Global hotspots of salt marsh change and carbon emissions

Abstract

Salt marshes provide ecosystem services such as carbon sequestration1, coastal protection2, sea-level-rise (SLR) adaptation3 and recreation4. SLR5, storm events6, drainage7 and mangrove encroachment8 are known drivers of salt marsh loss. However, the global magnitude and location of changes in salt marsh extent remains uncertain. Here we conduct a global and systematic change analysis of Landsat satellite imagery from the years 2000–2019 to quantify the loss, gain and recovery of salt marsh ecosystems and then estimate the impact of these changes on blue carbon stocks. We show a net salt marsh loss globally, equivalent to an area double the size of Singapore (719 km2), with a loss rate of 0.28% year−1 from 2000 to 2019. Net global losses resulted in 16.3 (0.4–33.2, 90% confidence interval) Tg CO2e year−1 emissions from 2000 to 2019 and a 0.045 (−0.14–0.115) Tg CO2e year−1 reduction of carbon burial. Russia and the USA accounted for 64% of salt marsh losses, driven by hurricanes and coastal erosion. Our findings highlight the vulnerability of salt marsh systems to climatic changes such as SLR and intensification of storms and cyclones.

Addressing Tidal Flooding Induced uncertainties in Satellite Derived Global Salt Marsh Change Studies: Impact on Blue Carbon Monitoring

“…As an example, Bishop-Taylor et al 6 highlighted in their Figure 3 that the sun-synchronous overpass of Landsat can be spatially biased in capturing tidal cycles across the Australia coastline. This could affect Campbell et al's 1 estimations of loss and gain in regions such as the northern and southern Australian coasts (Campbell et al 1 : Figure 1b).…”

“…In a recent study, Campbell et al 1 conducted a large-scale change analysis using Landsat satellite data from 2000 to 2019 to estimate global changes in the salt marsh ecosystem coverage and their impact on blue carbon stock. Based on 20-yr satellite change detection, they reported that more than 1400 km 2 of marshes had been lost, with the USA and Russia contributing 64% of the total loss.…”

“…To demonstrate these uncertainties, we conducted a test comparing the impact of tidal signal on loss and gain of salt marsh area estimations in Landsat pixels following the same temporal intervals presented in Campbell et al 1 (Figure 1). We measured per-pixel differences in epoch averaged normalized vegetation index (NDVI) for data with no tide filtering as used in Campbell et al 1 and with tidal filtering 5 . The data were reduced by an average of 38% per pixel for each epoch after tide filtering.…”

“…Further, flooding frequency on a per-pixel basis is spatially dependent, increasing the probability that certain pixels may be flooded more often during Landsat overpasses. In Campbell et al 1 , the authors restricted their analysis to peak biomass months (August-September). However, some regions experience high seasonal variability in flooding frequency and magnitude, and depending on the location, these months may experience different flooding patterns and intensities 5 .…”

“…Because tidal flooding magnitude and duration are highly variable across space at local 8 and regional to global scales 10 , uncertainties resulting from tide contaminated pixels will vary in global studies. Further, longterm change analyses such as Campbell et al 1 are more likely to be affected by temporal and sampling bias caused by decadal tidal cycles and sun-synchronous satellite overpasses, as discussed by Bishop-Taylor et al 6 . As an example, Bishop-Taylor et al 6 highlighted in their Figure 3 that the sun-synchronous overpass of Landsat can be spatially biased in capturing tidal cycles across the Australia coastline.…”