ParFlow is a numerical model that simulates the hydrologic cycle from the bedrock to the top of the plant canopy. It integrates three-dimensional groundwater flow with overland flow and plant processes using physically-based equations to rigorously simulate fluxes of water and energy in complex real-world systems. ParFlow is a computationally advanced model that can run on laptops and supercomputers and has been used in hundreds of studies evaluating hydrologic processes from the hillslope to the continental scale. Our code is open source and we promote a community of active users and developers interested in advancing computational hydrology and improving hydrologic understanding. Details about the model, example applications and links for downloading and getting started with the code are provided below.
ParFlow is used extensively for water cycle research in idealized and real domains as part of process studies, forecasting analysis, data assimilation frameworks, hind-casting tools and climate change projections. The model has been extensively benchmarked and has more than 130 publications describing its development and application to diverse systems around the world. ParFlow applications have been built for the continental US (CONUS) and Continental Europe in addition to more than a dozen watersheds around the world including the Big Thompson, CO; Klamath, OR; Little Washita, OK; San Joaquin, CA; Sante Fe, FL; Chesapeake, MD; Rur as well as several headwater catchments, Germany.
ParFlow is a parallel, integrated hydrology model that simulates spatially distributed surface and subsurface flow, as well as land surface processes including evapotranspiration and snow. It solves saturated and variably saturated flow in three dimensions using either an orthogonal or terrain-following, semi-structured mesh that enables fine vertical resolution near the land surface and deep (~1 km) confined and unconfined aquifers. ParFlow models dynamic surface and subsurface flow solving the simplified shallow water equations implicitly coupled to Richards’ equation; this allows for dynamic two-way groundwater surface water interactions and intermittency in streamflow. The model uses robust linear and nonlinear solution techniques and exhibits efficient parallel scaling to large processor counts, more than 100K cores, enabling very large extent simulations with fine spatial resolution. ParFlow has been coupled to various land surface and atmospheric models such as CLM, WRF, and TerrSysMP.
ParFlow has been widely used to simulate flow and transport systems worldwide; here we highlight some recent examples linked to their corresponding publications. (Clicking on the figures opens a link to the publication's website.)
Soil moisture monitoring and forecasting for agriculture over central Europe In the context of more frequent and consecutive droughts during the vegetation period over the last years (e.g. 2018, 2019, and 2020) in central Europe, we use ParFlow/CLM at ~600m resolution to monitor and forecast the soil water budget over Germany and the neighboring regions. The deterministic medium-range forecast over 10 days is extended by a 50 member ensemble to account for the uncertainty of the forecast, and especially precipitation, and its impact on the soil water budget. In addition, subseasonal probabilistic ensemble forecasts (3 months lead time) are calculated a few times per year to help farmers and their advisors to manage the water resources over a longer time span. Several parameters that are relevant for the stakeholders from the agricultural sector are derived from the standard ParFlow outputs, providing information on various aspects of the soil water budget for different root depths, e.g. plant available water, seepage water, subsurface water storage change, groundwater recharge, water table depth, and saturation.
Transient, integrated simulation of groundwater and surface water over the Continental US A representation of pre-development groundwater, surface water, and surface energy processes, which are driven by hourly forcing by NLDAS-II from the 1985 water year. At 1km lateral resolution, with 12TB of model output and 3TB of input, this is the first large-scale, high resolution simulation of its kind, capable of resolving complex interactions between climate, water and topography.
Groundwater-surface water interactions in the San Joaquin River Basin As one of the most productive agricultural regions of the United States and a major water resource for a growing population, the San Joaquin River basin in the Central Valley in California, is a case study in the water-food-energy nexus and sensitivity to a changing climate. Here, we seek to better understand the physical hydrology of the basin by simulating groundwater-surface water dynamics with ParFlow-CLM. Results suggest that mountain block hydraulic conductivity could account for 7-23% of total recharge in the Central Valley, an important finding for water resource management in California.
Groundwater-land surface-atmosphere feedbacks during the European 2003 heat wave By coupling ParFlow to land and meteorological models, the fully coupled water cycle from groundwater to atmosphere can be simulated, a novel exploration in that atmospheric models rarely incorporate a dynamic water table and three dimensional subsurface flow. This study couples ParFlow to the meteorological model over Europe during the 2003 heat wave in order to investigate the effects of various lower boundary conditions and configurations on land-atmosphere moisture exchange and thermal energy.
Continental water residence times Ever wonder how long water spends in the subsurface? Residence time and groundwater age are vital to ecosystem development and human consumption, but measuring residence time distributions is difficult at large scales. ParFlow may be used to estimate water residence time when used in conjunction with a Lagrangian particle tracking approach. A simulation of groundwater age across the continental United States allows unique insight into the relationship between geography, climate, and water residence time in major basins.
The effects of insect-induced tree mortality on water and energy in mountain headwaters The mountain pine beetle (MPB) has decimated the high elevation lodgepole and ponderosa pine forests of Western United States and Canada over the past two decades. ParFlow-CLM was used to diagnose feedbacks that land disturbance at this scale has on water and energy fluxes in the Big Thompson watershed in Colorado. Results show that insect-induced reductions in canopy interception, transpiration, and snow pack are largely mitigated by heightened ground evaporation and ablation, adding to the growing number of studies citing damping of MPB hydrologic signal at large scales.
Scale dependent parameterization in integrated hydrologic modeling The highly instrumented Wüstebach catchment in Germany allows the unique opportunity to explore the application of the information entropy concept in subsurface parameterization of three dimensional hydrological models. Results suggest that amplifying soil hydraulic conductivity in regions where aggregation of observations at model scale leads to loss of topographic information content may increase model performance, an important finding for high-resolution, large-scale physically based modeling that requires detailed subsurface parameterization.
Moisture dependent irrigation and its feedbacks with integrated hydrology In this example, ParFlow was used in conjunction with a novel linear optimization water allocation module to evaluate the impact of groundwater-surface water interactions on moisture dependent irrigation in the Little Washita River Basin, Oklahoma.
Modeling the Critical Zone in an ephemeral West-African Monsoon system West-African hydrosystems are driven by seasonal monsoon dynamics with high variability. While the extreme drought in the 70's and 80's was surprisingly associated with runoff and water table recharge increase and (due to simultaneous land use change) in the Sahel, consequences in the southern, more humid Sudanian zone were a major drop in streamflow. Streamflow generation processes in this area involve subsurface processes (as opposed to Hortonian-dominated Sahelian processes) including temporary connexion of perched and permanent water table. Water table drawdown during the dry season is controlled by land cover distribution and dominated by tree transpiration. This highly connected critical zone system (see e.g. Hector et al., 2015) requires integrated modeling to assess sensitivities to global changes and guide policymakers in this part of the world. The fully coupled ParFlow-CLM simulation of a small watershed in northern Benin (the Ara catchment) shows the intermittent streamflow generation (surface saturation as blue patches), saturation variations in the surface and unsaturated zone, and both permanent and perched water table changes (saturated, deep blue zones in the bottom of the domain) as a response of the interplay between precipitation (spatially uniform, temporally variable) and evapotranspiration (different vegetation classes across the domain). This simulation was successfully compared to a complete dataset (streamflow, water table, soil moisture, evapotranspiration, water storage...) produced by the AMMA-CATCH observatory (http://www.amma-catch.org/). It allowed to assess the importance in vegetation spatial distribution in partitionning the different terms of the water budget and thus calls to enhance the inclusion of land cover.
ParFlow is an open-source, community integrated hydrology model that is freely available on GitHub. It has well over 100 active users, and development is a callaborative effort between several institutions: Juelich Research Centre, Princeton University, Lawrence Livermore National Laboratory, Colorado School of Mines, Bonn University, Washington State University, Syracuse University and the Universty Grenoble Alpes. The Centre for High-Performance Scientific Computing in Terrestrial Systems and the Integrated GroundWater Modeling Center help anchor the development team.
For new ParFlow users, we recommend reading the Parflow user's manual. It contains lots of useful information on getting started, a complete table of published studies that have used ParFlow for a range of applications (which are listed below), some very helpful annotated examples and a complete library of the keys for running simulations and tools for postprocessing output. If you are building a model of a real domain there are helpful blogposts on a workflow for setting up and spinning up models on the ParFlow Blog. Also the blog contains posts on trouble shooting slow model performance and common errors when starting a simulation that will be useful to new users. Finally, the blog contains a lot of useful advice on getting ParFlow compiled on many different platforms.
ParFlow is released on GitHub. The stable release version, older versions, and the latest ParFlow developer versions are available on GitHub under https://github.com/parflow/parflow. Please refer to the GitHub documentation and the blog on how to obtain and build the code. If you are using ParFlow, please subscribe to the ParFlow users mailing list; we try to keep track of active users.
The latest stable release is v3.11.0 (2022-09-06), which also contains many test cases including data to get started.
ParFlow is released under the GNU LPGL license agreement.
If you use ParFlow in a publication, please cite the these papers that describe model physics:
If you use ParFlow coupled to CLM in a publication, please also cite two additional papers that describe the coupled model physics:
To report bugs or request features, please use the ParFlow Issue Tracker on GitHub. Please note that ParFlow is a community supported research code and while we will attempt to answer questions posted to this list your patience is appreciated.
List of publications that use ParFlow (last update: 2022-04-05):