Digital Earth Pacific – Approaches to monitoring fishery ecosystems

By Nicholas Metherall, Joeli Bili, Milika Sobey, Shyam Lodhia, Vanessa Dirking, Raphael Linzatti, Jesse Anderson and Sachindra Singh. 

Introduction
Mangrove and seagrass habitats provide a range of ecosystem services for fisheries as nursery sites that support spawning and recruitment (Unsworth et al. 2014; Nagelkerken et al. 2008). The breakdown of litter from both mangroves and seagrass beds supports microbial and planktonic food webs (Peduzzi and Herndl 1991), and these symbiotic relationships further support fisheries (Legendre and Rassoulzadegan 1995). However, many of these ecosystems face threats from environmental and climate-driven stressors as well as natural disasters (e.g. local processes on land, including sediment mobilisation and transport, resulting in downstream sediment burial of these fisheries). Global ocean processes, including thermal inertia and ocean acidification, are a further looming threat. These hazards present challenges for both future biodiversity and sustainable biomass yields of fisheries. Despite these ongoing challenges, there are limitations to the availability of verified data for monitoring mangrove forests, seagrass meadows and the impacts threats have on these ecosystems. Government authorities, environmental regulators and conservation agencies are constrained in their ability to systematically measure the extent of these ecosystems over time.

Digital Earth Pacific: A public technology infrastructure
Earth observation – using satellite imagery datasets – has emerged as an approach to support environmental monitoring across vast ecosystems, including those that support fisheries. Digital Earth Pacific (DEP) is a public technology infrastructure that has been built by the Pacific Community (SPC) in collaboration with its Pacific Island country and territory (PICT) members. The initiative supports these countries to access the cloud computer infrastructure of DEP, including levels of computer processing, access to satellite imagery databases and memory storage that, in the absence of DEP, would not be possible in the Pacific Islands region. To ensure that this technological capacity is made more accessible to PICTs, DEP has been made as an openaccess infrastructure that PICTs can use at no cost. This allows PICT member countries to save significant costs on accessing satellite data. To date, DEP has accessed 450,000 NASA Landsat and 300,000 European Space Agency Sentinel-2 satellite images. DEP has processed over 500 terabytes of data. If a country attempted to replicate this process, it would cost tens of millions in United States dollars (USD) to build the infrastructure, and a further USD 300,000 annually to operate. Through cloud computing, DEP can deliver this infrastructure at USD 30,000 per year; equivalent to a 1000% cost saving.

How earth observations can be used to support monitoring and decision-makers
Analysis of the large datasets accessed through DEP can produce a range of different data products. Evidence leads to insights, and these insights support policy-makers and land and ocean planners to make decisions. For example, as part of its coastline change data product, DEP has already mapped 22 years of seashore change along the 34,000 km of coastline within the Pacific Islands region. Similarly, the Water Observations from Space product has provided insights into surface water dynamics over the past 11 years. These decision-ready DEP products can help bridge the science–policy gap to support policy-makers.

MACBLUE supports DEP product for mangrove and seagrass extents
The Management and Conservation of Blue Carbon Ecosystems (MACBLUE) project is coordinated by the German Agency for International Cooperation (GIZ), in collaboration with the Secretariat of the Pacific Regional Environment Programme and SPC. MACBLUE seeks to support Fiji, Papua New Guinea, Solomon Islands and Vanuatu in mapping and monitoring seagrass and mangrove ecosystems. Mapping and remote sensing is a crucial part of blue carbon stock assessments, as the MACBLUE project intends to follow the Intergovernmental Panel on Climate Change guidelines (IPCC 2003). The mapped mangrove and seagrass habitats, in conjunction with verified ground truthing data, can be used to accurately quantify the carbon stocks within these ecosystems. The value of stored carbon, and the ecosystem services they provide, can be used as a basis for creating or improving policies to protect and conserve these ecosystems.

Through this MACBLUE project, participants from each of the participating countries will support in the co-design, data collection, calibration and validation of the DEP products, which will aim to map the extent of mangrove and seagrass ecosystems. Currently, DEP has generated datasets for the past seven years of mangrove forests throughout all PICTs. The next stages will include the further assessment, re-calibration and validation of these data products. Seagrass meadows will be a longer-term product, given the additional complexities associated with these commonly submerged intertidal ecosystems.

Localising earth observation: A case study from Papua New Guinea
The process of assessment, re-calibration and validation of mangrove and seagrass earth observation products has already begun in Papua New Guinea (PNG). This process has been initiated through the co-design of workshops facilitated by PNG’s Climate Change and Development Authority, GIZ and SPC. The workshop also involved collaboration with the Conservation and Environment Protection Authority, the PNG Forestry Authority, the National Fisheries Authority, and the University of Papua New Guinea. Through the workshop, participants provided input into areas of potential underestimation and overestimation of the extent of mangrove and seagrass areas based on existing earth observation products. Insights from this process were then incorporated into the design process for calibrating and validating the extent of DEP mangrove and seagrass areas.

Digital Earth Pacific’s methods for determining the extent of mangrove areas
The main methods used were based on peer-reviewed remote sensing studies, including those of Veldarrama-Landeros et al. (2018) and Tran et al. (2022). Both studies used spectral indices, including the normalised difference vegetation index-based classification approach. These kinds of approaches have been used for similar products such as Digital Earth Africa and Digital Earth Australia. The workflow includes the following seven steps:

1. Collation of datasets and images, including the Global Mangrove Watch (GMW) dataset, is generated as a baseline mask layer.
2. Sentinel 2 satellite imagery data is downloaded for all areas within the GMW baseline mask.
3. Sentinel 2 datasets, including many sentinel images across a single year, are pre-processed. These images are compressed into a median composite for each year.
4. The model is calibrated using spectral band values, including the normalised difference vegetation index (NDVI) and/or field data points. This is used to determine the extent of different vegetation types.
5. The model is used to classify or predict the extent of different vegetation types.
6. NDVI threshold values are applied to distinguish different categories of mangrove canopy density, including regular (lower density) canopy and closed (higher density) canopy cover.
7. Data may be validated by local partners and further field data collection.
8. With these inputs, the model may be re-trained and calibrated as part of a cyclical process to continuously improve model outputs.
9. Once there is a satisfactory output this output can be scaled across a wider region using (DEP) cloud computer capacity.

According to DEP’s technical product development team, there are several challenges in revising the methods used by global models to build mangrove extent maps for the Pacific region. These challenges include issues with cloud cover obscuring the view of satellite sensors, particularly during rainy seasons.

The next stages of incorporating further input from local countries into the DEP mangroves product will enable DEP to improve the GMW dataset. The next stages of development will make progress towards a first seagrass extent product and an enhanced mangroves product. Both DEP products will apply machine learning techniques and country-collated data points for capturing more refined extent and density. Incorporating these local inputs and localised data for validation and calibration enables DEP to provide a more accurate representation of the extent of mangrove areas at the local scale. To learn more, you can explore examples of these earth observations datasets and products at www.digitalearthpacific.org

References
_{IPCC (Intergovernmental Panel on Climate Change). 2003. Chapter 2 – Basis for Consistent Representation of Land Areas. In: IPCC Good practice Guidelines for Land Cover and Land Use Change and Forestry (LULUCF). Penman J., Gytarsky M., Hiraishi T. et al. (eds). Geneva, Swizerland: Intergovernmental Panel on Climate Change.}

_{Legendre L. and Rassoulzadegan F. 1995. Plankton and nutrient dynamics in marine waters. Ophelia 41(1):153–172. https://www.tandfonline.com/doi/ abs/10.1080/00785236.1995.10422042}

_{Nagelkerken I., Blaber S.J.M., Bouillon S., Green P., Haywood M., Kirton L.G., Meynecke J.-O., Pawlik J., Penrose H.M., Sasekumar A. and Somerfield P.J. 2008. The habitat function of mangroves for terrestrial and marine fauna: A review. Mangrove Ecology – Applications in Forestry and Coastal Zone Management 89(2):155–185. https://doi.org/10.1016/j. aquabot.2007.12.007}

_{Peduzzi P. and Herndl G.J. 1991. Decomposition and significance of seagrass leaf litter (Cymodocea nodosa) for the microbial food web in coastal waters (Gulf of Trieste, northern Adriatic Sea). Marine Ecology Progress Series 71(2):163–174. https://www.int-res. com/articles/meps/71/m071p163.pdf}

_{Tran T.V., Reef R. and Zhu X. 2022. A review of spectral indices for mangrove remote sensing. Remote Sensing 14(19):4868. https://www.mdpi.com/2072-4292/14/19/4868}

_{Unsworth R.K.F., Hinder S.L., Bodger O.G. and CullenUnsworth L.C. 2014. Food supply depends on seagrass meadows in the coral triangle. Environmental Research Letters 9: 094005. https://iopscience.iop. org/article/10.1088/1748-9326/9/9/094005}

_{Valderrama-Landeros L., Flores-de-Santiago F., Kovacs J.M. and Flores-Verdugo F. 2018. An assessment of commonly employed satellite-based remote sensors for mapping mangrove species in Mexico using an NDVI-based classification scheme. Environmental monitoring and assessment 190:1–13. https://link. springer.com/article/10.1007/s10661-017-6399-z/}

Authors
Nicholas Metherall, Joeli Bili, Milika Sobey, Shyam Lodhia, Vanessa Dirking, Raphael Linzatti, Jesse Anderson and Sachindra Singh. 

Contacts 
Aymeric Desurmont, Team Leader - Fisheries Information and Knowledge, Fisheries, Aquaculture, and Marine Ecosystems, SPC | [email protected] 
Tracey Holley, Science Networks and Knowledge Management Officer, Fisheries, Aquaculture, and Marine Ecosystems, SPC | [email protected] 

Download this article here. 
Download the full SPC Fisheries Newsletter here.