**3. Implications for survey design**

Standardization of monitoring protocols across lagoons, although a EU regulatory requirement [34], is challenging because of the complex and varied range of conditions encountered across such environments. Identification of the best location where specific samples of water quality, habitat or phytoplankton are to be taken is usually difficult to determine due to the spatio-temporal variability present within and between lagoon environments and a priori lack of knowledge of the conditions within the lagoon. Recent studies have looked at developing statistically robust sampling protocols to address this gap in knowledge. The use of robotics and autonomous systems introduces continuous monitoring capability. This makes survey design easier by prioritizing continuous data collection over point sampling. From a statistical perspective, such approaches to data collection enables the estimation of unbiased measures of dispersion and central tendency, with less intensive requirements on determining where point sample should be taken. This is of special relevance when trying to disentangle the effects that multiple factors (e.g. management practice) have on the quality of the lagoon.

Palma et al. [35] studied the effect of sampling design on coral reef characterization when collecting high-resolution (0.4 cm) RGB imagery with semi-autonomous water vehicles (**Figures 5** and **6**). The authors were interested in determining seascape metrics that would provide information about the configuration of coral reefs in Ponta do Ouro Partial Marine Reserve (Mozambique) and the morphology of the site (**Table 1**). A range of sampling scales (quadrats of size 0.5 m × 0.5 m, 2 m × 2 m, 5 m × 5 m, 7 m × 7 m) and densities (from 1 to 100 quadrats) were compared. Results showed that sampling scales equal to or coarser than 5 m × 5 m and sampling densities equal to or larger than 30 were most effective along the 1655 m2 case study area. The study highlighted that special attention needs to be given to the design of coral reef monitoring programmes, with decisions being based on

#### **Figure 5.**

*The driver propulsion system (DPV), a remotely operated vehicle (ROV), equipped with a waterproof (wp) tablet and cameras. The tablet is used to coordinate data collection and steer vehicle direction.*

**151**

*Autonomous Systems for the Environmental Characterization of Lagoons*

the seascape metrics and statistics being determined. Although the Ponta do Ouro Partial Marine Reserve is not classed as lagoon, the results obtained are transferable

*Coral reef area sampled at Ponta do Ouro partial Marine Reserve in Palma et al. [35]. The image shows the different sampling strategies compared in the study (0.5 m × 0.5 m, 2 m × 2 m, 5 m × 5 m and 7 m × 7 m). Each sampling strategy depicts a different spatial configuration of the number and coverage of species (colored* 

More recent studies, also transferable to lagoon environments, have looked at the combined use of structure-from-motion (SfM) approach and ROV to map coral reefs and reduce the need for destructive sampling. In particular, Palma et al. [36] developed a framework for wide-scale benthic monitoring which is transferable to lagoon environments. The authors estimated population structure, morphology and biomass automatically from imagery collected with a (i) a GoPro Hero4 Black Edition (Woodman Labs, Inc., San Mateo, CA, USA) recording maximal resolution still images (4000 pixels × 3000 pixels) and (ii) a Sony Alpha NEX7 Digital Camera (Sony Corporation, Minato, Tokyo, Japan) recording full high-definition (1920 pixels × 1080 pixels) videos mounted on a ROV—the driver propulsion system (DPV) (**Figure 7**). The point clouds generated with both cameras contained more than 6.5 million points. Both the point cloud and the high-resolution imagery collected enabled the estimation of coral morphometries, such as height, width and planar surface of coral colonies. With the methodology proposed in [36], the error in coral height estimation was always <12.6 cm. For coral width estimation, the error was always <14.7 cm, whereas for the estimation

develop the methodology further to estimate coral ash free dry weight (AFDW) from the imagery collected based on the planar surface estimated. AFDW is the biomass weight present within the coral after oxidation of the organic component occurs at high temperatures. Eq. (1) is specific for *Paramuricea clavata* [37]. The results provided information on the overall health of coralligenous habitats within the Marine Protected Area of Portofino (Punta del Faro, Italy). The technology

*AFDW* = *A* ∙ 0.0047 ∙ 0.1515 (1)

. Palma et al. [36] were also able to

within 6 minutes, with data analysis requiring under

*DOI: http://dx.doi.org/10.5772/intechopen.90405*

to lagoon environments.

*polygons) present within the area.*

**Figure 6.**

of the planar surface, the error was 533 cm<sup>2</sup>

enabled sampling of 52 m<sup>2</sup>

10 hours of post-processing work:

*Autonomous Systems for the Environmental Characterization of Lagoons DOI: http://dx.doi.org/10.5772/intechopen.90405*

#### **Figure 6.**

*Lagoon Environments around the World - A Scientific Perspective*

for lagoon characterization (**Table 1**).

**3. Implications for survey design**

advancing the technology towards achieving an integrated system that enables the collection of collocated spatio-temporal information of all the parameters required

Standardization of monitoring protocols across lagoons, although a EU regulatory requirement [34], is challenging because of the complex and varied range of conditions encountered across such environments. Identification of the best location where specific samples of water quality, habitat or phytoplankton are to be taken is usually difficult to determine due to the spatio-temporal variability present within and between lagoon environments and a priori lack of knowledge of the conditions within the lagoon. Recent studies have looked at developing statistically robust sampling protocols to address this gap in knowledge. The use of robotics and autonomous systems introduces continuous monitoring capability. This makes survey design easier by prioritizing continuous data collection over point sampling. From a statistical perspective, such approaches to data collection enables the estimation of unbiased measures of dispersion and central tendency, with less intensive requirements on determining where point sample should be taken. This is of special relevance when trying to disentangle the effects that multiple factors (e.g. management practice) have on the quality of the lagoon. Palma et al. [35] studied the effect of sampling design on coral reef characterization when collecting high-resolution (0.4 cm) RGB imagery with semi-autonomous water vehicles (**Figures 5** and **6**). The authors were interested in determining seascape metrics that would provide information about the configuration of coral reefs in Ponta do Ouro Partial Marine Reserve (Mozambique) and the morphology of the site (**Table 1**). A range of sampling scales (quadrats of size 0.5 m × 0.5 m, 2 m × 2 m, 5 m × 5 m, 7 m × 7 m) and densities (from 1 to 100 quadrats) were compared. Results showed that sampling scales equal to or coarser than 5 m × 5 m and sampling densities equal to or larger than 30 were most effective along the 1655 m2 case study area. The study highlighted that special attention needs to be given to the design of coral reef monitoring programmes, with decisions being based on

*The driver propulsion system (DPV), a remotely operated vehicle (ROV), equipped with a waterproof (wp)* 

*tablet and cameras. The tablet is used to coordinate data collection and steer vehicle direction.*

**150**

**Figure 5.**

*Coral reef area sampled at Ponta do Ouro partial Marine Reserve in Palma et al. [35]. The image shows the different sampling strategies compared in the study (0.5 m × 0.5 m, 2 m × 2 m, 5 m × 5 m and 7 m × 7 m). Each sampling strategy depicts a different spatial configuration of the number and coverage of species (colored polygons) present within the area.*

the seascape metrics and statistics being determined. Although the Ponta do Ouro Partial Marine Reserve is not classed as lagoon, the results obtained are transferable to lagoon environments.

More recent studies, also transferable to lagoon environments, have looked at the combined use of structure-from-motion (SfM) approach and ROV to map coral reefs and reduce the need for destructive sampling. In particular, Palma et al. [36] developed a framework for wide-scale benthic monitoring which is transferable to lagoon environments. The authors estimated population structure, morphology and biomass automatically from imagery collected with a (i) a GoPro Hero4 Black Edition (Woodman Labs, Inc., San Mateo, CA, USA) recording maximal resolution still images (4000 pixels × 3000 pixels) and (ii) a Sony Alpha NEX7 Digital Camera (Sony Corporation, Minato, Tokyo, Japan) recording full high-definition (1920 pixels × 1080 pixels) videos mounted on a ROV—the driver propulsion system (DPV) (**Figure 7**). The point clouds generated with both cameras contained more than 6.5 million points. Both the point cloud and the high-resolution imagery collected enabled the estimation of coral morphometries, such as height, width and planar surface of coral colonies. With the methodology proposed in [36], the error in coral height estimation was always <12.6 cm. For coral width estimation, the error was always <14.7 cm, whereas for the estimation of the planar surface, the error was 533 cm<sup>2</sup> . Palma et al. [36] were also able to develop the methodology further to estimate coral ash free dry weight (AFDW) from the imagery collected based on the planar surface estimated. AFDW is the biomass weight present within the coral after oxidation of the organic component occurs at high temperatures. Eq. (1) is specific for *Paramuricea clavata* [37]. The results provided information on the overall health of coralligenous habitats within the Marine Protected Area of Portofino (Punta del Faro, Italy). The technology enabled sampling of 52 m<sup>2</sup> within 6 minutes, with data analysis requiring under 10 hours of post-processing work:

#### **Figure 7.**

*Image depicting the structure-from-motion methodology developed by Palma et al. [36] to sample corals without the need for destructive sampling. Overall view of the sampled area within the marine protected area of Portofino (Punta del Faro, Italy); (a) detailed view of a scanned coral branch and the automated estimation of its surface area; (b) sequence of images showing the implementation of the estimation of the surface area of corals on-site using SfM methods: (b1) point cloud generation, (b2) delineation of outmost boundary and (b3) estimation of the coral surface area via a small set of polygons.*

Technological advances in RAS and data processing algorithms enable more comprehensive data sets to be produced that facilitate more informed management decisions. The increased quality and quantity of data collected provides a robust foundation for the use of more advanced statistical methods than the estimation of measures of central tendency and dispersion.

**153**

*Autonomous Systems for the Environmental Characterization of Lagoons*

Remote sensing approaches including the use of satellites, UAVs, remotecontrolled boats and underwater vehicles provide the potential for significant advances in the understanding of the environmental characteristics and

functioning of lagoons. They can facilitate a better understanding of the temporal and spatial variation of environmental quality parameters, of habitat extent and condition, of risks, pressures and resultant responses and of the effectiveness of mitigation measures. They can contribute to coordinating and implementing nature-related policies [2], to the standardization of monitoring programmes ([34]) and to identifying environmental management priorities. They could also be used

Recent studies [2] have highlighted the need to increase research and technology

Remote sensing approaches clearly have an important role to play in the baseline assessment of a lagoon enabling detailed characterizations of habitats, morphology and quality. They can then be used to determine how these parameters vary within and between years including the impact of climate change. In addition, they can enable a better assessment of the condition of a lagoon, the pressures, responses and effectiveness of interventions, than existing methodologies. Whether such detailed characterizations are needed for all lagoons will be for individual managers and

There are few agreed protocols for the collection and interpretation of data using these techniques. This can limit their use in demonstrating compliance with legislative requirements. However, if remote sensing techniques do gain greater utilization in terms of routine monitoring including for legislative purposes, then this will significantly increase data transfer and storage capabilities and requirements. These monitoring approaches generate significant quantities of data that will have to be managed—the transfer and storage of this data could be a challenge. Agreed data collection and analysis protocols would facilitate the exchange of information and

These technologies produce information that has not routinely been available previously [31, 38], for example, spatial and temporal variations in a range of water

development (RTD) to enhance current lagoon management practices. For example, current understanding of the functioning and ecological quality of European lagoons is currently impaired by limited and incomplete data sets [2] such as lack of water quality measurements, gauging records, climate stations or water level stations. Further data weaknesses identified included insufficient water quality data in spatial and temporal dimensions for lagoon model calibration and validation. Based on a total of four case study areas, the work by Stålnacke et al. [2] concluded that effective lagoon management critically depends on high-quality data in geospatial format. Such data can be obtained with the remote sensing RAS solutions described in previous sections. However, there are several challenges to the deployment of remote sensing approaches and their widespread uptake by those responsible for the management and oversight of lagoons. Many of the techniques are still predominately the domain of the research community. There is as yet no purpose driven overarching monitoring and surveillance protocol for lagoons into which the use of remote sensing can be easily positioned. Thought has to be given to the use that will be made of the data that will be collected. For example, is it being collected because it is now possible to collect it or it will inform and improve the

*DOI: http://dx.doi.org/10.5772/intechopen.90405*

to better understand climate change impacts.

**4. Management considerations**

**4.1 Key challenges**

management of a lagoon.

organizations to determine.

enable intercountry comparisons to be made.
