Mapping the Mangrove Forest Restoration Potential and Conservation Gaps in China Based on Random Forest Model

Abstract

Background and objective: The area of mangroves is gradually decreasing globally, and mangroves are already one of the most threatened ecosystems. Despite net growth in the mangrove areas in China, the restoration potential of mangroves is still insufficient. This study proposed the Random forest model as an excellent data mining method to map the restoration potential based on the predicted probability of mangrove habitat suitability.Methods: We demonstrated the vital environmental variables influencing habitat suitability. The de-cisive advantages of RFM were parsimonious (variables selection), cost-effective (us-ing existing open-source data), accurate (training AUC was 0.89, testing AUC was 0.91), highly efficient (fast-training speed); and its results had high explanatory power. Here, we first mapped the conservation gaps using the RFM.Results: The results showed that temperature was the most important environmental factor influencing the habitat suit-ability of mangroves. The northern limit of suitable areas was around 24°44' N. The theoretical suitable habitat area for mangrove was 196,566.6 ha (the highly suitable area was 32,551.4 ha, the medium suitable area was 164,015.2 ha). The potential area for mangrove restoration was 176,264 ha (Guangdong with 104215.4 ha, Guangxi with 65957.5 ha).Conclusion: We proposed 24 sites with conservation gaps for mangrove forests restoration and nine potential sites as examples for the further restoration plan. We took one example site with high restoration potential for further explanation: how the key environmental factors influence the habitat suitability and how to use the infor-mation to guide the restoration strategies. RFM can be used as a data mining algo-rithm for the utmost use of the presence-only ecological data, objectively evaluating the suitability of species distribution, and providing scientifically technical data for species restoration planning.