Assessing the aquatic chemistry of water bodies through sample collection is labor- and time-intensive with limits on discrete spatial coverage that may not provide a detailed representation of the system. A practical approach is to develop in situ sensors deployed aboard autonomous underwater vehicles (AUVs) for three-dimensional water chemistry mapping. For this purpose, a compact optical instrument (LEDIF) measuring fluorescence, absorbance, and scattering to quantify contaminants and natural substances in water bodies is packaged inside a pressure hull and attached to a highly modular and flexible AUV (Small Team of Autonomous Robotic FISH (STARFISH)). LEDIF-STARFISH was deployed at a reservoir in Singapore for in situ real-time chlorophyll a and turbidity data collection. Locations of potential algal hot spots were observed, providing unprecedented insight into the plankton biomass distribution of the reservoir at different times. The results showcase the instrument's potential in tracking spatiotemporal variability of substances in large water bodies.

INTRODUCTION

In situ water chemistry instruments offer many advantages over existing lab-based techniques such as flexible deployment options (e.g. autonomous vehicles, buoys), enhanced spatial and temporal data resolution, and quicker turnaround time from deployment to measurement results. Understanding the biogeochemical processes of large water bodies requires access to spatiotemporal variability of water chemistry information that is not feasibly obtained with the collection of water samples. With the advance of technology, robotic environmental sensing in general has increasingly become viable and has been attracting a swath of developments in the past decade (Bogue 2011; Dunbabin & Marques 2012). Researchers have deployed surface crafts to support water quality research in various degree of robotic autonomy (Casper et al. 2009; Dunbabin et al. 2009; Koprowski et al. 2013) with limited flexibility in vertical measurement. A practical approach is to deploy a water chemistry sensor aboard an autonomous underwater vehicle (AUV), and make use of the mobility of the platform to obtain real-time data for the generation of three-dimensional (3-D) water chemistry maps. The use of AUVs (Ellison & Slocum 2008; Hemond et al. 2008; Godin et al. 2011) for environmental monitoring has been demonstrated using customized instrument targeted for specific applications. In this paper, an optical sensing device (LEDIF) capable of detecting a wide range of contaminants and natural substances in real-time is instrumented aboard an AUV named STARFISH for 3-D chemical mapping. LEDIF is an optical sensor developed by Ng, Hemond, and Senft-Grupp at Center for Environmental Sensing and Modeling of Singapore-MIT Alliance for Research and Technology, for sensing a range of non-volatile substances in the freshwater and marine environments. STARFISH AUV is an autonomous platform developed at Acoustic Research Laboratory of National University of Singapore to support research in the area of environmental sensing and robotic intelligence.

The aim of this work is to test both systems, functioning as one instrument, to perform 3-D chemical mapping at a reservoir. The compound of interest, chlorophyll a, is chosen because it is a pigment produced by algae and is commonly regarded as a proxy for algal biomass, a key parameter of aquatic systems. Turbidity, another important metric, is also measured at the same time to determine the cloudiness of the water. The objectives are to complete long range (over 3 km, covering 1,000 × 400 m) surface and dive missions to verify LEDIF-STARFISH's capability in detecting spatiotemporal variation in chlorophyll a concentration, identify short-term spatial changes with additional surveys on the same day, and compare results with missions on later days. Results are reported in the form of 3-D water chemistry maps, providing insight into spatial distribution of plankton biomass in both the horizontal and vertical space of the reservoir.

METHODS

A compact modularized real-time sensor (LEDIF) was integrated into an AUV (STARFISH) for in situ sensing of water chemistry. A doppler velocity log (DVL) was also employed to improve underwater location tracking. The collected data was used to generate 3-D maps of chlorophyll a and turbidity of the reservoir and assess the extent of spatial and temporal variability of the signals.

LEDIF sensor

LEDIF is an in situ instrument developed to measure water chemistry in the field. It uses a custom tri-optical design consisting of multi-excitation fluorescence, broadband absorbance, and scattering to detect and quantify chemical compounds in water. The sensor uses six optically enhanced LED excitation sources for fluorescence and scattering measurements and an optically enhanced deuterium-tungsten lamp for absorbance measurements. The spectra are captured with a spectrometer. An ARM single board computer (or a microcontroller board) and a power management board are used for power regulation, data storage, data processing, and external communication (Figure 1). Sensor operations are controlled with custom software that allows users to define and automate sensing tasks based on operation needs. A detailed description of LEDIF design and materials is discussed in Ng et al. (2012a, 2012b).
Figure 1

Design layout and photo of LEDIF sensor packaged inside a pressure hull.

Figure 1

Design layout and photo of LEDIF sensor packaged inside a pressure hull.

LEDIF can operate autonomously, in conjunction with an autonomous vehicle, or under direct user control. LEDIF has been packaged inside a 200 mm diameter and 300 mm length pressure payload hull (Figure 1) for AUV and surface platforms deployment, a smaller (200 × 200 × 150 mm) rectangular box for portable sensing, and a T-shaped enclosure for long-term fixed location monitoring. LEDIF can be used to detect pigments such as chlorophyll a and b, dissolved organic carbon, high molecular weight hydrocarbons (e.g., polycyclic aromatic hydrocarbons), low volatile hydrocarbons (e.g., Benzene, Ethylbenzene, Toluene, and O-xylene), pesticides, among other organic analytes. The assessment of LEDIF together with the comparison of LEDIF with several commercial instruments (Figure 2) are discussed in Ng et al. (2014). LEDIF measurement matches well with the commercial instruments in fluorescence, absorbance, and turbidity measurements.
Figure 2

Comparison of LEDIF with several commercial laboratory instruments in (a1, a2) fluorescence (Legend: S_RF-5301 is Shimadzu RF-5301 spectrofluorometer), (b1, b2) absorbance (Legend: OO is Ocean Optics USB-ISS-UV/Vis light source coupled with USB4000 spectrometer. S_UV-2550 is Shimadzu UV-2550 spectrophotometer), and (c) scattering (Legend: O_AF is Orion AquaFast AQ3010 turbidity meter) measurements (Ng et al. 2014).

Figure 2

Comparison of LEDIF with several commercial laboratory instruments in (a1, a2) fluorescence (Legend: S_RF-5301 is Shimadzu RF-5301 spectrofluorometer), (b1, b2) absorbance (Legend: OO is Ocean Optics USB-ISS-UV/Vis light source coupled with USB4000 spectrometer. S_UV-2550 is Shimadzu UV-2550 spectrophotometer), and (c) scattering (Legend: O_AF is Orion AquaFast AQ3010 turbidity meter) measurements (Ng et al. 2014).

STARFISH AUV

STARFISH, abbreviated from Small Team of Autonomous Robotic FISH, are low cost AUVs designed to support scientific research in cooperative robotics and environmental sensing down to maximum depth of 100 m (Koay et al. 2011). STARFISH AUVs are comprised of cylindrical sections with distinct functionalities mated together to form a fully sealed torpedo shape vehicle. It is a highly modular system with well-defined electrical, mechanical, and software interfaces that allow straightforward addition of scientific payloads independently developed by collaborators.

The basic STARFISH AUV comprises of a nose sensor section, a command and control section, and a tail section. The configuration weighs less than 45 kg, measures 1.6 m in length with a nominal diameter of 20 cm. The AUV carries a set of basic sensors and actuators such as forward looking sonar, depth sensor, altimeter, compass, GPS, an internal roll ballast, thrusters, and control surfaces. Each section carries a standard electrical interface comprising of an Ethernet communication backbone, a system wide power bus and a battery bus. The inter-section communication backbone is based on an Ethernet link that is abstracted to a software interface that allows sections to either subscribe to periodic notifications of sensors and vehicle data or directly access a specific sensor. The ability of nodes to directly communicate is useful when additional information is required or when data latency must be below 1 ms. The STARFISH AUV power is distributed across battery packs in each section and connected via a battery bus, forming a single, pooled energy resource. The battery bus is then regulated and supplied to all sections through the power bus, allowing optimal energy distribution and use. Independent payloads, such as LEDIF, have the option to utilize STARFISH vehicle power, if needed. Once installed, the payload section can be accessed through the existing vehicular infrastructures (Figure 3). Mission plans, including sensor settings, can be communicated through the wireless network interface to the vehicle. Mission directives can be communicated over WiFi or GSM network (using SMS service) when the vehicle is on the surface. When the AUV is performing a dive mission, communication is through acoustic modems.
Figure 3

Payload integration and human-machine interface channels of STARFISH AUV.

Figure 3

Payload integration and human-machine interface channels of STARFISH AUV.

LEDIF aboard STARFISH

The STARFISH AUV is the preferred AUV platform to mobilize LEDIF in Singapore waters. Functioning as one instrument, LEDIF-STARFISH collects real-time water chemistry data according to a pre-planned path (Figure 4). With a DVL payload, LEDIF-STARFISH measures 2.4 m in length and weighs approximately 60 kg, allowing for easy 2 to 3 person deployment. The operational depth of LEDIF-STARFISH is 100 m. Figure 5 shows the calibration curves of chlorophyll a and turbidity constructed using laboratory prepared standards. It can be used to quantify the concentration of chlorophyll a and turbidity measurements shown in the 3-D water chemistry maps of this paper.
Figure 4

LEDIF-STARFISH underway to mission start point on the surface.

Figure 4

LEDIF-STARFISH underway to mission start point on the surface.

Figure 5

Calibration curves of (a) chlorophyll a and (b) turbidity utilizing laboratory prepared standards.

Figure 5

Calibration curves of (a) chlorophyll a and (b) turbidity utilizing laboratory prepared standards.

As STARFISH travels, water is drawn into the through hull liquid manifold of LEDIF for measurements. When the AUV surfaces, the collected data is communicated over WiFi to a laptop and used to generate targeted water chemistry maps with a simple level set method. To achieve significant spatial coverage in a short time, multiple AUVs can be deployed for concurrent mapping of different areas at different depths. While aboard the STARFISH, LEDIF can operate in several modes (e.g., surface chemical mapping).

In passive mode, LEDIF makes independent measurements as the vehicle executes its pre-planned mission. LEDIF receives sub-second position updates and mission status from STARFISH, and can adapt its sensing routines based on the observed vehicle state. In the under-development adaptive sampling mode, LEDIF makes use of STARFISH's distributed software architecture (Chitre 2008) and its agent based command and control framework (Tan & Chitre 2012) to influence the execution of the mission. LEDIF may request a detour from the original mission, insert a new task, change the execution parameters (e.g., vehicle speed), or even overwrite the remaining mission points with new ones. These requests are evaluated by the AUV command and decision making routines and, where appropriate and feasible, used to update the vehicle's overall objectives. An integration diagram between LEDIF and other agents in STARFISH AUV is shown in Figure 6. This adaptive sampling mode could eventually be used for delineating a plume or tracking the source of contaminants.
Figure 6

Interactions diagram between a typical software agent from payload (Scientist, LEDIF) and other agents in the STARFISH AUV.

Figure 6

Interactions diagram between a typical software agent from payload (Scientist, LEDIF) and other agents in the STARFISH AUV.

RESULTS AND DISCUSSION

LEDIF-STARFISH was deployed at a reservoir in Singapore on five separate days between October 2013 and March 2014. 3-D topographic and contour maps of chlorophyll a generated with the collected data in each mission are shown in Figures 79. A mission is defined as a complete collection of 3-D water chemistry data over the planned vehicle path line (cross symbol in figures) that is completed in a specific timeframe (mission time). The contour maps were generated using simple level set method with the collected data for the timeframe stated in the caption of the figures. Given LEDIF's sample rate and STARFISH's velocity, the instrument was capable of recording data points with an approximate resolution of 10 m. The data demonstrate the instrument's ability to identify that chlorophyll a concentration is a function of both location and depth. Missions repeated several hours apart and over several weeks also demonstrate a time dependence of chlorophyll a concentration in this system. It was observed that the chlorophyll a concentrations is higher in lower depth; however, understanding the profile and the cause of spatial variability of the reservoir will require a more rigorous and comprehensive studies. To assess the spatial variability of a water body with concentration profile that is highly dynamic, an additional unit of LEDIF-STARFISH can be employed for cooperative mission mapping to attain the same spatial coverage in a shorter mission time.
Figure 7

3-D chlorophyll a maps of different mission days. (a1) 18 October 13, mission time: 11:18:24 to 12:08:53, (a2) 18th October 13, mission time: 12:28:29 to 13:25:57, (b1) 25 October 13, mission time: 13:04:07 to 14:03:56, (b2) 25 October 13, mission time: 15:12:34 to 16:11:42, (c) 8 Nov. 13, mission time: 11:20:37 to 12:10:56. Cross symbol in maps represent LEDIF-STARFISH mission path line.

Figure 7

3-D chlorophyll a maps of different mission days. (a1) 18 October 13, mission time: 11:18:24 to 12:08:53, (a2) 18th October 13, mission time: 12:28:29 to 13:25:57, (b1) 25 October 13, mission time: 13:04:07 to 14:03:56, (b2) 25 October 13, mission time: 15:12:34 to 16:11:42, (c) 8 Nov. 13, mission time: 11:20:37 to 12:10:56. Cross symbol in maps represent LEDIF-STARFISH mission path line.

Figure 8

Typical 2-D chlorophyll a contour maps of (a1, a2) 18 October 13, mission time: 11:18:24 to 12:08:53, (b1, b2) 18th October 13, mission time: 12:28:29 to 13:25:57. Cross symbol in maps represent LEDIF-STARTFISH mission path line.

Figure 8

Typical 2-D chlorophyll a contour maps of (a1, a2) 18 October 13, mission time: 11:18:24 to 12:08:53, (b1, b2) 18th October 13, mission time: 12:28:29 to 13:25:57. Cross symbol in maps represent LEDIF-STARTFISH mission path line.

Figure 9

Surface chlorophyll maps on (a) 13 Feb. 14, mission time: 14:39:43 to 15:23:02 and (b) 6 March 14, mission time: 10:48:58 to 11:33:57. Cross symbol in maps represent LEDIF-STARTFISH mission path line.

Figure 9

Surface chlorophyll maps on (a) 13 Feb. 14, mission time: 14:39:43 to 15:23:02 and (b) 6 March 14, mission time: 10:48:58 to 11:33:57. Cross symbol in maps represent LEDIF-STARTFISH mission path line.

Additionally, LEDIF-STARFISH measurements were used to simultaneously estimate and map the turbidity of the reservoir (Figure 10). Three short missions on 28 March 2013 covering small areas reveal potential ‘hot spots’ on the scale of approximately 50 m. A longer mission on 18 October 2013 showed relatively homogeneous turbidity over different depths and the full extent of the mission area, signifying turbidity is low except at a few locations marked by LEDIF-STARFISH vehicle path in the mission. At locations out of the vehicle path and out of the ‘hot spots’, simple level set method was applied utilizing the collected data and it estimated a low turbidity value of a few Nephelometric Turbidity Unit in these areas. Future work will start to investigate the relationships between measured parameters.
Figure 10

Typical 3-D turbidity map generated using LEDIF-STARFISH real-time data. (a) 4 short missions (28 March 2013, mission time: 14:13:41 to 14:18:23, 14:54:34 to 14:57:30, 15:03:51 to 15:06:45, 15:16:52 to 15:29:39) and (b) a long mission (18 October 2013, mission time: 13:32:55 to 14:26:50). Cross symbol in maps represent LEDIF-STARTFISH mission path line.

Figure 10

Typical 3-D turbidity map generated using LEDIF-STARFISH real-time data. (a) 4 short missions (28 March 2013, mission time: 14:13:41 to 14:18:23, 14:54:34 to 14:57:30, 15:03:51 to 15:06:45, 15:16:52 to 15:29:39) and (b) a long mission (18 October 2013, mission time: 13:32:55 to 14:26:50). Cross symbol in maps represent LEDIF-STARTFISH mission path line.

The results demonstrate the repeated ability to deploy LEDIF aboard the STARFISH AUV and convert independent chemical measurements into 3-D, time-varying concentration maps. For these missions, with LEDIF operating in passive mode, concentration peaks and variations were easily identifiable. However, features such as precise ‘hot spot’ delineation, were not feasible. Those abilities require the next level of LEDIF and STARFISH integration to enable adaptive sampling and route planning, and identify the need for continuing research with this instrument and mobile platform. It shall be reiterated that out of LEDIF-STARFISH mission path line, the water chemistry maps were obtained using simple level set method.

CONCLUSIONS

An in situ real-time optical sensor (LEDIF) with fluorescence, absorbance, and scattering measurement capabilities is deployed aboard a STARFISH AUV for 3-D topographic chemical mapping. The instrument is capable of observing a variety of important compounds found in natural waters. This project provides a powerful tool to understand the spatiotemporal variability of water quality in water bodies and water management systems, and can help improve strategic planning, economical value, and control of regulated water systems. When coupled with adaptable formation control of multiple AUVs, it can be used to improve water safety and security of large consumable water sources. It also provided an unprecedented insight into the phytoplankton biomass spatiotemporal distribution at a local reservoir.

REFERENCES

REFERENCES
Bogue
R.
2011
Robots for monitoring the environment
.
The Industrial Robot
38
,
560
566
.
Casper
A.
Steimle
E.
Hall
M.
Dixon
B.
2009
Combined GIS and ROV technologies improve characterization of water quality in coastal rivers of the Gulf of Mexico
.
OCEANS 2009, MTS/IEEE Biloxi – Marine Technology for Our Future: Global and Local Challenges
2009
,
1
9
.
eISBN: 978-0-933957-38-1.
Chitre
M.
2008
Dssav – a distributed software architecture for autonomous vehicles
.
OCEANS 2008
,
1
10
.
DOI: 10.1109/OCEANS.2008.5151848
Dunbabin
M.
Marques
L.
2012
Robots for environmental monitoring: significant advancements an
.
Applications Robotics Automation Magazine, IEEE
19
,
24
39
.
Dunbabin
M.
Grinham
A.
Udy
J.
2009
An autonomous surface vehicle for water quality monitoring australasian
.
Conference on Robotics and Automation 2009
,
2
4
.
ISBN: 978-0-980740-40-0.
Godin
M.
Zhang
Y.
Ryan
J.
Hoover
T.
Bellingham
J.
2011
Phytoplankton bloom patch center localization by the tethys autonomous underwater vehicle
.
OCEANS
2011
,
1
6
.
Hemond
H.
Mueller
A.
Hemond
M.
2008
Field testing of lake water chemistry with a portable and an AUV-based mass spectrometer
.
Journal of the American Society for Mass Spectrometry
19
(
10
),
1403
1410
.
DOI: 10.1016/j.jasms.2008.04.019.
Koay
T. B.
Tan
Y. T.
Eng
Y. H.
Gao
R.
Chitre
M.
Chew
J. L.
Chandhavarkar
N.
Khan
R. R.
Taher
T.
Koh
J.
2011
STARFISH – a small team of autonomous robotic fish
.
Indian Journal of Geo-Marine Sciences
40
,
157
167
.
Koprowski
R.
Wrobel
Z.
Kleszcz
A.
Wilczyski
S. A.
Wonica
A.
Ozowski
B.
Pilarczyk
M.
Karczewski
J.
Migula
P.
2013
Mobile sailing robot for automatic estimation of fish density and monitoring water quality
.
BioMedical Engineering OnLine
1
,
12:60
.
DOI: 10.1186/1475-925X-12-60.
Ng
C.-L.
Hemond
H.
Senft-Grupp
S.
2012a
Highly compact multi-optical-junction optical flowcell and flexibility deployable optical sensing assemblies and systems for in-situ real-time spectroscopic measurements. Patent Corporation Treaty Application No. PCT/SG2012/000142
.
Ng
C.-L.
Senft-Grupp
S.
Hemond
H.
2012b
A multi-platform optical sensor for in situ sensing of water chemistry
.
Limnology and Oceanography: Methods
10
,
978
990
.
Ng
C.-L.
Teo
W.-K.
Cai
H.-T.
Hemond
H.
2014
Characterization and field test of a multi-platform optical sensor
.
Limnology and Oceanography: Methods
12
,
484
497
.
Tan
Y.-T.
Chitre
M.
2012
Hierarchical multi-agent command and control system for autonomous underwater vehicles
.
IEEE/OES 2012
,
1
6
.
DOI: 10.1109/AUV.2012.6380760
.