Changes in lake status are often interpreted as palaeoclimate indicators due to their dependence on precipitation and evaporation. The Global Lake Status Database (GLSDB) has since long provided a standardised synopsis of qualitative lake status over the last 30,000 ¹⁴C years. Potential sources of uncertainty however are not recorded in the GLSDB. Here we present an updated and improved relational-database framework that incorporates uncertainty in both chronology and the interpretation of palaeoenvironmental data. The database uses peer-reviewed palaeolimnological studies to produce a consensus on qualitative lake-status histories, whose chronologies are revised and standardized through the recalibration of radiocarbon dates and the application of Bayesian age-depth modelling for stratigraphic archives. Quantitative information on absolute water-level elevation is preserved if available from geomorphological sources. We also propose a new probabilistic analytical framework that accounts for these uncertainties to reconstruct synoptic, integrated environmental signals. The process is based on a Monte Carlo algorithm that iteratively samples individual lake-status histories within the limits of their uncertainties to produce many possible scenarios. We then use Recursively-Subtracted Empirical Orthogonal Function analysis to extract dominant patterns of lake-status variability from these scenarios.

As a proof of concept, we apply this framework to 67 sites in eastern and southern Africa whose lake-status histories cover part of the late Pleistocene and/or Holocene. We show that, despite the sometimes large temporal and interpretation uncertainties, and the inclusion of highly discontinuous lake-status time series, identifying the major known millennial-scale climatic phases during the last 20,000 years is possible. Our framework was also able to identify an antiphased response between the lake basins in eastern and interior southern Africa to these changes. We propose that our new database and methodology framework serves as a template for efficient lake-status data synthesis, encourages the incorporation of lake-status data in more palaeoclimate syntheses, and expands the possibilities for the use of such data in the evaluation of climate models.

由于其对降水和蒸发的依赖性，湖泊状况的变化通常被解释为古气候指标。长期以来，全球湖泊状况数据库（GLSDB）过去30,000^年以来一直为湖泊定性状况提供标准化的提要¹⁴C年。但是，不确定性的潜在来源未记录在GLSDB中。在这里，我们提出了一个更新和改进的关系数据库框架，该框架在年代学和古环境数据的解释中都包含了不确定性。该数据库使用经过同行评审的古湖泊学研究对定性的湖泊状态历史达成共识，通过放射性碳数据的重新校准和贝叶斯年龄深度模型在地层档案中的应用，对湖泊的时间顺序进行了修订和标准化。如果可以从地貌来源获得，则可以保留绝对水位高度的定量信息。我们还提出了一个新的概率分析框架，该框架考虑了这些不确定性，以重建天气，综合环境信号。该过程基于蒙特卡洛算法，该算法在不确定性的范围内迭代采样单个湖状历史，以产生许多可能的情况。然后，我们使用递归减去的经验正交函数分析从这些场景中提取湖状状态变异性的主导模式。

作为概念的证明，我们将此框架应用于非洲东部和南部的67个站点，这些站点的湖泊状态历史涵盖了晚更新世和/或全新世的一部分。我们表明，尽管有时存在较大的时间和解释不确定性，并且包括高度不连续的湖泊状态时间序列，但仍可以确定过去20,000年内主要的已知千年尺度气候阶段。我们的框架还能够确定非洲东部和南部内部的湖盆之间对这些变化的反相位响应。我们建议我们将新的数据库和方法框架用作有效的湖泊状态数据综合的模板，并鼓励将湖泊状态数据纳入更多的古气候综合中，