Yuh Yisa Ginath, N'Goran Kouamé Paul, Kross Angela, Heurich Marco, Matthews H Damon, Turner Sarah E
Department of Geography, Planning and Environment, University of Concordia, Montreal, Canada.
World Wide Fund for Nature, Regional Office for Africa/Cameroon Country Program Office, Yaoundé, Republic of Cameroon.
PLoS One. 2024 Dec 2;19(12):e0311816. doi: 10.1371/journal.pone.0311816. eCollection 2024.
The Congo Basin tropical forests are home to many endemic and endangered species, and a global hotspot for forest fragmentation and loss. Yet, little has been done to document the region's rapid deforestation, assess its effects and consequences, or project future forest cover loss to aid in effective planning. Here we applied the Random Forest (RF) supervised classification algorithm in Google Earth Engine (GEE) to map and quantify decadal changes in forest cover and land use (LCLU) in the Congo Basin between 1990 and 2020. We cross-validated our LCLU maps with existing global land cover products, and projected our validated results to 2050 under three climate change scenarios, using the Multiperceptron Artificial Neural Network and Markov chain algorithms of the Idrissi Land Change modeller from TerrSet. We found that, over 5.2% (215,938 km2), 1.2% (50,046 km2), and a 2.1% (86,658 km2) of dense forest cover were lost in the Congo Basin between 1990-2000, 2000-2010, and 2010-2020, totaling approximately 8.5% (352,642 km2) loss estimated between 1990-2020. For the period 2020-2050, we estimated a projected 3.7-4.0% (174,860-204,161 km2) loss in dense forest cover under all three climate change scenarios (i.e., 174,860 km2 loss projected for SSP1-2.6, 199,608 km2 for SSP2-4.5, and 204,161 km2 for SSP5-8.5), suggesting that approximately 12.3-12.6% (527,502 km2-556,803 km2) of dense forest cover could be lost over a 60-year period (1990-2050). Our study represents a novel application of spatial modeling tools and Machine Learning algorithms for assessing long-term deforestation and forest degradation within the Congo Basin, under human population growth and IPCC climate change scenarios. We provide spatial and quantitative results required for supporting long-term deforestation and forest degradation monitoring within Congo Basin countries, especially under the United Nations Framework Convention on Climate Change (UNFCCC) REDD+ (Reduce Emissions from Deforestation and Forest Degradation) program.
刚果盆地的热带森林是许多特有和濒危物种的家园,也是森林破碎化和丧失的全球热点地区。然而,在记录该地区迅速的森林砍伐情况、评估其影响和后果,或预测未来森林覆盖面积的损失以协助进行有效规划方面,所做的工作很少。在这里,我们在谷歌地球引擎(GEE)中应用随机森林(RF)监督分类算法,来绘制和量化1990年至2020年刚果盆地森林覆盖和土地利用(LCLU)的十年变化。我们将我们的LCLU地图与现有的全球土地覆盖产品进行交叉验证,并使用TerrSet的伊德里斯土地变化模型的多层感知器人工神经网络和马尔可夫链算法,在三种气候变化情景下将我们的验证结果预测到2050年。我们发现,在1990 - 2000年、2000 - 2010年和2010 - 2020年期间,刚果盆地分别有超过5.2%(215,938平方公里)、1.2%(50,046平方公里)和2.1%(86,658平方公里)的茂密森林覆盖面积丧失,1990 - 2020年期间总计估计丧失约8.5%(352,642平方公里)。对于2020 - 2050年期间,我们估计在所有三种气候变化情景下,茂密森林覆盖面积预计将丧失3.7 - 4.0%(174,860 - 204,161平方公里)(即SSP1 - 2.6情景下预计丧失174,860平方公里,SSP2 - 4.5情景下为199,608平方公里,SSP5 - 8.5情景下为204,161平方公里),这表明在60年期间(1990 - 2050年),约12.3 - 12.6%(527,502平方公里 - 556,803平方公里)的茂密森林覆盖面积可能会丧失。我们的研究代表了空间建模工具和机器学习算法在评估刚果盆地在人口增长和政府间气候变化专门委员会(IPCC)气候变化情景下的长期森林砍伐和森林退化方面的一种新应用。我们提供了支持刚果盆地国家长期森林砍伐和森林退化监测所需的空间和定量结果,特别是在《联合国气候变化框架公约》(UNFCCC)的减少毁林和森林退化所致排放量(REDD +)计划下。