Oil palm (Elaeis guineensis) cultivation in Central Africa (CA) has become important because of the increased global demand for vegetable oils. The region is highly suitable for the cultivation of oil palm and this increases pressure on forest biodiversity in the region. Accurate maps are therefore needed to understand trends in oil palm expansion for landscape‐level planning, conservation management of endangered species, such as great apes, biodiversity appraisal and supply of ecosystem services. In this study, we demonstrate the utility of a U‐Net Deep Learning Model and product fusion for mapping the extent of oil palm plantations for six countries within CA, including Cameroon, Central African Republic, Democratic Republic of Congo (DRC), Equatorial Guinea, Gabon and Republic of Congo. Sentinel‐1 and Sentinel‐2 data for the year 2021 were classified using a U‐Net model. Overall classification accuracy for the final oil palm layer was 96.4 ± 1.1%. Producer Accuracy (PA) and User Accuracy (UA) for the industrial and smallholder oil palm classes were 91.6 ± 1.7% and 95.0 ± 1.3%, 67.7 ± 2.8% and 70.0 ± 2.8%. Post classification assessment of the transition from tropical moist forest (TMF) cover to oil palm within the six CA countries suggests that over 1000 Square Kilometer (km2) of forest within great ape ranges had so far been converted to oil palm between 2000 and 2021. Results from this study indicate a more extensive cover of smallholder oil palm than previously reported for the region. Our results also indicate that expansion of other agricultural activities may be an important driver of deforestation as nearly 170 000 km2 of forest loss was recorded within the IUCN ranges of the African great apes between 2000 and 2021. Output from this study represents the first oil palm map for the CA, with specific emphasis on the impact of its expansion on great ape ranges. This presents a dependable baseline through which future actions can be formulated in addressing conservation needs for the African Great Apes within the region.

中文翻译：

---

绘制中非油棕种植园及其对森林和类人猿栖息地丧失的影响


由于全球对植物油的需求增加，中非 （CA） 的油棕 （Elaeis guineensis） 种植变得很重要。该地区非常适合种植油棕，这增加了该地区森林生物多样性的压力。因此，需要准确的地图来了解油棕扩张的趋势，用于景观规划、濒危物种（如类人猿）的保护管理、生物多样性评估和生态系统服务的供应。在这项研究中，我们展示了 U-Net 深度学习模型和产品融合在绘制 CA 内六个国家的油棕种植范围方面的效用，包括喀麦隆、中非共和国、刚果民主共和国 （DRC）、赤道几内亚、加蓬和刚果共和国。2021 年的 Sentinel-1 和 Sentinel-2 数据使用 U-Net 模型进行分类。最终油棕层的总体分类准确率为 96.4 ± 1.1%。工业和小农油棕类别的生产者准确率 （PA） 和用户准确率 （UA） 分别为 91.6 ± 1.7% 和 95.0 ± 1.3%、67.7 ± 2.8% 和 70.0 ± 2.8%。对 CA 六个国家从热带潮湿森林 （TMF） 覆盖过渡到油棕的分类后评估表明，迄今为止，在 2000 年至 2021 年期间，大型类人猿分布范围内超过 1000 平方公里 （km2） 的森林已转变为油棕。这项研究的结果表明，该地区的小农油棕覆盖率比以前报告的要广泛。我们的研究结果还表明，其他农业活动的扩大可能是森林砍伐的重要驱动因素，因为 2000 年至 2021 年期间，世界自然保护联盟 （IUCN） 范围内非洲类人猿的森林损失接近 170 000 平方公里。 这项研究的成果代表了 CA 的第一张油棕地图，特别强调了其扩张对类人猿活动范围的影响。这提供了一个可靠的基线，通过该基线可以制定未来的行动，以满足该地区非洲类人猿的保护需求。