Drought Decision Analysis

The Drought Decision Problem

Drought emerges slowly and may endure over multiple years. And at the start the decision-maker is faced with uncertainty: Is this going to be a deep drought? Will it last into next year, or the year after that? When is it time to start taking drought response actions? How aggressive of a response is called for?

Such decisions have an awkward structure: Some choices must be made now, before the full potential of drought is clear. Later it might not be possible to top off reservoirs, lease the snowmaking water, or plant an alternative crop. But the costs incurred now may not pay off if the drought does not worsen, and decisions may cause unnecessary loss: unhappy rate payers, lost farm income, or wasted investment. On the other hand, if the drought lingers and worsens, the decision-maker will face a different set of regrets if they choose to do nothing: Rate payers may be miffed at modest cut-backs now, but will surely be angry at steeper cuts next year.

In these ways the drought response decision is like most decisions made under uncertainty, and tools exist to help with such decisions. Drought is different from some weather and climate related decisions such as whether to evacuate a city in front of a hurricane, activate the orchard’s frost protection systems, or to call in the full snowplow staff overnight in the face of a winter storm forecast, mainly in terms of the time frame of the choices. But, like those other decisions under uncertainty, drought response follows and arc of: judging the options, weighing possible outcomes, and considering the regret costs of decisions made or not made.

Risk and decision analysis tools can help with such decisions. The formal process for analyzing and making such decisions is most well developed in water resources where best-practice templates and tools are available and widely used. In many other drought-sensitive endeavors, decision-makers cope with uncertainty in various ad hoc ways. In some cases they face a single choice: lease the extra water or not. In others, the choices evolve along with the drought, the season, and the budget available, some options foregone as the season and drought progresses, others newly emerging.

The Drought, Ranching, and Insurance Response (DRIR) Model

WWA's Climate Adaptation Decisions program began to focus on drought impacts in the range livestock industry ("ranching") during FY 15/16 as part of a collaborative project with the USDA Northern Plains Climate Hub and the DoI North Central Climate Science Center. The goals of the project, established at retreats among the three organizations, were to combine drought impact calculators developed by the USDA Agricultural Research Service (ARS) with a ranching drought decision model, developed by Jeffrey Tranel, Rod Sharp, and John Deering, economists with the Colorado State University Cooperative Extension Service (available here). With support from NIDIS we then expanded the modeling effort aiming to help ranchers make herd management decisions in extreme drought, given uncertainties about the market, feed prices, and next year's climate. Finally, in collaboration with CU's Earth Lab, we added a module that calculates likely pay-off of the USDA Risk Management Agency (RMA) range insurance program. The structure of the combined model is shown below.

A beta version is now available for downloading here, along with guidance on running the model.

DRIR Sample Runs

The model simulates a five-year period in which a drought is embedded, calculating annual profit and end-of-year net worth for a small set of scenarios of typical Colorado cow/calf ranch sizes, or the ranch parameters (e.g., herd size) can be set by the user. The version downloaded here includes actual rainfall data for the Central Plains Experimental Range (CPER) and insurance indemnification is calculated for the RMA grid cell that encompasses the CPER. The full DRIR model calculates the costs and revenues associated with five drought management options: no adaptation, buy additional feed, truck cow-calf pairs to rented pasture, sell pairs and replace cows, and sell pairs without replacement. The version available here gives results for the first three options, plus a normal, non-drought year, for baseline comparison.

The user can set the starting year, choose a ranch scenario, and set prices received as well as the costs of feed and rental pasture if they wish.

Results for a run starting with a drought in 2002, with and without the USDA insurance show both the costs of adapting and the marked improvement in income with insurance that makes up for extra costs and bring net income from the cow/calf operation up to a normal (no drought) year.

 

Note that since this is 5-year run based on real climate data it happens that the insurance pays off (based on the precipitation grid in which the ranch resides) in 2003, 2004, and 2006 in addition to the starting drought year (2002). But in all these years the ranch rain gage record, input to the drought calculator, showed more than 80% of normal forage likely for the season, and thus the rancher took no response. BUT the grid cell in which the ranch resides did show sufficient drought, given the policy selections made for this run, to receive an insurance pay out.

The runs below, starting with 2012, another significant drought year in the region, illustrate an unusual outcome of ranch drought decision making and insurance: in 2016 the ranch rain gage does trigger adaptations, but the gridded rainfall product on which the insurance is based does not trigger a payment. In that case it pays not to have purchased the insurance (and not incurring the premium expense), but the rancher cannot know when these decisions are made whether the insurance will pay off. When we do longer runs, over decades with and without insurance we find that overall it pays to buy the insurance, at least for this RMA grid cell and partly because the premium is subsidized; insurance pays off sufficiently frequently to cover the premium costs in the long-run.

 

Next Steps

The DRIR model is being further developed as a research tool in cooperation with CU's Earth Lab. This simulation tool (DRIR-R) is written in the R programming language to allow faster and more complex simulation runs, on-line experiments with many users, and in-lab simulations with actual producers trying out alternative strategies.

An on-line version of that model can be run at: http://www.ranching.io

The online version includes a screen-cast tutorial, a practice run, and a ten-year simulation. You can choose whether to have drought insurance or not.

A forecast version of the DRIR model is being built to assess the value of a seasonal drought forecast of various skill levels made available to the rancher at the point of decision.

 

Contacts

The Climate Adaptation Decision Models team includes: Bill Travis, Adam McCurdy, Joseph Tuccillo, Trisha Shrum, Max Roland, Evan Lih, and Travis Williams.

For information about the Ranch Drought project, contact: william.travis@colorado.edu
For information about the R-DRIR simulation tool, contacts: trisha.shrum@colorado.edu