Skip to main content
Dryad logo

Dispersal and coastal geomorphology limit potential for mangrove range expansion under climate change

Citation

Raw, Jacqueline (2022), Dispersal and coastal geomorphology limit potential for mangrove range expansion under climate change, Dryad, Dataset, https://doi.org/10.5061/dryad.4qrfj6qd7

Abstract

Latitudinal range limits for mangroves on high-energy, wave-dominated coasts are controlled by geomorphological features and estuarine dynamics. Mangroves reach a southern global range limit along the South African coastline, but the distribution is patchy, with stands occurring in only 16% of the estuaries in the region. Yet, the persistence of forests planted >50 years ago beyond the natural distribution limit suggests that additional estuaries could support mangroves. Understanding regional drivers is necessary to inform global scale estimates for how this important ecosystem is predicted to respond to climate change.

Here, we combine species distribution modeling (MaxEnt), Lagrangian particle tracking using an eddy- and tide-resolving numerical ocean model, and connectivity matrices, to identify suitable mangrove habitats at present, as well as under the IPCC RCP4.5 and RCP8.5 climate scenarios.

Within the current South African distribution range (± 900 km), eight more estuaries were identified as suitable under contemporary conditions. When considering potential range extension (± 110 km), an additional 14 suitable estuaries were identified. Connectivity matrices suggest limited long-distance dispersal, stranding mostly at or near the release location, and a decreased probability of connectivity towards the range limit. Under both future climate scenarios, 30% of estuaries currently supporting mangroves are predicted to become unsuitable, while an additional six estuaries beyond the current distribution are predicted to become suitable. However, there is limited connectivity between these new sites and established forests.

Synthesis: This study shows that dispersal substantially limits mangrove distribution at the southern African range limit and highlights the importance of including this process in species distribution models. Ultimately, our results provide new insight for mangrove conservation and management at range limits that are not controlled predominantly by temperature, as it has been assumed that mangroves will largely expand to higher latitudes under climate change.

Methods

Mangrove data

Mangroves occur in 31 of the 192 estuaries along the eastern South African coast, over a tropical to warm-temperate biogeographical transition zone (latitudinal range of 26°53’42.6’’ – 33°13’32.8’’ S). Data for mangrove area cover (ha) in South African estuaries were collated from the National Estuarine Botanical Database (updated in 2019) (http://bgis.sanbi.org/SpatialDataset/Detail/2687) (Adams & Rajkaran, 2021).

Environmental data

For the development of species distribution models (SDMs), environmental data are needed from the occurrence locations (as part of the background data), as well as from locations where mangroves do not occur (as part of the landscape data). As mangroves are restricted to estuaries, only environmental data from estuaries (and not areas along the open coast) were included in the model.

The estuarine environmental drivers of mangrove distribution along the South African coastline have previously been identified by Raw et al. (2019a). The following variables were therefore used as the basis for the SDMs in this study: average annual land temperature (°C), floodplain area (ha), mean annual runoff (m3 × 106), daily flushing rate (m3 × 106 d-1), and estuary mouth state (% time that the estuary is connected to the ocean). Data for these variables were obtained from the Council of Scientific and Industrial Research (CSIR) national database (updated in 2019) for the physical characteristics of South Africa’s estuaries (Van Niekerk et al., 2017). This dataset consists of a combination of published, observed data (> 20 years), and modelled data. For example, given the paucity of measured river inflow along the South African coast, mean annual runoff is based on modelled values over more than 50 years. The daily flushing rate is calculated from the estimated mean annual runoff, mapped open water extent, and average estuary depth. Environmental data were provided in tabular format for all estuaries along the east coast of South Africa, from Kosi Estuary in the north to Boknes Estuary in the south (~1000 km, n = 214).

The environmental variables for the 2050 projection models were developed by extracting temperature and rainfall projections for South Africa under the IPCC Representative Concentration Pathway (RCP) scenarios which consist of rectangular rasters at a spatial resolution of 8 km2 for each municipality (Engelbrecht et al., 2019). Because South African estuaries are relatively small, in most cases they were matched one-to-one with these grid cells. Furthermore, given the limited mangrove extent inland from estuary mouths, changes predicted within the first grid cell adjacent to the coast were considered the most relevant, even for longer estuaries. We considered two RCP scenarios, namely RCP4.5 and RCP8.5. The RCPs each define a different possible climate scenario based on greenhouse gas concentration trajectories for the 21st century and are labelled based on radiative forcing values by the year 2100 (2.6, 4.5, 6.0, and 8.5 W m-2 for each respective scenario). RCP4.5 is an intermediate scenario where emissions are projected to peak by 2040 and then decline towards 2100. In RCP8.5, emissions are projected to continue to rise throughout the 21st century, it is considered a worst-case, yet realistic, climate scenario.

The projected 2050 temperature and rainfall data were extracted for the geographical location of each estuary. The average temperature increase and average predicted change in rainfall (in mm) by 2050 were calculated for each scenario from the 10th and 90th percentile predictions. For temperature, the value for 2050 was calculated by adding the predicted increase to the current mean annual land temperature. This value ranged from 1.35 – 1.78°C under RCP4.5 and 1.73 – 2.13°C under RCP 8.5. For rainfall, the percent difference in the predicted value from the current rainfall was calculated and used to scale mean annual runoff, daily flushing rate, and estuary mouth state variables for 2050. Change in rainfall was calculated to range from a 7.8% decrease to a 3.3% increase under RCP4.5, while under RCP8.5 it was calculated to range from a 10.4% decrease to a 4.2% increase. Floodplain area was assumed to remain unchanged by 2050.

Modelling mangrove distribution with MaxEnt

MaxEnt is an open-source software that uses the principles of maximum entropy for modelling spatial species distributions from presence-only species records and environmental factors (predictor variables) that are relevant for the species’ habitat suitability (Phillips et al., 2019). This is achieved by MaxEnt randomly sampling background locations (i.e., where presence of the species is unknown) within the user-defined model domain (study area) and comparing characteristics of the predictor variables with environmental characteristics at the recorded occurrence locations. Areas of suitable habitat beyond the current occurrence locations can then be identified. The distribution model can be projected to new locations or under new environmental conditions.

All models were run in MaxEnt v. 3.4.1 (Phillips et al., 2019) initially with the data randomly partitioned into a 75% training set and a 25% testing set. After evaluating model fit, the models were re-run using the whole dataset as the testing set. The species-with-data format was used, as both the occurrence localities and environmental data were derived for individual estuaries and can therefore be represented as point-attributes (Elith et al., 2011). The models were run with linear, quadratic, and hinge features, which were selected based on prior defined relationships between the environmental variables and mangrove area (Raw et al., 2019a). Jack-knife tests of variable importance were used in each model and the AUC (area under curve) statistic was used as a measure of model performance.

To compare mangrove suitability between estuaries, the ‘cloglog’ output from the MaxEnt models was extracted; this provided an estimate for probability of occurrence between 0–1.

References

Adams, J. B., & Rajkaran, A. (2021). Changes in mangroves at their southernmost African distribution limit. Estuarine, Coastal and Shelf Science, 248, 107158. https://doi.org/10.1016/j.ecss.2020.107158

Elith, J., Phillips, S. J., Hastie, T., Dudík, M., Chee, Y. E., & Yates, C. J. (2011). A statistical explanation of MaxEnt for ecologists. Diversity and Distributions, 17, 43–57. https://doi.org/10.1111/j.1472-4642.2010.00725.x

Engelbrecht, F., Le Roux, A., Arnold, K., & Malherbe, J. (2019). Detailed projections of future climate change over South Africa. CSIR. https://pta-gis-2-web1.csir.co.za/portal/apps/GBCascade/index.html?appid=b161b2f892194ed5938374fe2192e537

Phillips, S. J., Dudík, M., & Schapire, R. E. (2019). Maxent software for modeling species niches and distributions (Version 3.4.1). https://biodiversityinformatics.amnh.org/open_source/maxent/

Raw, J. L., Godbold, J. A., Van Niekerk, L., & Adams, J. B. (2019). Drivers of mangrove distribution at the high-energy, wave-dominated, southern African range limit. Estuarine, Coastal and Shelf Science, 226, 106296. https://doi.org/10.1016/j.ecss.2019.106296

Van Niekerk, L., Taljaard, S., Ramjukadh, C.-L., Adams, J. B., Lamberth, S. J., Weerts, S. P., Petersen, C., Audouin, M., & Maherry, A. (2017). A multi-sector resource planning platform for South Africa’s estuaries. (Water Research Commission Report K5/2464). Water Research Commission.

Usage Notes

Data files are provided in CSV format.

Funding

Water Research Commission, Award: Project K5/2769

Department of Science and Innovation, South Africa, Award: UID: 84375

Nelson Mandela University