How Grid Size Shapes the Accuracy of Ocean Accounting: Lessons from Saleh Bay, Indonesia

Deep beneath the surface of Saleh Bay in West Nusa Tenggara, Indonesia, coral reefs, seagrass meadows and mangrove forests quietly underpin the livelihoods of more than 5,800 fishers. Now, a new study published in One Ecosystem has revealed a surprisingly technical but consequential question: how big should the grid cells be when scientists tally up these underwater assets? The answer, researchers found, can swing the numbers that governments use to manage oceans, budget conservation and value ecosystem services.

The study tested Marine Basic Spatial Units, or MBSUs, the grid-based statistical cells that underpin ocean accounting aligned with the United Nations System of Environmental-Economic Accounting for Ecosystem Accounting (SEEA EA). Ocean accounting aims to bring economic, social and environmental information together in a coherent, standardised frame, treating nature as stocks that generate flows of ecosystem services. But while the SEEA EA standard works reasonably well on land, the ocean presents a messier picture: complex biophysical flows, patchy data and ecosystems whose boundaries rarely align with administrative maps.

Using empirical data from Saleh Bay, the research team compared two grid cell sizes, 10 by 10 metres and 25 by 25 metres, and two methods for assigning each cell to an ecosystem type. The dominance method assigns a cell to the ecosystem with the largest coverage within it, while the hierarchy method follows a criteria-based order of management priority, in this case ranking mangroves first, then coral reefs, then seagrasses. That produced four schemes to compare: 10H, 10D, 25H and 25D.

The baseline data, called the Source, came from Sentinel-2 satellite imagery, field ground-truthing and official Indonesian government maps, including the Indonesian Mangrove Map 2021. Analysis showed Saleh Bay holds 4,809.01 hectares of coral reef, 1,625.00 hectares of seagrass and 2,172.95 hectares of mangrove, within a total ecosystem extent account covering 357,539.96 hectares. The researchers generated more than 36 million grid cells at the finer resolution and nearly 5.8 million at the coarser one, then joined the Source layers to each scheme using GIS tools in ArcMap 10.8.

The results were striking. Larger grid cells produced higher errors in estimating both total area per ecosystem and the spatial configuration of assets across the bay. The worst performer was the 25-hectare-scale hierarchy scheme, 25H, with percentage errors ranging from 16.54 to 44.06 percent relative to the Source, an overall accuracy of just 73.56 percent and a KHAT score of 0.59, rated only moderate. The coarse hierarchy approach systematically inflated the extent of ecosystems it prioritised, particularly mangroves, whose fine, intricate features were absorbed into neighbouring cells regardless of actual coverage.

The dominance method fared far better. Both dominance schemes, 10D and 25D, stayed above the roughly 85 percent overall accuracy benchmark commonly used in remote sensing, with 10D rated almost perfect alongside the Source itself and 25D rated very good. The researchers attribute this to a compensating effect: when a non-dominant ecosystem type is squeezed out of one cell, it tends to dominate an adjacent cell, so area lost in one grid is offset in its neighbour. Errors in total extent remained minor, even though local-scale spatial distributions still shifted at ecosystem boundaries.

But accuracy comes at a price. Computation times varied enormously: the 10 by 10 metre hierarchy scheme took 941 minutes, the 10 by 10 dominance scheme 1,604 minutes, the 25 metre hierarchy scheme just 286 minutes and the 25 metre dominance scheme 459 minutes. The finer grids required roughly 400 percent longer for grid generation alone, and the dominance method needed 120 to 250 percent more processing time than its hierarchy counterpart at the same cell size, largely due to additional data pivoting in RStudio. For countries with limited computing capacity, those trade-offs matter.

The stakes extend well beyond cartography. Ecosystem extent accounts form the foundation for condition accounts, ecosystem services accounts and asset accounts, so errors introduced at the gridding stage can propagate through the entire accounting chain. Monetary valuations are typically expressed per hectare, meaning extent errors directly distort economic estimates. Carbon stock accounting for blue carbon ecosystems such as mangroves and seagrasses depends on accurate extent, and misclassification can lead to misplaced conservation budgets or ineffective protection measures, either through overestimation that wastes costs or omission errors that leave ecosystems unprotected.

In Indonesia, the implications are concrete. The archipelagic nation, with roughly 17,508 islands and more than 91,300 kilometres of coastline, began testing ocean accounting in 2020 and has committed to national-scale accounts as a member of the High-Level Panel for a Sustainable Ocean Economy. Ocean accounts feed directly into marine spatial planning, permit granting, the Suitability of Marine Space Use evaluation (PKKPRL) and marine protected area management, including the Liang Ngali MPA and the developing Lipan Rakit MPA within Saleh Bay itself. Errors in the spatial frame would ripple into all of these frameworks.

The authors recommend that practitioners carefully select grid size and assignment method according to their country’s management goals, the ecosystems present and available computational capacity. Where hierarchy methods and larger grids must be used, they urge that error and accuracy values be clearly reported and disclaimers added to linked accounts. Future research, they note, should explore weighted scoring approaches that consider multiple environmental and management factors, along with the influence of geomorphology and natural patterns, while cloud-based GIS computing may help dissolve the hardware constraints that currently force such trade-offs. For a field trying to put a credible price on the ocean, the humble grid cell, it turns out, is anything but a trivial choice.

Subject of Research: Optimising Marine Basic Spatial Units for ocean accounting using empirical data from Saleh Bay, Indonesia

Article Title: Optimising Marine Basic Spatial Units (MBSU) for Ocean Accounting using empirical data from Saleh Bay, Indonesia

Article References: Rahayu, A. K., Arhatin, R. E., Gacutan, J., Agung, F., Pingkan, J., Rosdiana, A., & Yulianto, I. (2024). Optimising Marine Basic Spatial Units (MBSU) for Ocean Accounting using empirical data from Saleh Bay, Indonesia. One Ecosystem, 9, Article e125578. https://doi.org/10.3897/oneeco.9.e125578

Image Credits: AI Generated

DOI: 10.3897/oneeco.9.e125578

Keywords: ocean accounting, marine basic spatial units, SEEA EA, ecosystem extent, mangroves, coral reefs, seagrass, Indonesia, Saleh Bay, remote sensing, natural capital accounting, marine spatial planning