This study assesses the differences in modelled total runoff and modelled runoff delivered to people between different rainfall and population datasets in the Ankeniheny Zhamena Corridor (CAZ) of Eastern Madagascar. Runoff is estimated using the WaterWorld hydrological model driven by six rainfall datasets, and population is derived from five population datasets. Model results for runoff under different rainfall datasets lead to variability in runoff (coefficient of variation) up to 99% for single months and 60% in the dry season. These differences are much larger than differences in estimated runoff between baseline and complete deforestation scenario for each rainfall dataset. Population estimates for the CAZ range from 1.2 to 2 million between the population datasets. Differences in runoff under different rainfall datasets lead to an average of 356,000 people estimated to receive 90% more runoff and nearly 750,000 people estimated to receive 50% more or less runoff relative to a baseline rainfall dataset. Therefore, the choice of rainfall data in hydrological ecosystem services modelling has a large influence on estimates of ecosystem service flows highlighting the need for modellers to justify their data choices and report on uncertainties in results, particularly in light of potential policy decisions based on modelled outcomes.

这项研究评估了东部马达加斯加的Ankeniheny Zhamena走廊（CAZ）的不同降雨和人口数据集之间的模拟总径流和输送给人类的径流之间的差异。径流是使用由六个降雨数据集驱动的WaterWorld水文模型估算的，人口是从五个人口数据集得出的。不同降雨数据集下径流的模型结果导致径流变异性（变异系数）单个月高达99％，而旱季高达60％。这些差异远大于每个降雨数据集的基准线和完全毁林情景之间的估计径流量差异。在人口数据集之间，CAZ的人口估计数介于1.2到200万之间。在不同降雨数据集下径流的差异导致相对于基准降雨量数据集，平均有356,000人估计将增加90％的径流，而将近750,000人估计将增加或减少50％的径流。因此，水文生态系统服务建模中降雨数据的选择对生态系统服务流量的估计有很大影响，突出表明建模者需要证明其数据选择的合理性并报告结果的不确定性，尤其是考虑到基于建模结果的潜在政策决策时。