) gravel-supported D. polymorpha (high density). Phase I, experiments to examine the feasibility of using multibeam sonar to detect zebra mussels, is considered complete when the following objectives have been met: * indicates objective has already been met

1. Conduct lab experiments (summer 2018)* 2. Develop an empirical substrate classifier based on measured uncalibrated backscatter (fall/winter

2018)* 3. Investigate the feasibility of developing analytical substrate classifier based on measured calibrated

backscatter (spring 2019) 4. Develop a prototype field protocol for zebra mussel detection (spring/early summer 2019) 5. Write and disseminate findings (spring/early summer 2019)

Amendment Request January 31, 2019: An amendment is requested to adjust the supplies budget to account for unanticipated supply costs including materials to attach the hydrophone to the data carriage for the calibration experiment planned Spring 2019. This amendment request will have no impact on activity-level or overall budget totals. Additionally, we do not anticipate the changes in contracts and personnel impacting outcomes of the project.

• Personnel – decrease from $38,743 to $37,743 • Equipment/Tools/Supplies – increase from $9,035 to $10,035

4

Amendment Approved by LCCMR: 02/26/2019 Overall Subproject Outcomes and Results: Zebra mussels pose a serious threat to Minnesota lake and river ecosystems. However, monitoring zebra mussel populations is challenging because current methods for detecting and counting zebra mussel colonies rely on time consuming and expensive diving surveys, video imaging, or sampling of veligers (larvae), which limits the areas surveyed. Remote sensing techniques have been shown to quickly and efficiently gather spatially extensive information. Using this technology to detect zebra mussels would likely be much more efficient and more effective than traditional methods and could be used for early detection and warning in rivers, lakes and reservoirs and to track changes in zebra mussel density. This project was the first phase of research designed to test the utility of a swath mapping system, multibeam sonar, for detecting the presence and abundance of invasive mussels. Laboratory experiments were conducted to test the feasibility of using multibeam sonar to distinguish zebra mussel containing substrates. Acoustic backscatter data were collected in a two meter deep tank over sand, gravel, and mixed substrate containing high and low densities of zebra mussels and with native mussels using combinations of different sonar settings (frequencies and pulse lengths). Machine-learning was used to differentiate the acoustic backscattering signatures in a data-driven substrate classifier approach. Using these methods, we were able to classify substrate by size and mussel density. Classification errors decreased with more sonar settings. For minimum errors of less than 20%, 8 sonar settings are required, and for minimum errors of 10% or less for all substrates, 12 sonar settings. Each sonar setting corresponds to a separate boat survey of an area with a multibeam sonar in the field. Therefore, the next phase of this research is to further develop and test multibeam sonar monitoring approaches in the field (MAISRC Subproject 21.2: Field validation of mulitbeam sonar zebra mussel detection). IV. SUBPROJECT ACTIVITIES AND OUTCOMES: ACTIVITY 1: Acoustic backscattering properties of mussels Description: We will study of the acoustic backscattering properties of the Zebra mussel (Dreissena polymorpha) and a common species of native mussel, Threeridge (Amblema Plicata), under controlled laboratory settings. These experiments will utilize the same equipment that will be used in field surveys and parameters will include the optimal frequency or combination of frequencies, and pulse length. These experiments will also determine whether it is possible to acoustically discriminate between native and non-native mussels, as well as among various substrate types and the two mussel species. Experiments will be conducted within self-contained tanks on the St. Anthony Falls Laboratory model floor. This facility is equipped with a programmable instrumentation carriage that will allow for exact positioning of acoustic equipment. Summary Budget Information for Activity 1: ENRTF Budget: $ 96,549 Activity Completion Date:

Outcome Completion Date 1. Acoustic parameters to detect zebra mussels June 2019 2. Acoustic parameters to detect native mussels June 2019 3. Effect of substrate on detection June 2019

Activity Status as of January 31, 2019: The research team completed experiments designed to investigate the feasibility of using acoustic backscatter to differentiate between mussels and the substrate that supports them using live mussels (Amblema plicata and Dreissena polymorpha) as described in the project status above. To eliminate any risk of unintentional zebra

5

mussel transport, all native mussel and bare sediment experiments (sand, gravel, and mixed; Figure 1; Table 1) were conducted before any zebra mussels were introduced to the experimental area. Fifty A. plicata were collected by Mark Hove under permit from the MN DNR from Sturgeon Lake, a backwater of the Mississippi River near Prairie Island, MN in late August 2018. Mussels were held in a flow-through flume fed by Mississippi River water until moved into the experimental setup. Once an experiment was complete, the mussels were moved back to the holding facility. Native mussels were placed in the same orientation, on a 20 cm grid, buried approximately 1/3 into the sediment with their siphons pointing up at the start of each experiment. Zebra mussels were collected from Prior Lake under permit from the MN DNR in early September, 2018 and were transported back to SAFL in sealed containers. Zebra mussels were housed in a separate, non-flow through tank with secondary containment to prevent any spills. After the commencement of the native mussel and bare sediment experiments, zebra mussels were placed on sediment (sand or gravel). Zebra mussels that were attached to substrate (gravel or cobble) were left on the substrate and the zebra mussel containing substrate was added to the tank (Table 1). Upon completion of each experiment, the tank water was drained into a containment tank and was treated before disposal. Upon completion of all experiments, zebra mussels were frozen before disposal.

Figure 1. Cumulative grain size distribution of sediment used in experiments: sand, gravel, and a 50% mixture by weight of sand and gravel. Table 1. Sediment characteristics for experiments. D50 is the median grain size and D90 is the 90th percentile grain size.

Sediment D50 (mm) D90 (mm) Sand 0.67 ± 0.0 0.98 ± 0.0

Gravel 12.4 ± 0.1 15.2 ± 0.03 Mixed Sand/Gravel 0.97 ± 0.02 14.3 ± 0.1

Zebra mussel supporting substrate with attached mussels 38.0 69.9 The experimental setup consisted of a lined plywood tank (2 m x 1 m x 2 m deep) with 10 cm of sediment filled with Mississippi River water. The interior of the tank was lined with artificial grass to absorb sound and minimize reverberations (Figure 2). The bed was ensonified using a Norbit iWMBSh multibeam sonar, which was controlled using custom software, written in collaboration between project partner Buscombe and the multibeam sonar manufacturer Norbit Subsea, in order to control pulse timing and minimize reverberations in the tank further using pauses, and to automatically switch sonar settings and reload the sonar beamformer. The instrument cycled through 16 settings (four different pulse lengths and four frequencies) over a 120-degree

0102030405060708090

100

0.01 0.10 1.00 10.00 100.00

% fi

ner

Sieve size (mm)

gravel50-50sand

6

swath. The instrument was moved very slowly over the width of the bed using a computer controlled data carriage on SAFL’s model floor (Figure 3). This was done to ensure that all organisms present in the tank were ensonified, and from multiple angles, and to negate any acoustic effects of mussels moving during the experiment. Each raw experimental data set therefore consists of multibeam backscatter aggregated over the entire bed surface, and organized per incident sonar angle (-60 through 60 degrees, in one degree increments), incident frequency (100, 200, 400, or 700 kHz), and incident pulse length (100, 200, 500, and 800 µs).

Figure 2. Test tank built on the SAFL model floor to record the acoustic response of mussel containing substrates. The tank was lined with artificial grass to minimize sound reflections. Native mussels (A. Plicata) were placed at 20 cm spacing buried approximately 1/3 into the sediment. Mussels were allowed to move freely throughout the experiment.

Figure 3. Data carriage used to position the multibeam sonar (circled) in experimental tank.

7

Multibeam backscatter data processing In order to access the backscatter information inherent to the substrate, the raw echo features recorded by the sonar but unrelated to the substrate need to be corrected in the recorded signal. Two types of compensations are applied to each raw echo so that: (1) the characteristics of the sonar sensor (relative sound source level, reception sensitivity, acoustic spatial footprint) do not affect the estimation of the substrate’s backscatter strength; and (2) the propagation losses must be corrected according to local environmental conditions (in the laboratory, just water temperature is relevant) and acquisition geometry (the relationship between backscatter strength and incident angle). Subsequently, the corrected echo intensity can be considered as representing the substrate effect alone, and translated into the target backscatter strength (i.e. its inherent capability for sending back acoustic energy to the sonar). This backscatter relates to the mechanical characteristics of the sediment and any organisms present, the geometrical shape and size of both sediment grains and individual organisms, but also on the signal frequency and duration. Machine-learning substrate classifier Using the acoustical data, we have first explored the possibility of differentiating the acoustic backscattering signatures of mussels from the substrates that support them in a data-driven approach using only the composite surficial-volume scattering signal recorded by the MBES, i.e. without attempting to decompose the complex signal into its surficial and sub-surficial components. We have developed a discriminant function from a family of machine learning algorithms called Gradient Boosting Machines. This approach enables us to evaluate and rank the relative importance of input features (backscattering strength over various frequencies and pulse lengths) in order to prioritize optimal sonar settings for time efficiency for application of the method in the field. In the laboratory, we collected backscatter data using sixteen different sonar settings (all combinations of four frequencies and four pulse lengths). However, in the field, repeatedly surveying the same area with sixteen combinations of sonar settings would be prohibitively time-consuming. Based on field survey time considerations, we considered a hypothetical scenario of twelve sonar settings to be a practical maximum. We also constructed three smaller feature sets, so the four feature-sets used consisted of 12, 8, 4, and 2 backscatter features, respectively. The 12-feature set was constructed by eliminating the lowest and highest pulse lengths (respectively, 100 and 800 µs) from the full experimental dataset. An 8-feature set (Table 2) was constructed by further eliminating two frequencies (200 and 700 kHz). The 4-feature set was constructed by further eliminating the 100 µs pulse length, and finally a 2-feature set was constructed using two frequencies and 1 pulse length. For each feature set (Table 2), we developed two models for two hypothetical scenarios of biotic and abiotic substrate mixtures (Table 3), namely 1) no mussels (bare substrates only), and 2) a mixed native and invasive mussel population. The two models predict three and ten classes, respectively. Table 2. Four backscatter feature sets tested during this study.

Feature 12-feature set 8-feature set 4-feature set 2-feature set 100 kHz @ 100 µs X 200 kHz @ 100 µs X 400 kHz @ 100 µs X 700 kHz @ 100 µs X 100 kHz @ 200 µs X X X 200 kHz @ 200 µs X X X 400 kHz @ 200 µs X X X X 700 kHz @ 200 µs X X 100 kHz @ 500 µs X X X 200 kHz @ 500 µs X X 400 kHz @ 500 µs X X X 700 kHz @ 500 µs X x

8

Table 3. Two models for combinations of experimental substrate conditions.

Substrate/Model Model 1 (abiotic) Model 2 (mixed biotic) Sand X X

Mixed sand/gravel X X Gravel X X

A. plicata, Sand X A. plicata, Mixed sand/gravel X

A. plicata, Gravel X D. polymorpha (low density), Sand X D. polymorpha (high density), Sand X D. polymorpha (low density), Gravel X D. polymorpha (high density), Gravel X

F1 scores were computed to quantify how well each model performs in general, as an equal weighting of the recall and precision metrics. Precision and recall are standard accuracy metrics used to evaluate the classification performance of a machine learning model. Precision is the proportion of positive identifications was actually correct (a precision of 1 means there are no false positives), and recall is the proportion of actual positives identified correctly (a recall of 1 means there are no false negatives). Classification inaccuracies are generally very low for the 3-substrate model, across all four feature-sets (Table 4). Sand is classified with highest accuracy, with 1% error using 12, 8, and 4 features, rising to 18% using 2 backscatter features. Mixed sand/gravel classification errors are highest; up to 24% using the 2-feature set. Table 4. F1 scores for 3-substrate models trained on various feature sets.

12-feature set 8-feature set 4-feature set 2-feature set Sand 0.99 0.99 0.98 0.82

Mixed sand/gravel 0.98 0.96 0.92 0.76 Gravel 0.98 0.97 0.93 0.87

Classification inaccuracies are also low for each of the substrates in the 10-substrate model, across all four feature-sets (Table 5). Once again, sand is classified with highest accuracy, with 2- 3 % error using 12, 8, and 4 features, rising to 33 % using 2 backscatter features. Classification errors increase for every substrate class with progressive fewer backscatter features, however, the rate of increase is not the same for every substrate. Owing to the complexity of the backscattering of each substrate type, many backscatter features are required for high accuracy across all ten substrates; for minimum errors of less than 20 % across all ten substrates, 8 sonar settings are required, and for minimum errors of 10 % or less for all substrates, 12 sonar settings. Each sonar setting corresponds to a separate boat survey of an area with a multibeam sonar in the field.

9

Table 5. F1 scores for 10-substrate models trained on various feature sets.

12-feature set 8-feature set 4-feature set 2-feature set Sand 0.97 0.97 0.98 0.67

Mixed sand/gravel 0.95 0.87 0.92 0.38 Gravel 0.94 0.88 0.93 0.53

D. polymorpha (low density), Sand

0.97 0.88 0.68 0.36

D. polymorpha (low density), Gravel

0.95 0.9 0.75 0.46

D. polymorpha (high density), Sand

0.95 0.82 0.68 0.39

D. polymorpha (high density),

Gravel

0.94 0.84 0.7 0.43

A. plicata, Sand 0.98 0.91 0.81 0.47 A. plicata, Mixed

sand/gravel 0.96 0.87 0.71 0.38

A. plicata, Gravel 0.95 0.85 0.7 0.34 A manuscript is currently in preparation detailing the experimental work and the data-driven acoustic mussel-supporting classifier. The last experimental phase will be to conduct a short experiment to acoustically characterize the tank full of water and with a bed of bare substrate, using a hydrophone. Final Report Summary: The major goal of the project is to assess the feasibility of an acoustic remote sensing method for detecting Dreissenid and native mussels from their physical habitats, in high resolution for eventual and expanded use in monitoring and targeted research. This will involve the development of an optimal workflow for combining acoustic multibeam bathymetry and backscatter data and sparse ground truth data, to develop models for the likelihood of Dreissenid colonies. So far we have 1) collected the first known multibeam acoustic backscatter dataset for zebra and native mussels, 2) acoustically characterized the laboratory tank, and 3) developed a data-driven classifier that distinguishes between mussels and simple bare substrates. In April/May 2019, a high-frequency omnidirectional hydrophone was used to acoustically characterize the laboratory tank filled with water but with no mussels present in the substrate (Figure 3). The Norbit multibeam sonar was placed at the same height above the bed as during previous experiments. A gravel bed was used for these acoustic tests. The ping from the sonar was measured at 2 MHz using a data acquisition card mounted to the PCI slot of a PC motherboard with a 15 kΩ grounding resistor placed in-line to prevent DC offset in any measurements. These measurements have been used to determine the source level, attenuation, tone-burst duration and directivity pattern of the sonar. These measurements will better determine the angular reception sensitivity of the sonar, and better estimate the absolute values of the source level, therefore further diminishing the raw echo features recorded by the sonar that are unrelated to the substrate. These measurements have also been used to characterize any reverberations (damped echoes the result of facet reflections from the tank walls), which will be ultimately used in an uncertainty model for the backscatter data. Reverberations within the tank are minimal and therefore have no significant bearing on the major findings.

10

Figure 4. Images and results from hydrophone experiments conducted in April/May 2019. A) A view of the ensonified tank in the Norbit sonar data acquisition software; B) A spatial map of the peak acoustic power in the tank, relative to the sonar transmit array location. Reference intensity Iref is the measured peak acoustic pressure at 1 m below the sonar; C) Photograph of the unfilled experimental tank showing sonar and hydrophone on their respective mobile mounts; D) a close-up of the omnidirectional hydrophone sensing element.

The main experimental findings are listed below:

1. The ensonified area is narrow and spreads minimally, as expected (e.g. Figure 4B) 2. Tone-burst durations can be used to estimate acoustic footprints 3. Reverberations at incident and non-incident frequencies are minimal, quickly dampening within a few

microseconds 4. Measured source levels can be used to express all measurements collected in September 2018 on a

decibel scale

We are now confident that we have carried out sufficiently successful experimental work in the laboratory to now examine the portability of these methods to field settings. Phase II will involve an analysis of acoustic backscatter and bathymetric roughness toward a classification into a subset of the classes described within the Coastal and Marine Ecological Classification Standard (CMECS; https://www.cmecscatalog.org/cmecs/); and classification refinement using conditional random fields using multivariate predictors (Buscombe and Grams, 2018). V. DISSEMINATION: Description: The research team will participate in outreach activities through MAISRC including the annual Showcase event, media efforts and participation on committees. In addition, we will continue to communicate with Aquatic Invasive Species (AIS) managers about research progress and outcomes. Research will be disseminated through the publication of peer reviewed journal articles. We expect to publish 1-2 manuscripts detailing the methodology of mussel detection and discrimination between species and substrate. Status as of January 31, 2019:

11

The research project was presented to the public and other interested stakeholders at the annual MAISRC Research & Management Showcase in September. Research results are being compiled into 1-2 manuscripts for publication. Final Report Summary: Research results from Phase I will be disseminated through a peer-reviewed publication (in preparation) and will inform Phase II field testing starting July 2019 (MAISRC Subproject 21.2: Field validation of mulitbeam sonar zebra mussel detection). During this one-year project, we participated in MAISRC Fellows meetings and presented our project to the public at the annual MAISRC Research & Management Showcase. The Minnesota Environment and Natural Resources Trust Fund (ENRTF) will be acknowledged through use of the trust fund logo or attribution language on project print and electronic media, publications, signage, and other communications per the ENRTF Acknowledgement Guidelines. VI. SUBPROJECT BUDGET SUMMARY: A. ENRTF Budget: Explanation of Use of Classified Staff: N/A Explanation of Capital Expenditures Greater Than $5,000: N/A Number of Full-time Equivalent (FTE) funded with this ENRTF appropriation: 0.59 FTE Number of Full-time Equivalent (FTE) estimated to be funded through contracts with this ENRTF appropriation: 0.3 FTE B. Other Funds: N/A VII. SUBPROJECT STRATEGY:

A. Subproject Partners: This proposal leverages existing experimental facilities and expertise at SAFL, existing sonar, survey and video equipment and survey expertise at the U.S. Geological Survey, acoustics signal processing expertise at Northern Arizona University, and expertise in biology and ecology of native and invasive mussel populations at Macalester College. University of Minnesota, St. Anthony Falls Laboratory ($47,779) - Dr. Jessica Kozarek (Research Associate): Project Manager; Technical Staff; Undergraduate Student Macalester College, Environmental Studies ($14,787) - Dr. Daniel Hornbach (John S. Holl Professor and Chair); Mark Hove; Department of Geology - Dr. Kelly Macgregor (Professor) Northern Arizona University, Earth Sciences and Environmental Sustainability ($20,147) - Dr. Daniel Buscombe (Assistant Research Professor). Dr. Buscombe is an expert with more than 10 years of experience in hydrographic surveying and acoustics, including experience working with multibeam, singlebeam, sidescan, and sub-bottom profiling acoustic systems in freshwater. He has pioneered the study of acoustic substrate characterization in large regulated rivers in the US. He has published 6 journal papers on substrate characterization using sonar, with another in review. He has earned the runner-up prize in the 2017 R2Sonic Multispectral challenge, an international research competition on habitat mapping using new/emerging multispectral sonar acoustic technology. With this experience, Dr. Buscombe brings valuable expertise to the research team to be the first to study to use multispectral sonar in the lab.

12

U.S. Geological Survey, Southwest Biological Science Center, Grand Canyon Monitoring and Research Center ($13,837) - Dr. Paul Grams (Research Hydrologist). Dr. Gram’s research group specializes in physical and ecological studies in freshwater river systems. GCMRC has extensive experience in the use of multibeam sonar systems for mapping shallow river systems and is among the very few research facilities that are developing methods for acoustical bed substrate classification.

B. Subproject Impact and Long-term Strategy: This first phase of research, laboratory experiments, will provide a proof-of-concept for acoustic methods to detect and differentiate mussels from substrate and each other. This phase will be followed by a second phase to validate and further develop methodology in the field. Field measurements will allow the incorporation of a larger range of variables (mussel density, mixed substrates, water depth, etc.), once methodology has been tested in carefully controlled laboratory conditions. Typical methods including SCUBA surveys are limited by time and effort required to locate and quantify mussel beds. The methodology proposed in both phases of the current project will 1) provided an efficient method to detect zebra mussel beds over large spatial areas; and 2) enable greater understanding of the differences between invasive zebra mussel and native mussel habitat. This research will develop the necessary tools to further investigate and map zebra mussel spread and population changes over time, thus providing a basis for management of rivers to enhance native mussel habitat and limit zebra mussel spread.

C. Funding History: N/A VIII. REPORTING REQUIREMENTS:

• The project is for 1 year, will begin on 07/01/18, and end on 06/30/2019. • A periodic project status update report will be submitted January 31, 2019. • A final report and associated products will be submitted between June 30 and August 15, 2019.

IX. VISUAL COMPONENT or MAP(S): N/A

1

234567891011

12

13

14

15

16

17

18

19

20

2122

23

24

A B C D E F GEnvironment and Natural Resources Trust FundM.L. 2013 Subproject Budget

Subproject Title: MAISRC Subproject 21: Early detection of zebra mussels using multibeam sonarLegal Citation: M.L. 2013, Chp. 52, Sec. 2, Subd. 06aSubproject Manager: Jessica KozarekOrganization: University of MinnesotaSubproject Budget: $96,549Subproject Length and Completion Date: 1 Year; June 30, 2019Project Length and Completion Date: 6 Years; June 30, 2019Date of Report: August 9, 2019

ENVIRONMENT AND NATURAL RESOURCES TRUST FUND BUDGETBUDGET ITEM

Budget Amount Spent BalancePersonnel (Wages and Benefits) - Total $37,743 $37,655 $88 $37,743 $37,655 $88Project Manager/P&A (Kozarek); $18,003 ($13,415 salary, $4,588 benefits at 34.2% fringe rate); 20% FTE for 1 yearResearch Engineer/Civil Service (Milliren), $13,241 ($10,312 salary, $2,929 benefits at 28.4% fringe rate); 20% FTE for 1 year

Research Engineer/Civil Service (Feist); $698 ($544 salary, $154 benefits at 28.4% fringe rate); 1% FTE for 1 yearTechnician/Civil Service (Gabrielson); $2,371 ($1,846 salary, $524 benefits at 28.4% fringe rate); 4% FTE for 1 yearInstrument Specialist/Civil Service (Steen); $1,710 2710 ($2,110 1332 salary, $599 378 benefits at 28.4% fringe rate); 4% FTE for 1 Undergraduate Student; $1,720 ($1,720 salary, $0 benefits); 10% FTE for 1 year

Professional/Technical Services and Contracts - Total $48,771 $48,566 $205 $48,771 $48,566 $205United States Geological Survey, P. Grams - Advise acoustic remote sensing in experiments. Provide acoustic equipment. Contract includes personnel and travel.

$13,837 $13,837 $0

Northern Arizona University, D. Buscombe - Lead acoustic remote sensing in experiments. Contract includes personnel and travel.

$20,147 $20,146 $1

TOTALBALANCE

TOTAL BUDGET

TOTALSPENT

12

13

A B C D E F GENVIRONMENT AND NATURAL RESOURCES TRUST FUND BUDGETBUDGET ITEM

Budget Amount Spent BalanceTOTAL

BALANCETOTAL

BUDGETTOTALSPENT

25

2627

2829303132333435

363738394041

Macalester College, D. Hornbach, M. Hove, K. Macgregor - Mussel biology and field data collection. Provide research vessel and SCUBA equipment. Contract includes personnel, travel, and equipment maintenance (SCUBA).

$14,787 $14,584 $204

Equipment/Tools/Supplies - Total $10,035 $9,954 $81 $10,035 $9,954 $81Aquarium and lab supplies (tanks, secondary containment, coolers, aerators, sand/gravel, algae, plumbing, safety supplies, etc)

$6,878 $6,796 $81

Decontamination supplies (chemicals, pump, buckets, etc) $500 $500 $0Hydrophone $2,657 $2,657 $0

Capital Expenditures Over $5,000 - Total $0 $0 $0 $0 $0 $0NA $0 $0

Travel - Total $0 $0 $0 $0 $0 $0Travel - MN: Travel to field sites is covered under Macalester contract

$0

Other - Total $0 $0 $0 $0 $0 $0NA $0

COLUMN TOTAL $96,549 $96,175 $374 $96,549 $96,175 $374