Fully distributed rainfall-runoff modeling using spatial-temporal graph neural network

Recent studies using latest deep learning algorithms such as LSTM (Long Short-Term Memory) have shown great promise in time-series modeling. There are many studies focusing on the watershed-scale rainfall-runoff modeling or streamflow forecasting, often considering a single watershed with limited generalization capabilities. To improve the model performance, several studies explored an integrated approach by decomposing a large watershed into multiple sub-watersheds with semi-distributed structure. In this study, we propose an innovative physics-informed fully-distributed rainfall-runoff model, NRM-Graph (Neural Runoff Model-Graph), using Graph Neural Networks (GNN) to make full use of spatial information including the flow direction and geographic data. Specifically, we applied a time-series model on each grid cell for its runoff production. The output of each grid cell is then aggregated by a GNN as the final runoff at the watershed outlet. The case study shows that our GNN based model successfully represents the spatial information in predictions. NRM-Graph network has shown less over-fitting and a significant improvement on the model performance compared to the baselines with spatial information. Our research further confirms the importance of spatially distributed hydrological information in rainfall-runoff modeling using deep learning, and we encourage researchers to incorporate more domain knowledge in modeling.

Comparison of Rainfall-Runoff Simulation between Support Vector Regression and HEC-HMS for a Rural Watershed in Taiwan

To better understand the effect and constraint of different data lengths on the data-driven model training for the rainfall-runoff simulation, the support vector regression (SVR) approach was applied to the data-driven model as the core algorithm in the present study. Various features selection strategies and different data lengths were employed in the training phase of the model. The validated results of the SVR were compared with the rainfall-runoff simulation derived from a physically based hydrologic model, the Hydrologic Modeling System (HEC-HMS). The HEC-HMS was considered a conventional approach and was also calibrated with a dataset period identical to the SVR. Our results showed that the SVR and HEC-HMS models could be adopted for short and long periods of rainfall-runoff simulation. However, the SVR model estimated the rainfall-runoff relationship reasonably well even if the observational data of one year or one typhoon event was used. In contrast, the HEC-HMS model needed more parameter optimization and inference processes to achieve the same performance level as the SVR model. Overall, the SVR model was superior to the HEC-HMS model in the performance of the rainfall-runoff simulation.

Stream Stage Monitoring with Community Science-Contributed Stage Data

Volunteered Geographic Information, data contributed by community scientists, is an increasingly popular tool to collect scientific data, involve the community in scientific research, and provide information and education about a prominent issue. Johnson City, Tennnessee, USA has a long history of downtown flooding, and recent redevelopment of two land parcels has created new city parks that mitigate flooding through floodwater storage, additional channel capacity, and reduced impervious surfaces. At Founders Park, a project to collect stage data using text messages from community scientists has collected 1479 stage measurements from 597 participants from May 2017 through July 2021. Text messages were parsed to extract the stage and merged with local precipitation data to assess the stream’s response to precipitation. Of 1479 observations, 96.7% were correctly parsed. Only 3% of observations were false positives (parser extracted incorrect stage value) or false negatives (parser unable to extract correct value but usable data were reported). Less than 2% of observations were received between 11 p.m. and 7 a.m., creating an overnight data gap, and fewer than 7% of observations were made during or immediately following precipitation. Regression models for stage using antecedent precipitation explained 21.6% of the variability in stream stage. Increased participation and development of an automated system to record stage data at regular intervals will provide data to validate community observations and develop more robust rainfall–runoff models.

Formulation of Parsimonious Urban Flash Flood Predictive Model with Inferential Statistics

The curve number (CN) rainfall–runoff model is widely adopted. However, it had been reported to repeatedly fail in consistently predicting runoff results worldwide. Unlike the existing antecedent moisture condition concept, this study preserved its parsimonious model structure for calibration according to different ground saturation conditions under guidance from inferential statistics. The existing CN model was not statistically significant without calibration. The calibrated model did not rely on the return period data and included rainfall depths less than 25.4 mm to formulate statistically significant urban runoff predictive models, and it derived CN directly. Contrarily, the linear regression runoff model and the asymptotic fitting method failed to model hydrological conditions when runoff coefficient was greater than 50%. Although the land-use and land cover remained the same throughout this study, the calculated CN value of this urban watershed increased from 93.35 to 96.50 as the watershed became more saturated. On average, a 3.4% increase in CN value would affect runoff by 44% (178,000 m3). This proves that the CN value cannot be selected according to the land-use and land cover of the watershed only. Urban flash flood modelling should be formulated with rainfall–runoff data pairs with a runoff coefficient > 50%.