Identification of Dominant Runoff Controls Using Hydrologically Informed Machine Learning Approach

Relative dominance of the runoff controls, such as topography, geology, soil types, land use, and climate, may differ from catchment to catchment due to spatial and temporal heterogeneity of landscape properties and climate variables. Understanding dominant runoff controls is an essential task in developing unified hydrological theories at the catchment scale. Semi-distributed rainfall-runoff models are often used to identify dominant runoff controls for a catchment of interest. In most such applications, the model selection is based on either expert's judgement or experimental and fieldwork insights. Model selection is the most important step in any hydrological modelling exercise as the findings are largely influenced by the selected model. Hence, a subjective model selection without sufficient expert's knowledge or experimental insights may result in biased findings, especially for comparative studies like identification of dominant runoff controls. In this study, we use a physics informed machine learning toolbox based on genetic programming Machine Induction Knowledge Augmented - System Hydrologique Asiatique (MIKA-SHA) to identify the relative dominance of runoff controls. We find the quantitative and automated approach based on MIKA-SHA to be highly appropriate for the intended task. MIKA-SHA does not require explicit user selections and relies on data and fundamental hydrological processes. The approach is tested using the Rappahannock River basin at Remington, Virginia, United States. Two rainfall-runoff models are learnt to represent the runoff dynamics of the catchment using topography-based and soil-type-based hydrologic response units independently. Based on prediction capabilities, in this case, the topography is identified as the dominant runoff driver.

AI4Water v1.0: An open source python package for modeling hydrological time series using data-driven methods

Machine learning has shown great promise for simulating hydrological phenomena. However, the development of machine learning-based hydrological models requires advanced skills from diverse fields, such as programming and hydrological modeling. Additionally, data pre-processing and post-processing when training and testing machine learning models is a time-intensive process. In this study, we developed a python-based framework that simplifies the process of building and training machine learning-based hydrological models and automates the process of pre-processing of hydrological data and post-processing of model results. Pre-processing utilities assist in incorporating domain knowledge of hydrology in the machine learning model, such as the distribution of weather data into hydrologic response units (HRUs) based on different HRU discretization definitions. The post-processing utilities help in interpreting the model’s results from a hydrological point of view. This framework will help increase the application of machine learning-based modeling approaches in hydrological sciences.

Impact of Resolution and Source of Digital Elevation Model on Topological Attributes and Simulated Runoff by SWAT Model

The Digital Elevation Model (DEM) of a watershed is one of the most important inputs in most hydrological analyses and plays a key role in the accurate prediction of various hydrological processes. Comprehensive knowledge of the impact of different DEM sources on the performance of a model is essential before utilizing the model. In this study, we evaluated the influence of TOPO1:25000, ASTER, and SRTM DEMs, as input, on the performance of the Soil and Water Assessment Tool (SWAT) model for the prediction of surface runoff. We also investigated the effect of the resolution of the studied DEM sources on the accuracy of the SWAT model in the estimation of runoff. The second objective of this study was to identify the most influential and the least impactful input parameters on the performance of the SWAT model. We studied the Zarrineh River watershed in Iran as a case study to compare the effect of the aforementioned DEM types and DEM resolution on the output of the SWAT model. The outcomes of the study demonstrated that influential parameters on predicted runoff as well as a few watershed parameters, such as reach lengths, reach slopes, number of sub-basins, and the number of hydrologic response units (HRU), differs noticeably when the DEM source and resolution changes. It was also observed that simulated results over-predict the runoff during low precipitation periods and under-predict the runoff during high precipitation months, and the accuracy of the simulated results decreases by reducing the DEM resolution. The results showed that the SWAT model had the best performance when the TOPO1:25000 DEM was used as the input source. Low-resolution DEMs are available to a wider range of researchers. The outcomes of the current study can be employed to estimate the impact of low-resolution input data on the simulated result as well as substantially reduce the computation time by decreasing the input DEMresolution with only a minor reduction of accuracy.

SIMULATION OF SEDIMENT YIELD AND SUPPLY ON WATER FLOW IN DIFFERENT SUBBASINS OF TERENGGANU WATERSHED FROM 1973-2017

The catchment area of Terengganu has to be flooding during the monsoon season. The reason is climate change that increases water flow in most of the rivers. The analysis using ArSWAT2012 has simulated the whole watershed and the result as proven to have about 25 different sub-basins. Each sub-basin has its peculiar characteristics of Hydrologic Response Units (HRUs). Base on the morphological classification, the river has accumulated a lot of sediments. The sediment yield and concentration has been analyzed from 1973- 2017 through simulation. The study compared the simulations and found out the slide differences in the sediment loads that come in and the sediment that goes out. The sediment concentration also varies with the temporal morphological changes of the Terengganu watershed especially the river mouth.

Revisiting SWAT as a Saturation-Excess Runoff Model

The Soil Water Assessment Tool (SWAT) is employed throughout the world to simulate watershed processes. A limitation of this model is that locations of saturation excess overland flow in hilly and mountainous regions with an impermeable layer at shallow depth cannot be simulated realistically. The objective of this research is to overcome this limitation with minor changes in the original SWAT code. The new approach is called SWAT-with-impervious-layers (SWAT-wil). Adaptations consisted of redefining the hillslope length, restricting downward percolation from the root zone, and redefining hydrologic response units (HRUs) such that they are associated with the landscape position. Finally, input parameters were chosen such that overland flow from variable saturated areas (VSAs) corresponds to the variable source interpretation of the Soil Conservation Service (SCS) curve number runoff equation. We tested the model for the Town Brook watershed in the Catskill Mountains. The results showed that the discharge calculated with SWAT-wil agreed with observed outflow and results simulated with the original SWAT and SWAT-hillslope (SWAT-HS) models that had a surface aquifer that transferred water between groups of HRUs. The locations of the periodically saturated runoff areas were predicted by SWAT-wil at the right locations. Current users can utilize the SWAT-wil approach for catchments where VSA hydrology predominates.

Analysis of Hydrologic Response Units and Impact of Flooding in Kuala Terengganu Sub-basins River Catchment in Malaysia

One of the issue of environmental disasters occurring in a wet tropical environment is flood influenced by the climatic factor of rainfall with high intensity. Flood is the most frequent catastrophe in Kuala Terengganu, Malaysia. The flood occurs during the monsoon season inundating riverbank and displacing the inhabitant rendering them homeless. The application of the Soil Water Assessment Tool (SWAT) is employed to identify the Hydrologic Response Units (HRUs). The flood vulnerability simulation in the Terengganu sub-basins river catchment was done using the most affected sub-basins. In this study, the impacts of five out of the 25 sub-basins have been affected by high flooding risk. The sub-basin with the highest impact of Hydrologic Response Unit is the sub-basin Three and the lowest is found in sub-basin Five.

GSFLOW-GRASS v1.0.0: GIS-enabled hydrologic modeling of coupled groundwater–surface-water systems

Water flow through catchments sustains ecosystems and human activity, shapes landscapes, and links climate to the outermost layers of the solid Earth. The profound importance of water moving between the atmosphere and aquifers has led to efforts to develop and maintain coupled models of surface water and groundwater. However, developing inputs to these models is usually time-consuming and requires extensive knowledge of software engineering, often prohibiting their use by many researchers and water managers, and thus reducing these models' potential to promote science-driven decision-making in an era of global change and increasing water-resource stress. In response to this need, we have developed GSFLOW-GRASS, a straightforward set of open-source tools that develops inputs for and runs GSFLOW, the U.S. Geological Survey's coupled groundwater-surface-water flow model. As inputs, GSFLOW-GRASS requires at a minimum a digital elevation model, a precipitation and temperature record, and estimates of channel parameters and hydraulic conductivity. GSFLOW-GRASS is written in Python as a set of (1) GRASS GIS extensions, (2) input-file-builder scripts, and (3) visualization scripts. We developed a set of custom GRASS GIS commands that generate "hydrologic response units" for surface water, discretized topologically as sub-basins of the tributary network; build the MODFLOW grid; and add necessary attributes to each of these geospatial units. These GIS outputs are interpreted by a second set of Python scripts, which link them to hydrologic variables, build inputs to GSFLOW, and run GSFLOW. Lastly, GSFLOW output files are used to produce figures and time-lapse movies of simulation results using a third set of post-processing Python scripts. We demonstrate the broad applicability of these tools to diverse settings through examples based on: the high Peruvian Andes, the Channel Islands of California, and the formerly-glaciated Upper Mississippi valley in Minnesota.

Parameter regionalization of a monthly water balance model for the conterminous United States

A parameter regionalization scheme to transfer parameter values from gaged to ungaged areas for a monthly water balance model (MWBM) was developed and tested for the conterminous United States (CONUS). The Fourier Amplitude Sensitivity Test, a global-sensitivity algorithm, was implemented on a MWBM to generate parameter sensitivities on a set of 109 951 hydrologic response units (HRUs) across the CONUS. The HRUs were grouped into 110 calibration regions based on similar parameter sensitivities. Subsequently, measured runoff from 1575 streamgages within the calibration regions were used to calibrate the MWBM parameters to produce parameter sets for each calibration region. Measured and simulated runoff at the 1575 streamgages showed good correspondence for the majority of the CONUS, with a median computed Nash–Sutcliffe efficiency coefficient of 0.76 over all streamgages. These methods maximize the use of available runoff information, resulting in a calibrated CONUS-wide application of the MWBM suitable for providing estimates of water availability at the HRU resolution for both gaged and ungaged areas of the CONUS.