Deep learning approaches for improving prediction of daily stream temperature in data‐scarce, unmonitored, and dammed basins
Deep learning approaches for improving prediction of daily stream temperature in data‐scarce, unmonitored, and dammed basins
复制标题
DOI：
10.1002/hyp.14400
复制
发表时间：
2021-09
影响因子：
3.2
通讯作者：
F. Rahmani;Chaopeng Shen;S. Oliver;K. Lawson;A. Appling
中科院分区：
文献类型：
--
作者：
F. Rahmani;Chaopeng Shen;S. Oliver;K. Lawson;A. Appling
以流域为中心的长短期记忆（LSTM）网络模型最近被证明是一种非常强大的流温（Ts）时间预测工具（在一个时期训练，在同一地点的另一个时期预测）。然而，空间外推是对Ts建模的一个众所周知的挑战，并且不确定基于LSTM的每日Ts模型在未监测或筑坝盆地中的表现如何。在这里，我们编制了一个新的基准数据集，包括美国本土不同数据可用性组（DAG，意思是每日采样频率）的400多个流域，有和没有大型水坝，并研究了如何在有或没有温度监测的流域中组装合适的训练数据集进行预测。对于未监测流域（PUB）的预测，LSTM产生的均方根误差（RMSE）为1.129°C，R2为0.983。虽然这些指标从LSTM的时间预测性能下降，但它们远远超过了传统模型的PUB值，并且与传统模型在校准站点上的时间预测具有竞争力。即使对于具有主要水库的未监测流域，我们也获得了1.202°C的中位数RMSE和0.984的R2。对于时间预测，最合适的训练集是流域可以分组的匹配DAG（例如，60% DAG最适合数据可用性为61%的流域）。然而，对于PUB，包括所有流域数据的训练数据集始终是首选。输入选择集合适度减轻了属性过拟合。我们的结果表明，有影响力的潜在过程没有充分描述的输入（例如，地质、湿地覆盖），但时间波动仍然可以很好地预测，LSTM似乎是一个高度准确的Ts建模工具，即使是空间外推。
Basin‐centric long short‐term memory (LSTM) network models have recently been shown to be an exceptionally powerful tool for stream temperature (Ts) temporal prediction (training in one period and predicting in another period at the same sites). However, spatial extrapolation is a well‐known challenge to modelling Ts and it is uncertain how an LSTM‐based daily Ts model will perform in unmonitored or dammed basins. Here we compiled a new benchmark dataset consisting of >400 basins across the contiguous United States in different data availability groups (DAG, meaning the daily sampling frequency) with and without major dams, and studied how to assemble suitable training datasets for predictions in basins with or without temperature monitoring. For prediction in unmonitored basins (PUB), LSTM produced a root‐mean‐square error (RMSE) of 1.129°C and an R2 of 0.983. While these metrics declined from LSTM's temporal prediction performance, they far surpassed traditional models' PUB values, and were competitive with traditional models' temporal prediction on calibrated sites. Even for unmonitored basins with major reservoirs, we obtained a median RMSE of 1.202°C and an R2 of 0.984. For temporal prediction, the most suitable training set was the matching DAG that the basin could be grouped into (for example, the 60% DAG was most suitable for a basin with 61% data availability). However, for PUB, a training dataset including all basins with data was consistently preferred. An input‐selection ensemble moderately mitigated attribute overfitting. Our results indicate there are influential latent processes not sufficiently described by the inputs (e.g., geology, wetland covers), but temporal fluctuations can still be predicted well, and LSTM appears to be a highly accurate Ts modelling tool even for spatial extrapolation.