• Rob Van Kirk

Diversion and Streamflow in the Henry's Fork: Why Doesn't Lower Diversion Lead to Higher Streamflow?

Updated: 6 days ago

  • Due to increases in irrigation efficiency, annual diversion in the Henry’s Fork watershed has decreased by 215,000 ac-ft (20%) since 2000.

  • But, this has been offset by an equal decrease in stream reach gain, which consists primarily of groundwater return flow.

  • Mean water budget in the Henry's Fork has remained constant since 1978, roughly:

  • Inflow = 2,450,000 ac-ft

  • Outflow = 1,620,000 ac-ft

  • Net diversion = net watershed withdrawal (inflow minus outflow) = 830,000 ac-ft

  • Net withdrawal = 330,000 ac-ft consumptive use + 500,000 ac-ft Eastern Snake Plain Aquifer recharge


  • Even with equal streamflow, loss of groundwater inputs increases water temperature and decreases supply for irrigators.

  • Managed aquifer recharge can help increase groundwater inputs to the river, benefitting water users and fisheries.

  • Irrigation year 2020 aligned closely with expectations:

  • Total diversion was 98% of the 2001-2019 average

  • Inflow and outflow were both 92% of the 1978-2019 average.

  • Net diversion was 97% of the 1978-2019 average, while net watershed withdrawal was 96% of the 1978-2019 average.

  • Lower-watershed stream reach gain was 104% of average.

Irrigation efficiency

Irrigation methods changed substantially in the upper Snake River basin between the 1980s and 2000s from flooding and other direct surface-irrigation methods to sprinklers. Although unlined canals still remain the dominant method of conveyance of water from the river to fields, some irrigation companies in the Henry’s Fork watershed have converted their old canal conveyance systems to pipelines. Watershed-wide, these conversions have increased irrigation efficiency, defined here in broad terms as the amount of water used by crops as a fraction of water diverted from the river. Irrigation efficiency is the subject of much recent scientific literature and discussion among scientists and water managers. Without delving into the details, the consensus among most who study this subject carefully is that increased irrigation efficiency generally increases consumptive use of water by crops for a variety of economic, legal, and physical reasons. This is called the paradox of irrigation efficiency. In short, increased efficiency does not save water but instead usually increases water use.


Diversion, return flow, and water balance in the Henry’s Fork

One of the most important features of this paradox in the Henry’s Fork watershed is that more efficient irrigation requires less diversion from the river, which intuitively should result in higher streamflow. Instead, less diversion from the river has not increased the amount of water in the river because pipes and sprinklers result in less seepage into aquifers, which in turn reduces the amount of water that returns to the river through groundwater. Interactions between groundwater and surface water are extremely important to water management in the upper Snake River basin, particularly in the lower Henry’s Fork watershed. Because of this, I conducted a very careful and detailed analysis of trends in diversion and groundwater return flow in the watershed, comparing those trends against long-term trends in water supply (natural flow) and in regulated streamflow at the bottom of the watershed. The complete analysis is contained in the document linked here.


This blog presents the primary results of that analysis, illustrated by key graphics that plot irrigation-year 2020 observations on existing data from the 1978-2019 record. The 2020 data are preliminary and based on real-time observations. Final irrigation data are not published and approved by Water District 1 until March. Because I used preliminary 2019 data in the analysis, the numbers reported here differ slightly from those in the linked document.

I should also mention that London Bernier, an HFF intern from St. Lawrence University in 2020, conducted some more detailed analysis of reach gains in the Henry’s Fork watershed. She presented her findings at HFF’s summer seminar series in August. You can see her presentation here.


Diversion decreased abruptly by 220,000 ac-ft in 2001

Although diversion declined slowly but continuously throughout the 1980s and 1990s, the largest decrease in diversion occurred between 2000 and 2001. Mean annual diversion was 1,090,852 ac-ft from 1978-2000 but has averaged only 868,136 ac-ft since then, a decrease of 222,780 ac-ft (20%). Following that abrupt decrease, diversion has showed no statistically significant trend. Diversion in 2020 was 850,350 ac-ft, 98% of average. This figure does not include diversion into the Crosscut Canal, since that water is delivered into the Teton River, where it is diverted again. I include Crosscut diversion in my daily reports during irrigation season, because that diversion directly affects need for Island Park Reservoir draft and streamflow in the Henry’s Fork downstream, but in retrospective watershed-scale analysis I omit the Crosscut diversion to avoid double-counting. For reference, mean annual delivery of water from the Crosscut Canal to the Teton River was 45,527 ac-ft from 1978-2000 but only 37,524 ac-ft between 2001 and 2019, reflecting the watershed-wide drop in total diversion.

Importance: Irrigation practices changed dramatically in 2001 but not since then. Thus, current irrigation data are most appropriately compared against a period of record that begins in 2


The graph below shows that diversion during irrigation year 2020 closely followed the 2001-2019 average. The primary differences are lower-than-average diversion during spring rains and higher-than-average diversion during late summer and fall, which were very dry.

Reach gain in the lower Henry’s Fork decreased abruptly by 212,170 ac-ft in 2001

Water managers in the upper Snake River basin define “reach gain” as net change in streamflow in a particular reach, after accounting for diversions and reservoirs in the intervening reach. In a reach without reservoirs, for example the Henry’s Fork between Ashton Dam and St. Anthony, reach gain is calculated by the formula


reach gain = inflow - outflow + diversion


If reach gain is positive, water flowed into the reach. If reach gain is negative, the reach lost water. In the lower Henry’s Fork, reach gains and losses almost exclusively occur via interaction with the shallow aquifer. Gains occur when groundwater flows into the river, and losses occur when water flows from the stream channel into the aquifer. For watershed-scale purposes, I define lower watershed reach gain as watershed outflow (streamflow in the Henry’s Fork at Rexburg) minus lower-watershed inflow (Henry’s Fork at Ashton plus Fall River upstream of all diversions plus Teton River downstream of Crosscut Canal) plus total diversion from those river reaches. Thus, lower-watershed reach gain is the net gain/loss of water in the lower reaches of the Henry’s Fork, Fall River, and Teton River as they flow through the irrigated regions of the watershed.


Reach gain dropped by roughly the same amount as diversion did, and at the same time. Annual reach gain in the lower Henry’s Fork watershed averaged 247,677 ac-ft from 1978-2000 but only 35,507 ac-ft from 2001-2019. Reach gain in 2020 was 36,935 ac-ft, 104% of the 2001-2019 average. Although 4% above average seems like a substantial amount of water, that 4% is only 1,428 ac-ft over the whole irrigation year, equivalent to 2 cfs.

Importance: Diversion and reach gain have decreased by the same amount.


The graph below shows that reach gain during irrigation year 2020 closely followed the 2001-2019 average and illustrates the overall pattern of net loss from the river during the winter and net gain during irrigation season.

Reach gain is highly correlated with diversion

Given that reach gain and diversion both dropped by the same amount at the same time, we would expect that the two are highly correlated on a year-to-year basis. In fact that it is the case. Many decades of careful field measurements and modeling, including those made by two of my graduate students 10 years ago, clearly document a physical mechanism (irrigation seepage and groundwater flow) linking reach gains and diversion. Thus, in this case, correlation really is causation.

Importance: In the lower Henry’s Fork watershed, reach gain is essentially equivalent to irrigation return flow, that is, water that was diverted for irrigation but returned to the river before it could exit the watershed either as groundwater outflow or through evapotranspiration.


No long-term trends are apparent in net diversion.

Net diversion is equal to gross diversion minus surface return flow and measures the net amount of water that has been diverted from the surface-water system. In the Henry’s Fork watershed, net diversion is the difference between total diversion and reach gain. Given high correlation and nearly identical temporal trends in the two, we should not expect to see any trends in net diversion over time. Indeed this is true. Net diversion averaged 839,375 ac-ft from 1978-2019, with only relatively small variability around that long-term average. Net diversion in 2020 was 813,415 ac-ft, 97% of average, in line with total diversion in 2020 relative to average.

Importance: Decrease in diversion since 2000 have been offset by an equal decrease in return flow, yielding no change in net diversion from the Henry’s Fork and its tributaries since 1978.


No long-term trends are apparent in watershed inflow or outflow.

The amount of water available for diversion, as well as overall water conditions that could be reflected in reach gain, depends on water supply, which I measure as watershed-total natural flow. To determine whether trends in diversion and reach gain are related to trends in water availability, I analyzed watershed inflow (total natural flow over the water year). Although there is a very slight drop apparent in 2000, that drop is not statistically significant. In other words, no systematic temporal trends are apparent in natural flow, although of course it varies from year to year according to variability in precipitation. The 1978-2019 average watershed inflow is 2.454 million ac-ft. Natural flow in 2020 was 2.256 million ac-ft, 92% of average. As reported in a separate blog, total precipitation for water year 2020 was 90% of average. Good baseflow last winter pushed total natural flow up a little bit compared with precipitation.

Importance: Average water supply has not changed since 1978, so observed changes in diversion and reach gain since 2000 have occurred independent of water supply.


Outflow from the Henry’s Fork watershed is measured by streamflow in the river at Rexburg, downstream of all diversions and tributaries in the watershed but upstream of influence of water diverted from the South Fork Snake River that is delivered to areas in the very lower part of the Henry’s Fork watershed. No statistically significant trends in watershed outflow are apparent. Watershed outflow strongly reflects inflow and has averaged 1.624 million ac-ft since 1978. Outflow in 2020 was 1.495 million ac-ft, also 92% of average.


Importance: Despite a 20% decrease in diversion since 2000, streamflow in the lower Henry’s Fork has not changed over the past four decades and in particular has not increased. This reflects one major component of the paradox of irrigation efficiency: decreased diversion has not led to increased streamflow.


If neither watershed inflow nor watershed outflow have changed since 1978, we would also expect that the difference between the two (inflow minus outflow) has not changed. Indeed that is the case. The difference between watershed inflow and outflow is net withdrawal of surface water from the watershed, in other words, water that originates in the watershed but does not exit it as surface-water streamflow. I’ll discuss the fate of that withdrawn water below. Mean watershed withdrawal from 1978-2019 was 829,843 ac-ft, while net withdrawal in 2020 was 794,333 ac-ft, 96% of average, again in line with the gross diversion and total diversion figures reported above.

Importance: Despite changes in irrigation practices, the net amount of water withdrawn from the watershed has not changed since 1978.


The following table summarizes these key quantities and shows how irrigation year 2020 compared with previous years.

Net diversion = net watershed withdrawal.

By now, it should be pretty apparent that net diversion, as measured by the difference between gross diversion and reach gain, is the same thing as net watershed withdrawal, as measured by the difference between watershed inflow and outflow. Some algebra shows that in fact the two calculations are equivalent, except for two small differences. One is evaporation from and direct precipitation on the watershed’s three reservoirs, a quantity included in the inflow calculation but not reflected in either diversions or reach gain. The second is difference in reservoir system storage between the end of one irrigation year and the beginning of the next, a quantity also included in the inflow calculation but not reflected in diversion or reach gain. Despite these differences, net diversion and net withdrawal are statistically indistinguishable. Both average around 830,000 ac-ft, with slight differences caused by the reservoir evaporation/ precipitation and change-in-storage terms, whether water years or irrigation years are used to define annual quantities, and lag between the time groundwater is recharged by irrigation seepage and when it returns to the river as surface flow.

Importance: The net balance of surface water in the Henry’s Fork watershed is defined very simply as outflow = inflow minus net diversion.


Since there have been no systematic changes since 1978 in watershed inflow, outflow, net diversion, or net withdrawal, calculation of average irrigation-year hydrographs of these quantities provides a meaningful picture of water balance in the Henry’s Fork and one that has remained constant since the late 1970s, despite substantial changes in irrigation practices and reservoir management over that time.

The solid blue curve is the difference between inflow and outflow, which reflects both net diversion and upstream change in storage. In particular, diverting water from the river and/or storing it in reservoirs reduces outflow. The area underneath the blue curve is water that flows into the watershed (natural flow) that is not flowing out at Rexburg. The red dashed curve is net diversion, that is, diversion minus return flow. The area under the red curve is net diversion of water from the Henry’s Fork and tributaries. When the blue curve lies above the red curve, there is less surface water leaving the watershed than can be accounted for by net diversion. That difference is water stored in the reservoirs, indicated by the blue-hatched polygons. When the red curve lies above the blue curve, there is more water being diverted than natural flow available in the system. The difference is made up by reservoir draft, as depicted by the red-hatched polygon.


It’s easy to see that on average, the reservoir system is drafted from late June to mid-September to meet irrigation needs when natural flow is insufficient to maintain diversion and leave a little water in the river at the bottom of the watershed (my “600-cfs rule”). The reservoir system fills during the remainder of the year. From analysis of raw Island Park Reservoir data alone, the 1978-2020 average timing of reservoir draft is June 23 through September 15, which aligns with the watershed-scale data depicted in the graph above.


Because net withdrawal and net diversion equal each other over whole irrigation years, the total area under the blue curve equals the total area under the red curve. That also means that the total area of the blue polygons equals the area of the red polygon. That is, reservoir storage, on average, must equal reservoir draft, because the reservoir starts each irrigation season full.


The net effect of diversion, return flow, and reservoir storage/draft on streamflow in the Henry’s Fork watershed is shown more clearly by comparing the watershed inflow and outflow graphs alone. This is shown below. Note that watershed outflow at Rexburg is always lower than inflow but that the hydrographs have roughly the same shape. Hydrograph shape is a good indicator of the ability of a stream to maintain natural ecological processes in the stream channel and floodplain. In the case of the lower Henry’s Fork, the functional stream channel and floodplain is smaller than it would be under unregulated conditions, but ecological functionality is very high there, as evidenced by a dynamic stream channel and extensive riparian forests.


Three questions remain.

1. What happens to the 830,000 ac-ft of water withdrawn from the Henry’s Fork?

Using a model of crop acreage, crop types, growing-season precipitation, and crop evapotranspiration (ET) needs (data independent of those shown so far), we estimated mean annual crop ET at 330,000 ac-ft. This is roughly equivalent to 16 inches of irrigation per year applied to around 250,000 irrigated acres in the Henry’s Fork watershed. To put this in perspective, annual ET for grain in our watershed is around 20 inches, that for alfalfa cut 2-3 times per year is around 30 inches, and that for potatoes is around 36 inches. Growing-season precipitation averages 6-10 inches, depending on location in the watershed, and that, along with soil moisture, is sufficient to grow grain without irrigation in the wetter locations of the watershed. However, it comes up as much as 30 inches short of crop needs for alfalfa and potatoes in the lowest-elevation areas of the watershed. More details on ET calculations and the watershed-scale water budget are given in a technical report accompanying the Henry's Fork Basin Study.

The remaining 500,000 ac-ft withdrawn but not used by crops leaves the watershed via groundwater flow in the Eastern Snake Plain Aquifer (ESPA). This water eventually makes its way back to the Snake River downstream of the Henry’s Fork watershed. It is likely that the distribution of withdrawn water between crop consumption and ESPA recharge has changed over the decades. Prior to irrigation efficiency conversions, the ESPA figure was probably somewhat higher than 500,000 ac-ft, and the crop ET figure was probably a little lower. Without fine-scale data dating back to the 1970s and earlier, it is difficult to estimate these differences directly.


2. If streamflow hasn’t changed since 1978, does it really matter that diversion and reach gain have decreased since 2000?

YES! Although streamflow hasn’t changed over an annual basis since 1978, the annual hydrograph of reach gain has changed substantially, as shown in the graph below.

The difference between the 1978-2000 and 2001-2020 reach gain hydrographs is greatest during the middle of the summer, when total gain is around 400 cfs less now than it was prior to 2000. August streamflow in the Henry’s Fork at Rexburg averages 1,356 cfs, so, the difference in reach gain between the two time periods is roughly 30% of the river’s total flow at Rexburg during late summer. Decreased gains leave less water in the river for irrigators on the lower Teton River whose water rights depend on reach gains rather than on storage. Furthermore, groundwater inputs are cooler than surface water in the river, so even at an equivalent flow, water temperatures are lower when a greater percentage of streamflow comes from groundwater than from surface water.


Importance: Even though neither net withdrawal from the river nor annual streamflow has changed since 1978, loss of groundwater returns has negatively affected both water users and summertime water temperatures in the lower Henry’s Fork Watershed.


3. Is there a way to increase stream reach gains without returning to flood irrigation or increasing draft of Island Park Reservoir?

YES! Managed aquifer recharge provides this mechanism, and water management at the end of the 2020 irrigation season illustrates how this can work.


Although administrative irrigation season ended on October 31, around 430 cfs of total diversion continued in some canals in early November for managed aquifer recharge. Administratively, the water being recharged is/was storage water rented by groundwater users and provided to surface water users under terms of a settlement agreement. Because of good reservoir carryover in 2019 and near-average water supply in 2020, the rented water was not needed for irrigation, and 58,330 ac-ft was donated to the Idaho Water Resource Board to be used for managed aquifer recharge. Numerous irrigation entities, including Fremont-Madison Irrigation District (FMID), began recharging this water on September 3.