Prediction of the time for perennial outflow from a natural basin to recede from the mean to some low flow value is an important practical and difficult problem in water resource management. This study aims to gain further understanding of this complex problem and to put forward new practical and accurate methods for predicting flow recessions between nominated limits in both gauged and ungauged basins. For a gauged basin, a three parameter recession model is employed to estimate the recession time, from day-to-day, as flow recedes from mean flow using previously measured site recession curves and a library of recession curve shapes generated by the RObust Parameter Estimation algorithm. The model is tested using data from 10 New Zealand basins which are diverse in low flow hydrological behavior and yields a Median Absolute Error (MAE) of 1 day. Another new model is also developed to predict recession time in an ungauged basin using catchment characteristics and information from master recession curves in a suite of 10 reference basins geologically and hydrologically similar to the ungauged basin, as assessed by a Random Forest model. Model performance is robust with a MAE of 1 day and the models advance the use of past flow records to enable more accurate predictions to be obtained. They can be applied elsewhere with confidence although further testing is desirable.

中文翻译：

---

有测量和无测量流域的水流衰退曲线的预测

预测自然流域的常年出水量从平均值下降到某个低流量值的时间是水资源管理中一个重要的实际难题。本研究旨在进一步了解这一复杂问题，并提出新的实用和准确的方法来预测测量和未测量盆地中指定限制之间的流量衰退。对于测量盆地，使用三参数衰退模型来估计每天的衰退时间，因为流量使用先前测量的场地衰退曲线和由 RObust 参数估计生成的衰退曲线形状库从平均流量中衰退算法。该模型使用来自 10 个新西兰流域的数据进行测试，这些流域在低流量水文行为方面各不相同，并产生了 1 天的中值绝对误差 (MAE)。还开发了另一个新模型，使用集水特征和来自一组 10 个在地质和水文上与未测量盆地相似的参考盆地的主衰退曲线的信息来预测未测量盆地的衰退时间，这是由随机森林模型评估的。模型性能稳健，MAE 为 1 天，模型推进使用过去的流量记录，以获得更准确的预测。尽管需要进一步测试，但它们可以放心地应用于其他地方。模型性能稳健，MAE 为 1 天，模型推进使用过去的流量记录，以获得更准确的预测。尽管需要进一步测试，但它们可以放心地应用于其他地方。模型性能稳健，MAE 为 1 天，模型推进使用过去的流量记录，以获得更准确的预测。尽管需要进一步测试，但它们可以放心地应用于其他地方。