Spatiotemporal convolutional long short-term memory for regional streamflow predictions.

Rainfall-runoff (RR) modelling is a challenging task in hydrology, especially at the regional scale. This work presents an approach to simultaneously predict daily streamflow in 86 catchments across the US using a sequential CNN-LSTM deep learning architecture. The model effectively incorporates both spatial and temporal information, leveraging the CNN to encode spatial patterns and the LSTM to learn their temporal relations. For training, a year-long spatially distributed input with precipitation, maximum temperature, and minimum temperature for each day was used to predict one-day streamflow. The trained CNN-LSTM model was further fine-tuned for three local sub-clusters of the 86 stations, assessing the significance of fine-tuning in model performance. The CNN-LSTM model, post fine-tuning, exhibited strong predictive capabilities with a median Nash-Sutcliffe efficiency (NSE) of 0.62 over the test period. Remarkably, 65% of the 86 stations achieved NSE values greater than 0.6. The performance of the model was also compared to different deep learning models trained using a similar setup (CNN, LSTM, ANN). An LSTM model was also developed and trained individually to predict for each of the stations using local data. The CNN-LSTM model outperformed all the models which was trained regionally, and achieved a comparable performance to the local LSTM model. Fine-tuning improved the performance of all models during the test period. The results highlight the potential of the CNN-LSTM approach for regional RR modelling by effectively capturing complex spatiotemporal patterns inherent in the RR process.

A reduced latency regional gap-filling method for SMAP using random forest regression.

The soil moisture active/passive (SMAP) mission represents a significant advance in measuring soil moisture from satellites. However, its large spatial-temporal data gaps limit the use of its values in near-real-time (NRT) applications. Considering this, the study uses NRT operational metadata (precipitation and skin temperature), together with some surface parameterization information, to feed into a random forest model to retrieve the missing values of the SMAP L3 soil moisture product. This practice was tested in filling the missing points for both SMAP descending (6:00 AM) and ascending orbits (6:00 PM) in a crop-dominated area from 2015 to 2019. The trained models with optimized hyper-parameters show the goodness of fit (R2 ≥ 0.86), and their resulting gap-filled estimates were compared against a range of competing products with in situ and triple collocation validation. This gap-filling scheme driven by low-latency data sources is first attempted to enhance NRT spatiotemporal support for SMAP L3 soil moisture.

Hydroinformatic tools and their potential in the search for missing persons in rivers.

This research proposes a methodological approach to search floating bodies or floating body parts in rivers using hydroinformatic tools. These tools may allow possible search locations and potential release sites to be determined. The approach draws from hydrometric field information, two-dimensional (2D) hydrodynamic modeling, particle tracking (PT) models, and large-scale particle image velocimetry (LSPIV). This methodology was applied to a case study of the La Miel river in Colombia, where existing reports of deceased persons are available. A series of hydraulic accidents have been defined to represent the hydrodynamic and transport processes that occur in such situations. The results indicate that potential search locations, namely places where human floating bodies or floating body parts may be found, are principally on the left and right shores of recirculation systems.

Knowledge-based modularization and global optimization of artificial neural network models in hydrological forecasting.

Natural phenomena are multistationary and are composed of a number of interacting processes, so one single model handling all processes often suffers from inaccuracies. A solution is to partition data in relation to such processes using the available domain knowledge or expert judgment, to train separate models for each of the processes, and to merge them in a modular model (committee). In this paper a problem of water flow forecast in watershed hydrology is considered where the flow process can be presented as consisting of two subprocesses -- base flow and excess flow, so that these two processes can be separated. Several approaches to data separation techniques are studied. Two case studies with different forecast horizons are considered. Parameters of the algorithms responsible for data partitioning are optimized using genetic algorithms and global pattern search. It was found that modularization of ANN models using domain knowledge makes models more accurate, if compared with a global model trained on the whole data set, especially when forecast horizon (and hence the complexity of the modelled processes) is increased.