Climate: Record-breaking cold in October, a generally dry winter, a 9th-inning rally that brought snowpack up to average just in time for it to melt, and a very dry summer. Mean temperature for water year 2020 was equal to the 1989-2019 average, but high variability in springtime temperatures melted snow rapidly during warm periods around May 1 and June 1. Water-year precipitation ended up at 90% of average, but July-September precipitation was only 52% of average.
Natural streamflow: 92% of the 1978-2019 average: 88% in upper Henry’s, 94% in Fall River, and 96% in Teton River. Streamflow dropped rapidly in all three subwatersheds after snowmelt peaks in early May and early June and after late June rain. Since 2009, natural flow in the upper Henry’s Fork has exceeded the 1930-2020 average in only three years: 2011, 2012, and 2019.
Irrigation Management: Draft of Island Park Reservoir started on July 4 (11 days later than average) and ended on September 9 (6 days earlier than average). Delivery of water to the Teton River through the Crosscut Canal started on July 8 and ended on September 24. Over the period of draft, streamflow in the Henry’s Fork downstream of all diversions averaged 469 cfs, compared with this year’s target of 350 cfs. Using the target downstream of all diversions rather than the St. Anthony gage saved 1,226 ac-ft in Island Park Reservoir while increasing streamflow downstream of diversions by 5.4%.
Island Park Reservoir: Winter flow during water year 2020 was 532 cfs, compared with 560 cfs in 2019 and a 1978-2019 average of 356 cfs. The reservoir reached a minimum of 54% full but ended the water year 56% full, compared with 46% full on average. Outflow during the upcoming winter is expected to be around 400 cfs.
Predictive Model Performance: Based on April 1 conditions, predictions were within 10% of actual values for hydrologic and water-management parameters that depended on spring and early summer conditions. However, the model under-predicted need for Island Park Reservoir draft by 20-30%, depending on the particular metric. Given observed flow in the lower Henry’s Fork, the model over-predicted Island Park Reservoir carryover by 27.2% (97,135 ac-ft vs. 76,366 ac-ft).
Climate
Water year 2020 started off wet and cold, with numerous low temperature records set October 29-31. Some individual station temperatures were over 20 degrees F colder than average. However, after the cold, wet start to the year, November and December were warm and dry, leaving snow water equivalent (SWE) at less than 80% of average on January 1. A series of storms increased SWE to average in early February, but dry weather dropped that below 90% of average in March. Only some very wet weather in April saved the snowpack, which peaked at 100% of average on April 17, within a few days of average peak timing. However, SWE varied greatly across the watershed. When SWE peaked on April 17, SWE was 111% of average in the Teton subwatershed, 98% in Fall River, and 96% in upper Henry’s Fork.
The spring of 2020 can best be characterized by high variability in both temperature and precipitation. Averages do not tell the whole story. Mean temperature for the spring ended right up at average. At the nine SnoTel stations, mean April-June temperature was 42.6 degrees F, compared with 42.6 degrees on average and with 42.4 degrees F last year. The past two springs have provided relief from a four-decade trend of increasing springtime temperatures at these nine stations. Based on that trend, about 1.1 degrees F per decade, last spring’s mean temperature was predicted to be 44.1 degrees F and this spring’s was predicted at 44.2 degrees. So, in both springs, temperatures were about 1.6 degrees below expectations.
However, daily variability in springtime temperatures was much greater this year than last, which is readily apparent in the temperature graphs. A useful measure of variability is the coefficient of variation, defined as standard deviation divided by the mean. Daily coefficient of variation in temperature this spring was 24%, compared with 19% last year and 21% on average. Starting with the week before Memorial Day, we had four episodes during which temperatures dropped from 5-10 degrees above average to 5-20 degrees below average over the course of 24-36 hours. These very strong and unseasonable cold fronts were generally accompanied by high winds and brought heavy precipitation in three of the four cases—those being the Memorial Day weekend, June 16-18 and June 28-30.
Despite average snowpack and average temperatures, the early melt was a result of two extended periods of well above average temperatures, one around May 1 and the other around June 1. During the first extended period of warm weather at the end of April and beginning of May, all of the low-elevation snow melted rapidly, and SWE remained below average the rest of the spring except for a day or two around the Memorial Day weekend. The warm period around June 1 melted the mid-elevation snow and most of the high-elevation snow. Thus, the high variability in temperatures accelerated snowmelt over what would be expected strictly by looking at averages.
In comparison to last year, snowmelt this spring was about two weeks earlier than last year. This was due in part to the high temperature variability mentioned above but was also due to a much lower snowpack this year than last. More snow takes longer to melt, regardless of temperature. This year’s snowpack peaked at the same time as last year’s, but peak SWE was 4.6 inches (14%) lower than last year. As can easily be seen from the SWE graphic, that difference was maintained throughout the spring. During the mid-April to late-June melt period, the difference between last year’s SWE and this year’s ranged between 2.9 inches and 7.5 inches, averaging 5 inches, pretty close to the initial difference in peak SWE between the two years.
After the cold, rainy period at the end of June, precipitation was sparse. Total July-September precipitation was only 52% of average. The water year ended with accumulated precipitation at 90% of average: 94% of average in the valleys, 87% in upper Henry’s Fork, 89% in Fall River headwaters, and 92% in Teton River headwaters. Compare that to water-year precipitation in 2019 that ended up at 114% of average over the whole watershed: 124% of average in the valleys, 119% of average in upper Henry’s Fork, 108% of average in Fall River headwaters, and 109% of average in Teton headwaters. As a result, three-year average precipitation dropped from 43 inches (4 inches above average) in late September of 2019 to 37 inches (1 inch above average) in late September of 2020. By the beginning of the new water year, most of the western U.S. was covered by drought, the headwater areas of the upper Snake River basin being a notable exception.
The only two of the 12 individual climate stations in the watershed that received above-average precipitation in 2020 were Rexburg at 103% of average and Ashton at 111% of average. In fact, precipitation in the valleys this year was high enough that grain harvest was delayed by several weeks until grain moisture dropped to desired levels, and dry-farmed (non-irrigated) grain yields were high. In 2019, Ashton was also the winner among all stations, at 133% of average precipitation. However, in 2019, all 12 stations received above-average precipitation.
Summer temperatures started off near average in July but were above average for most of August and September. Mean August-September temperature was 2 degrees F above average. When combined with dry weather, accumulated moisture surplus in the valley areas dropped rapidly during the month of September. I define moisture availability in the agricultural regions of the watershed as the difference between precipitation and evapotranspiration, averaged over the previous year. On average, evapotranspiration exceeds precipitation by 37 inches per year. In theory, this is the amount of irrigation needed to fully irrigate an alfalfa crop. In early September, this difference was only 34 inches, indicating a surplus of around 3 inches. However, that 3-inch surplus had turned into a 1.5-inch deficit by the end of the water year, resulting in above-average irrigation demand during August and September.
Although water year 2020 brought many extremes in temperature—and extreme changes in temperature that brought high winds—one event stood out because of its spatial scale and magnitude. A very strong cold front moved into the watershed from the northeast on Labor Day, dropping temperature in Ashton from 74 degrees at 2:15 p.m. to 35 degrees by 7:00 p.m. This cold front was accompanied by strong northeast winds that persisted for 24 hours. This wind event was rare in three characteristics: 1) direction (prevailing wind, and that accompanying most weather fronts blows from the southwest here), 2) magnitude (northeast winds are rarely very strong), and 3) duration (sustained for 24 hours—rare even for the prevailing southwest winds). Furthermore, this cold front generated strong winds throughout the western U.S., fanning wildfires throughout Washington, Oregon, and northern California, and causing extensive damage along the Wasatch Front.
Closer to home, this event had the effect of pushing around 1,000 ac-ft of water in Island Park Reservoir temporarily out to the west end of the reservoir, where large areas of reservoir bottom were exposed because the reservoir was only 54% full at the time. When that water returned back to the east side of the reservoir, where the dam is located, it carried a large sediment load that increased turbidity and sediment in the river downstream for over two weeks. We are still analyzing data from throughout the watershed to quantify the fate of that sediment as it passed through the watershed. I will provide that information in the form of a blog once we have the analysis done.
Natural Streamflow (water supply)
Watershed natural flow in 2020 was 92% of average: 88% in upper Henry’s Fork, 94% in Fall River, and 96% in Teton River. Natural flow in water year 2020 ranked 26th out of the 43 water years since 1978 and closely mirrored the subwatershed precipitation figures shown in the table above. Last year, natural flow was 98% of average, despite 114% of average water-year precipitation. Some of the discrepancy between precipitation and natural flow in 2019 was due to percolation of fall rain and spring snowmelt into soils and aquifers, deferring some of the effect of 2019’s abundant precipitation until 2020.
In fact, good baseflow was the primary reason streamflow in water year 2020 was as good as it was, considering how low precipitation was. October-March natural flow, which depends primarily on groundwater, was 98% of average over the whole watershed, reflecting good precipitation in 2019. April-September natural flow, which depends primarily on snowpack and direct runoff from spring and summer rain, was only 89% of average, reflecting relatively low precipitation this year. July-September precipitation in 2020 was only 52% of average, which is one reason why natural flow dropped from 98% of average over the fall and winter to only 89% of average over the spring and summer.
As mentioned above, springtime temperature this year was highly variable. Low-elevation snow melted rapidly in the last few days of April and first few days of May, producing a very high but short-duration spike in natural flow in the upper Henry’s Fork watershed. Natural flow there dropped rapidly after that spike and stayed below average the rest of the spring and summer. This is readily apparent in the graph of natural watershed inflow between Henry’s Lake and Island Park.
Another period of very warm weather in early June melted mid-elevation and most high-elevation snow, producing the seasonal peaks in Fall River and Teton River. Streamflow in both of those rivers dropped rapidly during early June, enough that it looked like reservoir draft would be needed as early as June 20 to meet irrigation demand. However, heavy rain at the end of June increased streamflow enough to push need for reservoir draft out into July, a few days later than average. Once dry weather took hold in July, natural flow dropped rapidly to around 80% of average, where it stayed the rest of the summer and into the beginning of water year 2021.
The most disappointing news from water year 2020 was a large decline in natural flow in the upper Henry’s Fork. After improving for three consecutive years following the 2013-2016 drought, natural flow in the upper Henry’s Fork dropped from 97% of the 1978-2019 average in 2019 to only 88% of average in 2020. In general, the groundwater springs in the upper Henry’s Fork respond to precipitation over a time frame of around three years. At the end of last water year, the three-year average precipitation had been increasing steadily for three years, peaking at 4 inches above average. Over the past year, three-year average precipitation has decreased steadily, losing 3 inches over the year. Natural flow in the upper Henry’s Fork has reflected this pattern over the past four years: three years of steady increase followed by a large decrease. Upper Henry’s Fork streamflow had just managed to recover to the 1930-2019 average last year for the first time since 2012, before falling below average again this year. Good fishing this year and four consecutive years of good carryover in Island Park Reservoir have masked the reality that the upper Henry’s Fork is still in a long period of dry conditions. Over the past 11 water years, natural flow in the upper Henry’s Fork has exceeded the 1930-2020 average in only three years: 2011, 2012, and 2019.
Irrigation Management
Irrigation diversion followed typical patterns for most of the spring and summer. During periods of cool, wet weather during the springtime, diversion dropped below average, increasing again during warm, dry periods. This kept diversion below average prior to July. Diversion peaked at 3,670 cfs on July 9, right on the median peak date. Diversion dropped rapidly after July 23 as hay was being cut and grain was done being irrigated. Because the summer was very dry, diversion was above average for most of August and September, although it dropped fairly consistently throughout that time period. There were two very small, brief increases in diversion in mid-August and mid-September, associated with watering of third hay crops, grain stubble and potato fields pre-harvest. Although a full accounting of diversion will not be possible until final data are reported after the end of administrative irrigation season, cumulative diversion was 96% of average at the end of the water year.
Draft of Island Park Reservoir started on July 4 (11 days later than average) and ended on September 9 (6 days earlier than average). My “600-cfs” rule of thumb performed well again; start and end of reservoir draft occurred within a few days of when natural equaled 600 cfs of total diversion. The 600 cfs figure is roughly what is needed in excess of diversions to account for stream channel losses and to leave enough flow in the South Fork Teton River and the lower Henry’s Fork to meet demand at the downstream-most canal diversions and leave some water in the stream channel. As expected, need for storage draft coincided well with need for delivery of water to the Teton River through the Crosscut Canal. Crosscut delivery started on July 8 and ended on September 24. Crosscut delivery peaked at 318 cfs on July 22-23 before being reduced for hay and grain harvest during late July and early August. A second, smaller peak of 276 cfs occurred on August 17, associated with the above-mentioned increase in diversion.
Two new features of irrigation management in 2020 improved delivery of water to the downstream-most users and streamflow in the lower Henry’s Fork and South Fork Teton River, while also saving water in Island Park Reservoir. The first of these was installation this spring of remote-controlled headgates on the Crosscut Canal and the Teton River splitter. These two headgates are critical to delivering water to users on the Teton River. In the past, changes at either location required out-of-the-way travel down dirt roads to make manual adjustments. The result was often a delay of 12-36 hours between when changes were needed and when they were made. If an increase in delivery was needed, these delays could leave users on the lower Teton River short of water, whereas if a decrease in delivery was needed, the delay could cost unnecessary draft of Island Park Reservoir. This year, adjustments were made remotely in real time, improving precision in delivery, to the benefit of water users, streamflow, and Island Park Reservoir.
The second feature of management in 2020 was experimental use of a new irrigation-season flow target in the lower Henry’s Fork. Over the past few years, the Henry’s Fork Drought Management Planning Committee had set a target of 1,000 cfs in the Henry’s Fork at the St. Anthony gage. Fremont-Madison Irrigation District (FMID) personnel then used that target as a guide for requesting increases or decreases in outflow from Island Park Reservoir during the irrigation season. This year, after discussion by the Committee, FMID decided to try a different target this year, namely 350 cfs in the river downstream of all diversions. Modeling showed that over the whole irrigation season, this target would be roughly equivalent to 1,000 cfs at St. Anthony from a reservoir management standpoint but would ensure sufficient water at the Consolidated Farmers Canal, the downstream-most point of diversion on the Henry’s Fork. There are four canals downstream of the St. Anthony gage, and because diversion decreases as the summer goes on, fixing a target flow at St. Anthony results in very low streamflow downstream of Consolidated Farmers Canal in July, when water temperatures are warmest, and higher flow in late August and September, when temperatures are cooler. Using a target downstream of Consolidated Farmers Canal would result in more stable flow there throughout the summer, particularly avoiding very low flows during July.
So, how did the two new management options perform? First, on the South Fork Teton River, shortage at Rexburg Irrigation Canal, the downstream-most diversion, is indicated by streamflow in the South Fork Teton River downstream. In 2020, streamflow dropped below 10 cfs downstream of Rexburg Irrigation Canal on only 1 day, compared with 2 days in 2019 and an average of 9 days per year over the 2004-2019 period of record. The coefficient of variation in daily streamflow—equal to the standard deviation divided by the mean—provides a measure of variability. Over the heart of the irrigation season—July 15 to August 31—coefficient of variation in daily flow was 47% in 2020, compared with 66% in 2019 and an average of 78%. This indicates more stable and reliable streamflow, to the benefit of both water users and fisheries.
On the lower Henry’s Fork, streamflow downstream of all diversions dropped below 300 cfs on 4 days this year, compared with 14 days in 2019 and a 2001-2019 average of 8 days. July 15-August 31 coefficient of variation in daily flow was 18%, compared with 35% in 2019 and 30% on average. Again, these figures indicate more reliable and stable streamflow. Over the period of Island Park Reservoir draft, streamflow in the Henry’s Fork downstream of all diversions averaged 469 cfs this year, compared with the target of 350 cfs and 445 cfs in 2019. Streamflow at St. Anthony during the period of draft averaged 1,099 cfs, compared with the target of 1,000 cfs used in previous years and an actual flow of 1,078 cfs in 2019. Thus, compared with last year, St. Anthony flow was 21 cfs (1.9%) higher, while flow downstream of diversions was 24 cfs (5.4%) higher. Statistical analysis shows that after accounting for the effects of total diversion and water supply on Island Park Reservoir draft, a one cfs change in streamflow at St. Anthony during the period of reservoir draft produces a change of 4.3 ac-ft of carryover in Island Park Reservoir. So, compared with 2019, the 21 cfs increase in St. Anthony flow during the period of reservoir draft cost around 90 ac-ft of storage in Island Park but resulted in a 5.4% increase in streamflow in the river downstream of diversions. Thus, paying closer attention to flow downstream of diversions rather than at St. Anthony provided a substantial improvement in streamflow magnitude and stability, at relatively little cost to Island Park Reservoir carryover.
However, these figures don’t tell the full story. Because Island Park Reservoir was drafted much more this year than over the past three years, the Drought Management Planning Committee set a strategy to fill the reservoir as soon and as quickly as possible this year. Outflow was decreased as allowable in September and early October to achieve maximum fill rates while still meeting the lower Henry’s Fork flow target. Reservoir fill started this year on September 10. As of October 7, streamflow in the Henry’s Fork since September 10 had averaged 532 cfs downstream of diversions and 953 cfs at St. Anthony. Had the 1,000-cfs St. Anthony target been used since September 10, reservoir outflow would have been 47 cfs higher, at a cost of 1,316 ac-ft of fill. Subtracting the 90 ac-ft discussed above, this year’s management strategy improved irrigation-season streamflow by 5.4% in the lower Henry’s Fork, while also increasing reservoir storage by a net of 1,226 ac-ft, compared with last year’s strategy.
Island Park Reservoir Management
Because of good reservoir carryover in 2019 and near-average baseflow, winter outflow averaged 532 cfs, compared with 560 cfs in 2019 and a 1978-2019 average of 356 cfs. Winter flow averaged 532 cfs over the past three years, the highest three-year average since water years 1999-2001, at the end of an extended period of above-average water years.
The reservoir remained constant at around 120,000 ac-ft through the winter and into the spring. This is close to the maximum volume desirable when the reservoir is still covered with ice, to avoid ice scour on spillway infrastructure. With ice cover still in place in late April and hot weather forecast, the opportunity arose to deliver a managed freshet designed to move sediment out of the Harriman reach at relatively little cost to reservoir fill. The managed freshet average 1,760 cfs over a five-day period between April 29 and May 4. Because inflow from snowmelt was so high, the freshet delivery reduced reservoir content by only 522 ac-ft, a difference easily made up during fill operations after ice melted. As it turned out, this managed freshet was very close to the river’s natural freshet over that time frame. Net scour of sediment from the river between Island Park Dam and Pinehaven between April 15 and May 3 was 555 tons, about 25% of the mean annual transport of sediment out of that reach. Most of that occurred during the 5-day freshet.
After the freshet, the reservoir was filled over the next few weeks, reaching full pool on June 1. It remained full until draft began on July 4, 11 days after the mean draft start date of June 23. Draft continued until September 9, 6 days earlier than average. Minimum reservoir content was 73,581 ac-ft, 54% of average. However outflow was reduced enough over the last two weeks of September that the reservoir gained 2,755 ac-ft between September 9 and September 30 to end the water year at 73,336 ac-ft (56% full), compared with the 1978-2019 average of 62,675 ac-ft (46% full).
Reservoir carryover has averaged 91,344 ac-ft (71% full) over the past four years. Reservoir carryover has not been this high four years in a row since 1994-1997. Reservoir carryover in 2017-2020 is the fourth highest since 1978, behind only the record-high water years in the mid- to late-1990s. In 1994-1997, watershed-wide natural flow was 118% of average, while that over the past four years is only 102% of average. These figures show how improved reservoir management over the past few years has increased carryover in Island Park Reservoir to the benefit of water users and fisheries. Fremont-Madison Irrigation District and U.S. Bureau of Reclamation deserve a lot of credit for managing the Henry’s Fork irrigation system so carefully to keep the reservoir as full as possible.
Direct precipitation on the reservoir surface added a net volume of 7,582 ac-ft over the water year, compared with an average figure of 6,829 ac-ft. This is water added to the reservoir over and above inflow from streams. The net contribution from precipitation was above average, despite below average precipitation, because the reservoir stayed at around 89% full all winter and so had much larger surface area to capture precipitation than it would have had under average conditions. This is one more benefit to keeping the reservoir as full as possible.
Despite better-than-average reservoir carryover, winter outflow from the reservoir is not going to be as high as it has been over the past three years. Lower carryover and lower baseflow this year will limit winter flow. Reservoir inflow is expected to be around 415 cfs this winter, 96% of average. Outflow will be maintained at 200 cfs for the remainder of October and November, which is expected to fill the reservoir to around 106,000 ac-ft (78% full) by December 1. Outflow will then be increased to somewhere around 400 cfs, at which the reservoir will continue to fill slowly over the winter, reaching around 112,000 ac-ft by April 1. The 1978-2020 average April 1 content is 112,280 ac-ft. If winter outflow is 400 cfs as expected, four-year average outflow for 2018-2021 will be 500 cfs, the highest since 1998-2001.
Predictive Model Performance
In the spring of 2017, I developed the first version of a numerical simulation model to predict water supply, irrigation-system management, and Island Park Reservoir carryover for the upcoming irrigation season. The model uses hydrologic and climatic data available on April 1 of each year to predict conditions from the April 1 through September 30. Model outputs include an expected value (the average of 5,000 different random simulations) and a 90% prediction interval around the expected value. The prediction interval encloses the middle 90% of outcomes of the 5,000 different simulations. By definition, the probability of an outcome outside of the interval is less than 10%, given the data available on April 1. I have made incremental improvements to the model since its initial construction and will continue to refine it each year based on performance. Today I will compare the model’s predictions with actual conditions during 2020. You can see the original 2020 predictions in my blog post of April 2, 2020.
Generally, the model overestimated water supply and underestimated need for Island Park Reservoir draft. The model predicted April-September natural flow at 97% of average, while the actual figure turned out to be 89% of average. The model error, defined as difference between the prediction and actual value, divided by the actual value, was 8.6% (model over-predicted), compared with 8.1% in 2019, 4.3% in 2018, and -7.7% (under-prediction) in 2017. The model predicted April-September natural flow at 1.579 million ac-ft, compared with the actual value of 1.455 million acre-feet. For context, the difference of 125,000 ac-ft between the predicted and the observed values is 92% of the storage capacity of Island Park Reservoir. By subwatershed, relative prediction errors were 16.1% for upper Henry’s Fork, 1.6% for Fall River, and 6.3% for Teton River.
I have summarized the most important irrigation-season parameters in the table below. You can see from that table that the model performed well in predicting all parameters that depended primarily on conditions in the spring and early summer, including total April-September streamflow, mean April-June temperature, June 30 outflow from Island Park Dam, and need for start of reservoir draft and Crosscut Canal delivery.
However, the model performed progressively worse as the summer went on, under-predicting need for Island Park Reservoir draft by 25-30%, depending on the particular parameter considered. The model also did not perform well in predicting September management of Island Park Reservoir, assuming that the reservoir would be held constant once draft was no longer needed. Instead, outflow was lowered in the middle of the month to fill the reservoir as much as possible once draft was no longer needed. However, as shown in the graphs, with the exception of Island Park Reservoir management decisions made in real time during spring (the freshet) and fall (early fill), observed streamflow and reservoir values fell within the model’s 90% prediction intervals most of the summer.
The single most important output of the model is Island Park Reservoir carryover, which I define as the content of the reservoir on September 30. The observed value was 76,336 ac-ft. The model predicted a value of 105,941 ac-ft, an over-prediction of 38.8%. However, reservoir carryover is a decreasing function of irrigation-season streamflow in the lower Henry’s Fork. The model prediction is based on perfect attainment of the target flow, which was set at 350 cfs downstream of all diversions on the Henry’s Fork. Given the observed streamflow of 469 cfs there, the model predicted a reservoir carryover of 97,135 ac-ft, an over-prediction of 27.2%. While better than 38.8%, this error is still very high. The model has now been applied in water years 2017, 2018, 2019, and 2020. Errors in predicting end-of-season content in Island Park Reservoir in those years were <1%, 7.4%, and 6.8%, and 27.2% respectively. The over-prediction in 2020 was due largely to lower-than-expected streamflow during July, August, and September. Another factor resulting in over-prediction of reservoir carryover was underestimation of need for Crosscut Canal delivery, which was a function both of lower-than-expected natural flow on the Teton River and underestimation of the amount of streamflow left in the Teton River downstream of diversions. I will use a higher value for the latter in next year’s version of the model in an attempt to improve performance.
