top of page

Why does HFF's reported snowpack differ from others?


The primary reason for the difference is geography.



The standard snowpack products--tables and maps provided by NRCS and other agencies--report a single basin or watershed-wide average. The basin-scale figure refers to the Snake River basin between Heise and King Hill (between Twin Falls and Mountain Home), which includes a large number of SnoTel stations south of the river in small, low-elevation watersheds such as Willow Creek, Blackfoot, Portneuf, Raft, and Salmon Falls. HFF's analysis uses nine specific SnoTel stations to represent SWE in the Henry’s Fork watershed.






There are two secondary reasons why HFF's numbers differ. First, HFF reports SWE relative to the arithmetic average (aka the “mean”), whereas the default depiction on the NRCS maps is relative to the median. Second, HFF uses a reporting period that starts in 1989—the first water year in which most climate parameters were recorded at most SnoTel stations in the watershed—and ends with the most recent complete water year. The NRCS is currently using 1991–2020 as the period of record. These two records differ slightly—by less than 1% in the case of the Henry’s Fork.


These observations logically prompt two more questions: 1) why does HFF use the 1989–2023 mean vs. 1991–2020 median, and 2) won’t above-average SWE in the southern half of the Snake River basin make up for below-average SWE in the northern portion to give us average basin-wide streamflow?


The answer to the first question is that HFF uses a longer period of record to provide more context for what is currently happening. Granted the difference between 1989–2023 and 1991–2020 is very small, but in 2031, when NRCS updates the 30-year period to 2001–2030, that period will be highly skewed toward drier years, whereas HFF's period at that point will be 1989–2030, which will include the wet years of the 1990s. This difference will likely be substantial, with the 1989–2030 SWE figures being much higher. There are good reasons to use the NRCS period, namely that current data will be compared to more recent years, thereby reflecting the current climate as it shifts towards warmer, drier years. Apples to apples, in some sense. On the other hand, keeping the colder, wetter years of the 1990s in the record will emphasize the warmer, drier trend as it is happening. After all, our water allocation system and irrigation storage reservoirs and other infrastructure were established prior to the mid-20th century, and our expectations for streamflows and fishing conditions were generally set in the mid-20th century. Thus, HFF chose to compare current conditions to a period that better reflects the climate when the irrigation system and our expectations for water supply were established.


The use of the mean vs. median somewhat reflects a bias toward depicting the water supply more conservatively (on the low side) but also a mathematician’s viewpoint. The median is a rank-based measure, namely the observation in the middle of all of the years in the record. In this sense, the median is the “average year” or “normal year.” The mean is the arithmetic average—add up the numbers for all years and divide by the number of years. In water-year measures such as precipitation, streamflow, and SWE, the mean is larger than the median because a few very wet years pull this average up, even though the majority of years are somewhat drier. A fair comparison to the median year is the rank position or percentile of the current year. For example, current water-year precipitation ranks 14th out of the last 36 years, at the 63rd percentile. By that measure, current precipitation is above normal—higher values have historically occurred in less than 50% of all years. On the other hand, current water-year precipitation is 97% of average, which is below normal by this measure of “normal”. Both of these statistics are correct, true, and mathematically consistent. The one that is less so is percent of median, which would be 103% in this case. This measure mixes the mathematical operation of division with a rank-based statistic, producing something that doesn’t exist within strict mathematical analysis. So, HFF chose to report both rank and percent of average rather than mix the two, producing more information and context.


To answer to the second question, around 75% of the total water supply in the upper Snake River basin (remember, that huge basin that extends all the way downstream to King Hill) comes from the Henry’s Fork and South Fork watersheds, combined. All of the other watersheds in the basin—including the southside watersheds such as the Blackfoot and Portneuf and the northside watersheds such as the Big Lost and Wood rivers—provide the other 25%. The distribution of SnoTel stations is not in proportion to where the water actually comes from. So, for example, if SWE is 85% of average in the Henry’s Fork and South Fork watersheds and 125% of average everywhere else, the total water supply would be 95% of average, not 105% of average, as would be obtained by averaging the 85% and the 125% with equal weight.


HFF addresses this issue by using 12 weather stations that accurately reflect water supply in the Henry’s Fork watershed. Essentially, water supply in the Henry’s Fork is proportional to elevation, so HFF uses the 12 stations whose elevations average to the average elevation of the whole watershed. This is why there can be such a disparity between precipitation in the valley areas and total watershed precipitation. If the valleys receive much heavier precipitation relative to average than the higher-elevation locations, because such a small fraction of our total water supply originates in the valleys, the watershed mean will still be below average, accurately reflecting water supply.

104 views

Comments


bottom of page