Desalinized seawater is a vital freshwater source for regions with coastal water scarcity. Mapping seawater desalination plants enables a spatially detailed water resource assessment. Here, which is the first of its kind, we investigated the potential application of species distribution models (SDMs), which are widely used in ecology, to predict the global spatial distribution of seawater desalination plants. Two regression SDMs, a generalized linear model and a generalized additive model, along with two machine learning SDMs, a random forest model and a generalized boosted regression model, were trained and tested using the cross-validation method at 0.5°. For each SDM, we considered four explanatory variables: aridity, distance to coastline, gross domestic product per capita, and annual domestic and industrial water withdrawal. Our results showed that machine learning SDMs had a relatively strong performance in capturing the historical locations of seawater desalination plants. We then mapped the future distribution of seawater desalination plants under different shared socioeconomic pathways (SSPs) and representative concentration pathways (RCPs). Our predictions showed that the number of predicted locations of seawater desalination plants increased by 31%, 47%, 55%, 57% in 2030, 2050, 2070, and 2090, respectively, relative to 2014. The largest increase occurred under SSP3_RCP7.0, while the lowest increase was found under SSP1_RCP2.6, which is mainly determined by the differences in the volume of annual domestic and industrial water withdrawal. Our study provides an insight into how SDMs can be used to predict the geographic locations of water management facilities.

中文翻译：

---

使用物种分布模型绘制全球当前和未来的海水淡化厂地图

淡化海水是沿海缺水地区的重要淡水资源。绘制海水淡化厂地图可以进行空间详细的水资源评估。在这里，这是同类中的第一个，我们研究了广泛用于生态学的物种分布模型（SDM）在预测海水淡化厂的全球空间分布方面的潜在应用。两个回归 SDM，一个广义线性模型和一个广义相加模型，以及两个机器学习 SDM，一个随机森林模型和一个广义增强回归模型，在 0.5° 下使用交叉验证方法进行了训练和测试。对于每个 SDM，我们考虑了四个解释变量：干旱度、到海岸线的距离、人均国内生产总值以及年度生活和工业取水量。我们的结果表明，机器学习 SDM 在捕获海水淡化厂的历史位置方面具有相对较强的性能。然后，我们绘制了不同共享社会经济路径 (SSP) 和代表性浓度路径 (RCP) 下海水淡化厂的未来分布图。我们的预测显示，2030年、2050年、2070年和2090年海水淡化厂的预测选址数量相对于2014年分别增加了31%、47%、55%和57%。增幅最大的是SSP3_RCP7.0 , 而在 SSP1_RCP2.6 下增幅最低, 这主要是由年生活和工业取水量的差异决定的。我们的研究深入了解了如何使用 SDM 来预测水管理设施的地理位置。