Evaluating Mangrove Conservation Effectiveness

Protected areas have been considered an important tool to maintain ecological integrity. However, protected areas are increasingly under threat and their effects on forest conservation are in doubt. Mangroves are highly productive ecosystems which can provide various ecosystem services and goods. Cambodia contains sizable mangroves, however, it also experiences significant mangrove loss and fragmentation. An improved understanding of how effective protected areas can preserve mangroves is required. The study aimed to examine both temporal and spatial mangrove cover change by comparing mangrove cover and fragmentation pattern change inside and outside Peam Krasop Wildlife Sanctuary (PKWS) in southwest Cambodia between 2014 and 2020 to investigate its conservation effectiveness. 

Mangroves are a group of salt-tolerant woody plants that are the foundational species forming intertidal ecosystems in tropical and subtropical regions. Covering between 150,000 and 188,000 km2 in estuarine areas around the world, mangroves are highly productive ecosystems providing a wide range of ecosystem services and goods. Despite the importance of mangrove ecosystems, the coverage continues to decline due to human activities and climate change. Southeast Asia contains the majority of the world’s mangrove forests, however, it also experiences the most mangrove loss with annual loss rates nearly double the global average.

My study area includes PKWS and the 5 km buffer around it as the non-protected area. Located in the coastal strip of Koh Kong province, southwestern Cambodia, PKWS is one of the 23 protected areas in Cambodia established by Royal Decree. PKWS maintains sizable mangrove forests at the western part and evergreen forest to the east. At least 64 mangrove species can be found inside PKWS. However, even after PKWS had been declared a protected area in 1993, it experienced severe forest fragmentation during the late 1990s due to infrastructure development, illegal logging and land conversion for agriculture and aquaculture.

My main reference data was the land cover map made by the  SERVIR-Mekong project  which can help me decide land cover classes and draw training polygons. After collecting data, I conducted unsupervised classification which can also assist in determining the number of land cover classes. A Maximum Likelihood algorithm was used to produce land cover maps using Landsat 8 OLI/TIRS Level-2 Tier 1 images in ENVI. The image classification was followed by a map accuracy assessment using a confusion matrix. Land cover transition matrix between 2014 and 2020 for both PKWS and the surrounding area was obtained to describe the quantity of conversions from one particular land cover to other land cover classes. Additionally, a mangrove cover change thematic map was made to detect the location of mangrove loss, gain, and persistence. Landscape metrics were calculated to detect forest fragmentation change using the software FRAGSTATS and landscapemetrics package in RStudio. 

• Accuracy assessments

The overall accuracies of the 2014 and 2020 classification results were at 98.15% and 97.56% respectively. Exposed soil and build up and non-forest vegetation achieved the worst results with the lowest accuracies for the 2014 classification map (87.02% and 80.17% respectively). Additionally, non-forest vegetation and open-canopy mangrove had the lowest accuracies for the 2020 classification map (78.78% and 75.77% respectively). As a result, exposed soil and build up, non-forest vegetation, and open-canopy mangrove had relatively higher proportions of mismatched pixels

• Land cover classification and change detection inside and outside PKWS

A total of six classes (i.e., open-canopy mangrove, closed-canopy mangrove, terrestrial evergreen forest, non-forest vegetation, exposed soil and build-up, and water) were produced for 2014 and 2020 classification maps. PKWS maintained sizable mangroves at the western part and terrestrial evergreen forests to the east. However, the surrounding area had only a few mangrove patches whilst the dominant forest cover type was terrestrial evergreen forest. Additionally, the surrounding non-protected landscape had a larger area of exposed soil and build up than PKWS in both years. 

About 91% of closed-canopy mangrove and 63% of open-canopy mangrove remained unchanged inside PKWS.

The net mangrove loss including both types of mangroves was 74.16 ha.

The net mangrove loss outside PKWS was 72.54ha which was smaller. However, higher proportion of mangrove cover change was observed.

Only 67% of closed-canopy mangrove and 56% of open-canopy mangrove remained stable in the surrounding non-protected area.Besides, the surrounding area had higher proportion of land conversion to exposed soil and build up. 

Net mangrove deforestation may be compensated by reforestation and masked the true level of mangrove loss. Therefore, it was necessary to calculate gross deforestation. 

There was more mangrove gross loss (908.82 ha) than gain (834.39 ha) inside PKWS. The surrounding non-protected area had much less mangrove persistence (1432.17 ha) than PKWS due to the low initial mangrove coverage. However, mangrove loss and gain outside PKWS (812.79 ha and 740.61 ha respectively) were similar to those measured inside PKWS.

• Vegetation fragmentation patterns inside and outside PKWS 

PKWS had higher proportions of closed-canopy and open-canopy mangroves and a slightly higher proportion of non-forest vegetation compared to the surrounding non-protected area. 

Higher edge density relates to more severe fragmentation. There was a decrease in edge density for mangrove patches in both areas indicating that the degree of fragmentation decreased. However, the edge density inside PKWS was larger than that of the surrounding area. Further analysis was needed to evaluate the fragmentation pattern. 

Considering both mangrove types, the number of mangrove patches decreased in both PKWS and the surrounding area indicating that the degree of fragmentation decreased. Although the number of patches was smaller in the surrounding area, we need to take into account the mean patch size to determine which area had more fragmentation. 

PKWS had larger mean patch size for both mangrove types, but the difference was not statistically significant. However, the core area proportion of closed-canopy mangrove inside PKWS was significantly larger in both dates indicating that PKWS had less closed-canopy mangrove fragmentation. 

Similar to other studies conducted in tropical and subtropical regions, PKWS did protect mangrove quantity and quality by decreasing fragmentation and limiting deforestation rate, however, its function was ineffective as compared to the surrounding non-protected area because the surrounding area had slightly less fragmentation. 

Missing field data may impact the accuracy of land cover classification.

Incorporating remote sensing and GIS techniques with field surveys as well as local demographic and livelihood data can be more effective for evaluating conservation efforts and taking further actions.

Ideally, the time interval should include both before and after the protected area establishment.