About the Author

Virtual Farms for Water Resource Decision-Making

By Karthika Swamy Cohen

Large agricultural regions often see overpumping of aquifers beneath their surface due to the huge demand for water for irrigation. Combined with long-term droughts, this has led to near extinction of underground water resources in such areas. Overuse of water leads to salt water infiltration of coastal aquifers, reduction in drinking water, and land subsidence. 

In order to resolve these issues, it is important to consider community-based solutions with an emphasis on stakeholder objectives. 

To help with this sort of decision making, Lea Jenkins of Clemson University and Katie Fowler of Clarkson University, along with their research teams, have developed a virtual farming tool to help strawberry farmers in California improve water resource and land allocation decision-making. The simulation-based optimization tool allows users to work with a multi-component region and account for environmental, urban, and agricultural factors for evaluation. 

At the SIAM Conference on Computational Science and Engineering being held in Atlanta, GA this week, Jenkins described their virtual farm model as part of a minisymposium on “New Approaches to Complex Coupled Multiscale Systems.”

The virtual model helps farmers make decisions about what crops to plant at what times of the year, how long to plant them, what to plant in the interim periods between crops, and so on. It also gives guidance on water usability requirements, and crop yields. In some cases, it can even tell farmers what they should be growing on their farms! 

Jenkins described a situation where plugging in parameters into the model determined that strawberry farmers would be better off growing raspberries! “We said, ‘We came up with a solution - you guys should be planting raspberries. You have a lot of rain, you should be able to store it.’”

The virtual farm model allows the testing of various scenarios and formulating objectives to model stakeholder interests along with many decision variables. This allowed the research team to advise famers on maintaining profitability under water restrictions, and rotating and removing crops from their portfolios.

At a minisymposium during the SIAM Conference on Computational Science and Engineering, Lea Jenkins describes a virtual farming tool to improve water resource and land allocation decision making. Photo credit: Karthika Swamy Cohen.
Driscoll, which purchases strawberries from the farmers, was interested in developing a community-based solution. Use of water had to be justified since some of the other stakeholders in the region would be affected by overuse. This was complicated by the fact that local water management agencies in the region had no enforcement paradigm. 

The virtual model is important so decisions can be made in a virtual setting rather than encountering water shortage in the real world. For instance, the central valley region of California grows almond trees. As Jenkins put it, “The problem is you can't really pull almond trees from the ground as easily as you can pull strawberry plants.”

The model is built on top of MODFLOW, the USGS's three-dimensional finite-difference groundwater model. Integrating MODFLOW with the One-Water Hydrologic Flow Model (MF-OWHM) enables it to track all the water continuously over various irrigation states over time. 

There are so many different moving parts in the system as a whole that it is important to know how these different parts interact, Jenkins said. 

OWHM simulates the consumption and redistribution of surface water and groundwater for farming. It also allows researchers to get precipitation and other weather-related stats and pull it into the model. 

The hydrologic model is then coupled with an optimization package to optimize agricultural water use. The package the team used was the DAKOTA optimization package. 

The linked MODFLOW-OWHM with the DAKOTA software tool was used to analyze three crops with competing properties in terms of water use, profitability, and demand. Using optimization, the team determined the distribution of crops a farmer should plant by generating trade-off curves and analyzing competing objectives such as water-usage, profit, and meeting the market demand. 

“We are not the first ones to pair farming models with optimization techniques but it was the first with a multi-objective approach,” said Jenkins.  “We are developing the tool by combining existing software frameworks to take advantage of expertise in large-scale farm simulations and optimization algorithms.”

This has required the team to write wrappers to allow for communication between the tools, and it has necessitated that researchers work closely with a variety of industry partners to ensure the results are representative of the problem they want to resolve. "The challenge has been to work within the paradigm we were given and make our problem fit within the software tools we could access," said Jenkins.

The framework for crop management and water conservation efforts will no doubt be a useful tool for agricultural sectors in various regions. The open source approach also means that the model can work with new or existing models. Future work will involve better uncertainty quantification, model improvement, and alternative agricultural and hydrological settings.

Read a detailed account of Fowler and Jenkins’ work on virtual farms in the April issue of SIAM News. 

  Karthika Swamy Cohen is the managing editor of SIAM News