Results

We used the application in the Explore Datasets tab to compare a range of mangrove extent, aboveground biomass and soil c maps. Here are some highlights of what we found:

Mangrove Extent

The top 10 mangrove holding nations are provided in Figure 1. For each mangrove area dataset we provide the 10 countries with the largest mangrove area. Alongside each of these are the mangrove areas in the other datasets for comparison. Indonesia contained the largest mangrove area in all datasets. Brazil had the second largest mangrove area in all datasets except in the CIFOR dataset where the USA was the second largest mangrove holding nation, although it did not appear in the top 10 for any other dataset. Australia, Malaysia, Nigeria, Myanmar, Papua New Guinea and Bangladesh feature in the top 10 of almost all of the datasets although their order is not consistently ranked.

A full matrix of the mangrove area maps are provided in Figure 2*. Each dataset is compared with every other dataset on a per country basis, demonstrating the level of agreement between maps. The CIFOR dataset is largely in agreement with the other maps for countries with smaller mangrove area, but increases in variability where mangrove extents are approximately 500,000 ha and above. The GMW and MFW datasets strongly agree and both differ only slightly with that of Mangrove Atlas for the larger mangrove holding nations. However, both demonstrate a consistent increase in mapped mangrove area over that of the CGMFC-21 map. Indeed, each of the mangrove maps shares this relationship with the CGMFC-21 map.

Globally, the distribution of mangrove forests varies between datasets with both agreement and disagreement within countries. Figure 3** shows the difference of each dataset from the median mangrove area per country as a percentage. The GMW and MFW map exhibit small differences in mangrove extent per country and are the most similar maps in terms of difference from the median area. The CGMFC-21 map shows greater differences particularly in the Middle East and South Asia. This is also true of the Mangrove Atlas dataset which has much larger mangrove areas in the Middle east then the median area and mangroves in nations under the administration of Russia at the time of data collection.

These trends, however, are not reflected in the standard deviation of the mapped areas (Figure 4**), where only a small range in values is mapped for much of Africa and the Middle East, with a larger range in values across Southeast Asia and Oceania. Variability is also mapped across the Americas, particularly in the United States, Mexico and Brazil.

It should be noted that, in general, the CIFOR dataset appears to vastly overestimate mangrove extent, with mangrove forests mapped at unsuitable latitudes and is judged to be the least reliable extent map. Furthermore, the CGMFC-21 map estimates tree cover, which underestimates mangrove extent in comparison to the other presence/absence maps.

*Indonesia is omitted from this analysis as its very large mangrove area has a significant impact upon the country-by-country regression which masks differences between countries with smaller differences between datasets

**The CIFOR dataset is omitted from these analyses. This map exhibits large erroneous differences from the other datasets and are therefore not included.

Mangrove Aboveground Biomass (AGB)

There is little agreement in the top 10 most dense biomass mangrove holding nations between the Simard and Hutchinson products (Figure 5). Only Papua New Guinea, Brunei and Indonesia feature in the top 10 in both datasets, each occupying a different position in the list. Small island nation states feature heavily in the datasets, including Guam, Solomon Islands, Northern Mariana Islands and Federated States of Micronesia. The top 10 AGB Simard values have approximately a 100% difference in AGB between position 1 and 10, while the Hutchinson top 10 show little variation, with each containing approximately 350 Mg/ha⁻¹. Of the top 10 biomass holding nations in the Simard and Hutchinson datasets, 8 of the datasets feature in both products (Figure 5). Indonesia is ranked first on both lists, dwarfing the other countries in the top10. Aside from Indonesia, none of the products share the same order with the exception of Myanmar which is ranked 8 in both datasets. The Simard dataset features Venezuela and Cameroon, while the Hutchinson dataset features Cuba and Mexico, accounting for the differences between the two top 10 lists.

There is variability in both the mean AGB density and total ABG per country between the Hutchinson and Simard datasets (Figure 6). The largest differences between the total AGB stored in each country is between that of Brazil and Bangladesh, which exhibit the largest negative and largest positive differences, respectively. The largest negative residual describes almost 60 Tg more of AGB in the Hutchinson dataset than the Simard. This was reversed in Bangladesh where the Simard dataset contained almost 20 Tg more AGB. However, the majority (51) of countries contained less than 1 Tg of difference. However, as a % of the Hutchinson total, different trends were observed, with 9 of the countries exhibiting a difference less than +/- 10%. Similarly, 21 of the countries had a residual mean density of within +/- 20 Mg/ha, demonstrating the variability on AGB density between individual countries.

Mangrove Soil Carbon (C)

Across all four soil c datasets, Indonesia contained the greatest cumulative total of mangrove soil C. This is predominantly due to Indonesia containing the greatest mangrove area. Brazil, Malaysia, Mexico, Australia and Nigeria frequently appear in the top 6 in each dataset although appear in different orders, again reflective of these countries being large mangrove holding nations (Figure 7). The Sanderman dataset predominantly predicts the greatest mangrove soil C across all top 10 countries (per dataset) while the Rovai dataset consistently predicts the lowest soil C content. It is important to note that the Atwood dataset is based upon the CGMFC-21 extent map which measures extent as a function of tree cover, despite tree cover not being a good proxy for soil cover/type.

While the top 10 total mangrove soil c nations are in general agreement, this is not true of the mean soil c per country. Figure 8* demonstrates this on a per country basis with very little agreement between datasets with low r-square and high MAE values. No two datasets are in agreement, unlike the mangrove area maps, demonstrating the variability that must be accounted for when estimating soil c values per country, across a large geographical domain.