Photo of Fall River.
  • Cool, wet weather from mid-February to mid-April turned an average water supply into one that is decidedly above average.
  • As of May 9, the Henry’s Fork reservoir system is 94% full and filling rapidly.
  • Based on early-April conditions, summertime water supply in the Henry’s Fork watershed is forecast to be above average.
  • More storage water will be delivered from Island Park Reservoir this summer than in 2017, but higher inflows will compensate, resulting in a very high probability of better-than-average carryover at the end of the irrigation season.

April Precipitation 156% of Average

Precipitation during the month of April was above average at all 12 weather stations in the watershed and was 156% of average for the watershed as a whole. Below-average temperatures early in the month were offset by above-average temperatures late in the month, and the month ended up within 1 degree F of average.

Climate summary for the month of April 2018.

April climate summary table

Cool temperatures through the middle of April capped a nine-week run of average to below-average temperatures, allowing snow to continue to accumulate until the middle of the month and preventing high runoff until late April.

Graph of temperature departure for 2018

Graph of temperature departure from average so far this water year.

Snow-water-equivalent (SWE) peaked in late March at the two low-elevation SnoTel sites in the northern part of the watershed but peaked between April 15 and April 27 at the other 7 SnoTel sites. Over the whole watershed, SWE peaked on April 18, 8 days later than average. Peak SWE accumulation was 117% of average, and the month of April ended with 112% of average SWE on the ground.

SWE Summary for the month of April 2018.

SWE summary table for April 2018

As of May 9, 74% of this year’s peak SWE remains on the ground, compared with an average of 82%. However, because 2018’s peak started out higher than average, current SWE is still 106% of average despite high melt rates over the past week.

Graph of SWE in 2018

Watershed-averaged SWE through May 9.

All in all, a long stretch of cool, wet weather from mid-February through mid-April turned an average water supply into one that was decidedly above average at the end of the month. The current warm spell is melting snow rapidly, and snow remaining on the ground will probably equal average in a few more days. The following graph shows that natural streamflow was a little above average all winter and didn’t really start to reflect snowmelt until late April. Note how much higher flow was in March and early April of 2017, when runoff began very early.

Graph of natural streamflow in 2018

Total natural streamflow in the Henry’s Fork watershed through May 8.

Predictions for Summer 2018

Last year I developed a numerical computer model to simulate water supply and reservoir operations in the Henry’s Fork watershed. I made several refinements this year to make it more realistic and flexible (see details in next section). One of the new components I added this year was a variable low-flow target in the Henry’s Fork at St. Anthony.

Importance of low-flow target at St. Anthony

During irrigation season, just enough water is delivered from Island Park Reservoir to meet irrigation demand in the Henry’s Fork watershed. The downstream-most diversion point on the Henry’s Fork is the Consolidated Farmers Canal, which is downstream of St. Anthony. This canal does not have a full diversion structure spanning the entire width of the river, so at least 500 cfs must be delivered to that point for the canal to divert its full right. Excess remains in the river downstream. Experience on the part of water users and managers has shown that if flow in the Henry’s Fork at the St. Anthony gage is around 1,000 cfs, sufficient water makes it past the intervening three diversions to allow diversion at the Consolidated Farmers canal. Although we do not have solid biological data, the general sentiment among anglers and fisheries biologists is that 1,000 cfs is enough to sustain the fishery downstream of St. Anthony, although more water would certainly result in more fish habitat in the lower Henry’s Fork. However, increasing the low-flow target at St. Anthony increases required delivery from Island Park Reservoir, to the detriment of fishing experience, water quality, subsequent winter flow and trout recruitment in the reach between Island Park Dam and Riverside.

So, I used the model this year to quantify for the first time the relationship between summertime flow at St. Anthony and Island Park Reservoir storage carryover, and hence, next year’s winter flow.

Primary model results: St. Anthony target set at 1,000 cfs

Overall, natural streamflow is predicted to be above average in the upper Henry’s Fork (upstream of Ashton) and quite a bit higher than last year. Fall River will contribute above-average flow but not nearly as much as last year. Flow in Teton River will be near average. Runoff timing will be a few days earlier than average across the board, a little later than last year in the upper Henry’s Fork and a little earlier in Fall River and Teton River. The end result is that natural flow in the Teton River will fall below average in the middle of summer, so more water will need to be delivered from the Henry’s Fork through the Crosscut Canal than in 2017. This will require more delivery out of Island Park Reservoir than last year, at the same St. Anthony flow target of 1,000 cfs that was used in 2017.

The following graphs each illustrate a key model result. In all graphs, the solid blue line is the predicted value (the average over all of the random simulations—see below), and the gray shaded area depicts the values that will occur with 90% probability, given the April-1 conditions on which the model was based.

Inflow to Island Park Reservoir

is predicted to be higher than average early in the spring because of earlier-than-average snowmelt, lower than average from May through mid-July for the same reason, and average to slightly above average during the late summer, reflecting good baseflows after two years of above-average snowpack. For the same reason, inflow this year is predicted to be about 200 cfs higher during July, August, and September than what was observed last year. This also reflects the second consecutive year of above-average snowpack in the springs that feed the upper Henry’s Fork.

Graph of simulated inflow to IP Reservoir

Streamflow in Fall River at Chester

is expected to be above average during the spring because of above-average snowpack and early runoff timing, and average to slightly above average most of the summer. Late-summer flow is expected to be a little lower than last year, but overall, natural flow in Fall River will be sufficient to meet irrigation demand on Fall River and leave 300-500 cfs in the river down to the Henry’s Fork confluence.

Simulated flow in Fall River

Natural streamflow in the Teton River

will have similar runoff timing to that in Fall River but will drop below average  by late June and stay there most of the summer.

Simulated flow in Teton River

Delivery of water to the Teton River

through the Crosscut Canal is expected to begin in late June. Predicted delivery will be higher than that in 2017 during the early part of the summer and lower during the late summer. However, uncertainty is high, and higher-than-average delivery is possible through July and August.

Simulated flow in Crosscut Canal

Streamflow in Henry’s Fork at St. Anthony

generally reflects runoff in Fall River during the spring and early summer. From June onward, flow is predicted to be near average. At worst, the low-flow target—in this case 1,000 cfs—will determine system operations from late June through mid-September, but under the wettest scenarios, streamflow will remain near 2,000 cfs all summer without requiring any storage delivery. The expected flow is somewhere in between, generally around 1,200-1,500 cfs, which happens when the low-flow target constrains operations for only a short time period during the summer.

Simulated flow in HF at St. Anthony

Outflow from Island Park Reservoir

is expected to fall from around 1,000 cfs once the reservoir fills in May down to 660 cfs by late June, when irrigation delivery will be needed. Because of the 660-cfs power-plant constraint, the 90% prediction interval is pretty narrow during mid- to late-June, when flows will be between 660 cfs and 750 cfs with high probability. Under a few scenarios, flow could be as low as 480 cfs during this time frame. Once irrigation delivery starts, outflow will increase to around 800 cfs and remain there for most of July and August, generally a few hundred cfs higher than that in 2017. Under the worst-case scenario, outflow will reach 1,300 cfs from early July through early August.

Simulated outflow from IP Reservoir

Island Park Reservoir

will fill to capacity in May and remain there until late June, when irrigation delivery begins. Higher outflow this year than in 2017 will be offset by higher inflow, resulting in reservoir draft that follows the 2017 curve very closely, ending up at around 110,000 ac-ft (81% full). At worst under the 1,000-cfs St. Anthony target, the reservoir will end up very close to average, a little less than 60,000 ac-ft (44% full). Under the very wettest scenarios, no storage will be needed.

Graph of simulated volume in IP Reservoir

Effect of low-flow target at St. Anthony

The following graphs show the effect of the low-flow target at St. Anthony on Island Park Reservoir carryover and 2018-2019 winter flow. One thing to note right away is that the bottom of all of the 90% prediction intervals is very low. This is because 95% of all possible outcomes are better than the bottom of the gray shaded region (it’s a 90% interval because 5% of the outcomes lie below the bottom and 5% lie above the top). In other words, a lot of things have to go wrong to be that bad—a hot dry summer, lowest possible flow in all streams given this year’s snowpack, and high irrigation demand. So think of the bottom of the gray shaded interval as the worst case scenario, given where we started on April 1, with an above-average snowpack and no substantial snowmelt during the month of March.

Reservoir carryover

decreases as the St. Anthony flow target increases, and the curve becomes steeper as the St. Anthony flow target increases. Predicted reservoir carryover ranges from 120,000 ac-ft (89% of capacity) at a target of 800 cfs to 100,000 ac-ft (74% of capacity) at a flow target of 1,200 cfs. Under all St. Anthony targets between 800 cfs and 1,200 cfs, the wettest scenarios still allow the possibility that no reservoir storage is needed. On the bottom end, the effect of the St. Anthony target is even more pronounced; the bottom of the 90% prediction interval falls from 80,000 ac-ft to 20,000 ac-ft as the St. Anthony target increases from 800 cfs to 1,200 cfs. This is because above-average flow at St. Anthony is inconsistent with below-average supply and above-average irrigation demand, requiring disproportionately high storage delivery to maintain.

IP Reservoir carryover vs. St. Anthony flow

Winter flow in 2018-2019

also decreases as the St. Anthony flow target increases. Winter outflow is predicted to be lower than that in 2017-2018 but still above average. However, under the worst-case scenario, winter flow drops from around 270 cfs to 100 cfs as the St. Anthony target increases from 800 cfs to 1,200 cfs.

Graph of winter flow vs. St. Anthony flow

The graph above assumes constant outflow from October 1 through March 31. In practice, outflow is reduced during October and November to store more water when temperatures are warmer and aquatic plants provide fish habitat. This allows winter outflow to be higher. If outflow is reduced to 100 cfs during October and November, winter flow improves to about 80 cfs higher at the 800-cfs St. Anthony target, but the improvement possible by decreasing flows during October and November shrinks to 0 at a St. Anthony target of 1,200 cfs.

Winter flow with storage early in fall

Improving our predictive capabilities

Now that we have quantified how sensitive reservoir carryover and winter flow are to the St. Anthony target, HFF will be putting a large amount of effort into studying relationships among streamflow at St. Anthony, streamflow in the reach between St. Anthony and the Parker-Salem highway, groundwater return flows from irrigation seepage and from managed recharge in the lower watershed, water temperatures, and trout habitat availability. The goal will be to develop physically and biologically based flow targets at St. Anthony—as well as managed aquifer recharge objectives—that maximize benefits to the fishery in the lower Henry’s Fork while minimizing delivery of storage from Island Park. This study will constitute the Ph.D. research of former HFF intern and research assistant Christina Morrisett, who will begin her Ph.D. program at Utah State University in August of this year. Her academic adviser is Dr. Sarah Null, one of the most accomplished and knowledgeable watershed scientists in the western U.S. Dr. Null specializes in management of heavily regulated rivers like the Henry’s Fork to balance aquatic habitat with water use. By the time Christina is finished, we will have an integrated groundwater-surface water model of the whole watershed that will help refine management to the benefit of all stakeholders.

Overview of the Simulation Model

The model is what we call a stochastic model, which means that the inputs to each different simulation are randomly drawn from multivariate probability distributions that reflect the range of possible outcomes that could happen, given snow, streamflow, and weather conditions known or forecast on April 1. For each management scenario (in this case, different values of the St. Anthony flow target), 5,000 different random combinations of inputs are selected and used in a season-long simulation. The predicted value of a given flow variable is the average of the 5,000 different outputs. The 5th and 95th percentiles of the 5,000 simulations define the bottom and top of the 90% prediction interval.

Model variables

At daily time scale, model tracks:

  • Natural streamflow
    • Henry’s Lake inflow
    • Henry’s Lake to Island Park Reservoir inflow
    • Island Park to Ashton inflow
    • Total Fall River natural flow
    • Total Teton River natural flow
  • Irrigation diversion
  • Lower-watershed river-reach gains and losses
  • Reservoir storage and delivery
  • Resulting regulated streamflow

Operational assumptions

  • Fill reservoirs (Henry’s, Island Park, Grassy) as soon as possible.
  • After fill but before delivery needed, set reservoir outflow to equal inflow.
  • After that, set Henry’s Lake outflow to 70 cfs for the remainder of the summer.
  • Set low-flow target in HF at St. Anthony (default is 1,000 cfs).
  • Set Crosscut Canal delivery to 0 except when Teton natural flow does not meet demand plus stream reach losses.
  • When delivery of water to the Teton River is needed through the Crosscut Canal, diversion into the Crosscut Canal equals the amount needed in the Teton River, adjusted for gains or losses in the canal itself. For example, if Teton River natural flow is 400 cfs, total irrigation diversion in the Teton River downstream of the Crosscut Canal is 600 cfs, and channel loss in that reach of the Teton River is 200 cfs, then 400 cfs needs to be delivered to the Teton River through the Crosscut Canal. If canal loss in the Crosscut is 40 cfs, then 440 cfs is diverted from the Henry’s Fork into the Crosscut Canal.
  • On first day storage is needed (or July 15, whichever is later), set Grassy Lake outflow to 50 cfs.
  • Deliver 3,000 ac-ft from Grassy Lake (30 days at 50 cfs).
  • Last, set Island Park Reservoir outflow to keep flow at St. Anthony no lower than target.
  • When required Island Park outflow is between 480 and 660 cfs, set it to 660 cfs, which is roughly one power-plant turbine at capacity and the second at its minimum operating flow.

Random components

The following components are selected randomly from probability distributions predicted by statistical models:

  • Total volume of natural streamflow over the April-September period (predicted from last winter’s streamflow and April-1 SWE)
  • Streamflow timing (predicted from April-1 SWE  and projected Apr-Jun temperature)
  • April-June watershed-averaged temperature (projected from 28-year trend)
  • Daily natural streamflow (calculated from a randomly selected streamflow volume and a hydrograph matching a randomly selected value of runoff timing)
  • Correlation in volume and timing among subwatersheds
  • Diversion (projected from 36-year trend)

The following graphs show some of the random model components, so you can see how they are generated.

Example of natural-flow probability distribution

based on 5,000 random values of summertime flows in the reach between Henry’s Lake and Island Park Reservoir.

Histogram of randomly selected flows

Mean April-June temperature

is an important predictor of runoff timing and has followed a steadily increasing trend over the past 30 years. Colors represent individual SnoTel stations. The thick black line is the watershed-wide trend, and the diamond is the predicted value for 2018, with its 95% prediction interval. Temperatures were randomly drawn from the distribution around the 2018 prediction.

Graph of April-June temperature trend

Years used in the simulation of runoff timing

were selected randomly based on the probability that runoff timing in a given year matches that predicted in 2018 from April 1 SWE and temperatures selected from the distribution depicted above. Note that this year’s runoff timing has a much greater chance of matching that in 2012 or 2016 than any other year going back to 1978. In fact, no year prior to 1999 was selected in the random sample of 5,000 years, reflecting the systematic increase in temperature shown above, which has produced a corresponding trend toward earlier runoff timing. In other words, under current climatic conditions, the probability of experiencing runoff timing similar to that observed in any year between 1978 and 1999 is essentially 0.

Table of years selected randomly for runoff timing

Simulated daily streamflow

is calculated by first randomly selecting a water year from the distribution shown in the table above. The April-September hydrograph from this water year matches the predicted timing of runoff. This hydrograph is then scaled to have a total volume of 1 so that it contains information only on timing of streamflow throughout the summer but not any information about total volume of streamflow. Then this “unit” hydrograph is multiplied by a randomly selected streamflow volume to obtain a streamflow hydrograph that has both correct volume and correct flow timing. An example is given below, which was generated randomly from a volume that was about 8% higher than average and hydrograph shape equal to that in 2002.   

Example hydrograph simulated by model

Correlation in flow among river reaches

is critical to ensure that the random streamflow simulations are consistent across the whole watershed. Note that correlations are fairly strong; a wet year in one river reach is associated with a wet year in another river reach. The correlation between Fall River and Teton River is particularly strong because both are dominated by snowmelt and share a common drainage divide at the northern end of the Teton Range. (HL = Henry’s Lake inflow, HLtoIP = Henry’s Lake to Island Park inflow, IPtoAsh = Island Park to Ashton inflow, FR = Fall River total natural flow, TR = Teton River total natural flow).

Correlations among simulated flow in different river reaches

Annual volume of irrigation diversion

has been steadily decreasing over the past 35 years because of increases in irrigation efficiency. Statistical analysis of this trend shows that the best prediction of diversion in a given year is last year’s value plus a little adjustment for the trend over the past few years. Thus, the predicted diversion in 2018, is just a little higher than that in 2017. However, the 95% prediction interval is large.

Trend in irrigation diversion since 1981

Years used in the simulation of diversion

were selected randomly based on the probability that total diversion volume in a given year matches that predicted from the time-series in the graph above. Only years since 1988 were used in the random selection to reflect modern irrigation practices and water-rights accounting data. As with the runoff-timing simulations, years with diversion similar to that predicted for 2018 are more likely to have occurred in the last 15 years than in the previous 15. In particular, nearly 62% of the years in the random sample of 5,000 occurred in the past 15 years, versus less than 38% in the previous 15 years.

Distribution of years used to simulate diversion