The eutrophication of lakes and the subsequent algal blooms have become significant environmental issues of global concern in recent years. With ongoing global warming and intensifying human activities, water quality trends in lakes worldwide varied significantly, and the trend of algal blooms in the next few decades is unclear. However, there is a lack of comprehensive quantitative research on the future projection of lake algal blooms globally due to the scarcity of long-term algal blooms observational data and the complex nonlinear relationships between algal blooms and their driving factors. We aimed to develop a global projection model to evaluate the future trend in algal bloom occurrences in large lakes under various socio-economic development scenarios. We focused our research on 161 natural lakes worldwide, each exceeding 500 km2. The results indicated that the Random Forest model performed best (Overall Accuracy: 0.9697, Kappa: 0.8721) among various machine learning models which were applied in this study. The predicted results showed that, by the end of this century, the number of lakes experiencing algal blooms and the intensity of these blooms will worsen under higher forcing scenarios (SSP370 and SSP585) (p < 0.05). In different regions, lakes with increasing algal blooms are mainly distributed in Africa, Asia, and North America, while lakes with decreasing occurrence are primarily found in Europe. Additionally, underdeveloped regions, such as Africa, exhibit greater sensitivity to different SSP scenarios due to high variability in population and economic growth. This study revealed the spatiotemporal distribution of algal blooms in global lakes from 2020 to 2100 and suggested that the intensifying algal blooms due to global warming and human activities may offset the effort of controlling the water quality.

中文翻译：

---

全球湖泊藻华对未来气候和土地利用变化的预计响应：机器学习方法


近年来，湖泊的富营养化和随之而来的藻类大量繁殖已成为全球关注的重大环境问题。随着全球持续变暖和人类活动的加剧，全球湖泊的水质趋势差异很大，未来几十年藻类大量繁殖的趋势尚不清楚。然而，由于长期藻华观测数据的稀缺以及藻华与其驱动因素之间复杂的非线性关系，目前缺乏对全球湖藻华未来预测的全面定量研究。我们旨在开发一个全球预测模型，以评估在各种社会经济发展情景下大型湖泊藻华发生的未来趋势。我们将研究重点放在全球 161 个天然湖泊上，每个湖泊的面积超过 500 km2。结果表明，随机森林模型在本研究应用的各种机器学习模型中表现最佳 （总体准确率： 0.9697， Kappa： 0.8721）。预测结果表明，到本世纪末，在更高的强迫情景下（SSP370 和 SSP585），经历藻华的湖泊数量和这些藻华的强度将恶化（p < 0.05）。在不同地区，藻华增加的湖泊主要分布在非洲、亚洲和北美，而藻华减少的湖泊主要分布在欧洲。此外，由于人口和经济增长的高度可变性，非洲等欠发达地区对不同的 SSP 情景表现出更高的敏感性。 这项研究揭示了 2020 年至 2100 年全球湖泊藻华的时空分布，并表明由于全球变暖和人类活动导致藻华加剧可能会抵消控制水质的努力。