Effectiveness of protected areas in preventing forest loss in a tropical mountain region

Abstract

As forest loss is accelerating in tropical mountains globally, protected areas (PAs) are seen as bastions to protect sensitive ecosystems, preserve biodiversity, and safeguard headwater catchments from degradation. However, the effectiveness of PAs in preventing forest conversion has rarely been determined. Complicating the issue is that many PAs are inhabited to some extent by long-standing residents, causing illegal logging that is commonly reported. To assess the effectiveness of PAs in preserving forests, as well as investigate the drivers of forest loss in/near PAs, we compare forest loss rates inside and outside PAs before and after their establishment in a tropical mountain region (northern Thailand, the epicenter of mainland Southeast Asia, including 84 PAs). Over the 17-year period from 2000 to 2016, we found that the percentage of forest loss was lower within the PAs than outside (1.69% versus 4.94%). Mean annual forest loss in the PAs was 20% of that in unprotected area. Total forest loss inside PAs included 888.12 km2 (1.93%) in national parks, 325.18 km2 (1.34%) in wildlife sanctuaries/conservation areas and 16.37 km2 (0.65%) in no hunting areas. Forest loss also tended to be highest along boundaries within a 300-m buffer both inside and outside the PAs. Using gradient boosting decision trees, we determined that accessibility variables (elevation, and distance to road) and population were key drivers associated with forest loss in PAs. Further, we found a two-year lagged correlation between forest loss in PAs and international maize price (R2 = 0.73, p < 0.001), indicating the sensitivity of forest loss in some locations to commodity agriculture trends. Finally, we acknowledge the difficulty of managing forest loss in PAs because of the large populations of people living within the boundaries who rely on forest products to support their livelihoods, as well as difficulties in enforcing illegal logging laws.

Keywords

• Conservation
• Deforestation
• Forest management
• Remote sensing
• Machine learning model