search

vicelab / cen-sierra-pywr

Code base for modeling the central Sierra Nevada hydropower systems
4 stars 5 forks source link

Update SJVI to account for actual calculation #6

Open drheinheimer opened 4 years ago

drheinheimer commented 4 years ago

Alan just reminded me of a potentially significant inaccuracy in our SJVI-dependant IFR calculations. We assume perfect knowledge of SJVI starting in October, which is of course incorrect. While I don't think we'll ever be too far off, it might be worth updating our SJVI calculation accordingly. I.e., we could use the actual formula, even if we assume perfect hydrologic foresight in the calculation when some forecasting is needed.

This inaccuracy is attenuated somewhat by both the formula itself, and also by the fact that the larger IFR volumes are going to be later in the year, when presumably our assumption of perfect foresight is not too wrong.

Either way, we need to either justify our current approach or revise our SJVI calculation scheme.

alanccai commented 4 years ago

Here is how DWR describes the timing of SJV WY Index calculations, taken from http://cdec.water.ca.gov/reportapp/javareports?name=WSIHIST:

alanccai commented 4 years ago

For Water Years 2011- 2019, the first of month forecasts are available here: http://cdec.water.ca.gov/cgi-progs/previous/WSI.

drheinheimer commented 4 years ago

Good. Do operators go with a standard DWR calculated exceedance probability when making realtime IFR releases? We'll need a way to approximate real-time (monthly) SJVI calculation that accounts for whatever the typical operational criteria is.

alanccai commented 4 years ago

Based on my scanning of the IFR language, here's how it breaks down.

(1) Some operators make a decision based not on the SJVI index, but on projected inflows at their specific location (ex. into New Melones). This projection is presented as a single number, and seems to be the 50% exceedence probability, aka the median value of what probability distribution there is for that value.

Then there is a simple classification based on ranges of values for this single number. (No exceedance probabilities involved). For example, for the Donnells Reach of Stanislaus River, if the value is between 350K and 676K acre-feet, then it's classified as Dry. Note that this classification has nothing to do with the SJVI.

Complicating factors: (a) depending on the IFR, sometimes this "projected inflows" value means the projected inflow values for the entire water year. Sometimes it's just for April - July. The CDWR forecast issued in Feb, March, April, and May (known as DWR Bulletin 120 - https://cdec.water.ca.gov/snow/bulletin120/) offers values for both. (b) Some IFRs require a new classification to be made each time a new Bulletin comes out from Feb to May, while others just rely on the May one...

(2) Some IFRs use the SJVI to make their classification. In this case, the DWR Bulletin 120 only present the classification based on the 75% exceedence value, as the procedure calls for. The values for other exceedence values are not included in the Bulletin.

drheinheimer commented 4 years ago

I think this issue relates to how the SJVI is calculated, rather than how it is used. I.e., the Apr-Jul forecast is separate from SJVI. Or not? I mean, is Apr-Jul also part of the SJVI?

For the actual calculation, I think we can do a 75% exceedance probability based off of snowpack, pending confirmation from DWR.

alanccai commented 4 years ago

San Joaquin Valley Water Year Index = 0.6 Current Apr-Jul Runoff Forecast (in maf) + 0.2 Current Oct-Mar Runoff in (maf) + 0.2 * Previous Water Year's Index (if the Previous Water Year's Index exceeds 4.5, then 4.5 is used)

Yes, forecasted Apr-Jul is also part of the SJVI.

alanccai commented 4 years ago

Relevant document:

DWR California Cooperative Surveys Forecasting Methodologies.pdf