Assessment of the San Joaquin Cloud Seeding Program using WRF based Sub Basin Quantitative Precipitation Forecasts as Formal Controls for Unseeded Precipitation: Preliminary Results obtained from the 2009 and 2010 Winter Seasons
Traditional assessment procedures involve comparisons of “seeded” precipitation that falls in a target area with “non-seeded” precipitation that falls in control areas. Changes in the amounts of precipitation that falls in the target area are compared with precipitation that falls in the control areas to establish changes in precipitation attributed to seeding. This situation leads to statistical designs that require large sample sizes to identify an often “small seeding signal” embedded within a larger noisy natural precipitation signal to prove that seeding has had some effect on the amount of precipitation that falls in the target area. Although the problem of natural variability of precipitation cannot be side-stepped, it is feasible to minimize the problem by identifying in precipitation, specific physical or chemical parameters connected with the seeding process and subsequently use these as a means of identifying those portions of the precipitation involved in the seeding process. The focus of the assessment is sharpened by eliminating from the analysis non-seeded periods, thereby reducing the uncertainty in the seeding assessment process.
The San Joaquin watershed is located about 100 km northeast of the Hanford KHNX Radar. The target ranges in elevation from about 2000 to 4000 m. KHNX has a fairly unobstructed view of the windward side of the target area, west of Kaiser Ridge, but as with most Sierra watersheds clutter filtering and/or terrain blocking prevent the radar from viewing the higher elevations of the target area. This prevents development of meaningful radar based QPE's using KHNX, although KHNX frequently paints the upper portions of clouds passing over the target area so it can be used to provide temporal estimates of frontal passage or times other higher clouds pass over the watershed.
For those portions of the target located on the windward side of Kaiser Ridge quantitative precipitation estimates (QPE's) are formed by combining radar (KHNX) and precipitation gauge data. These methods have been described by Clarke and Eilts, 2007.
To measure precipitation in the higher elevations of the target area a combination of snow course, streamflow and precipitation gauge data are combined to produce sub-basin QPE's. Most of the modern snow course sites in the Upper San Joaquin basin and streamflow measurements were established before 1913 when the Big Creek project came into operation and are still maintained today. This combination of datasets form the QPE database for 2009 and 2010 winter seasons in areas of the target that are not visible to KHNX.
Traditional streamflow assessments of the San Joaquin program have continuously improved by using more sophisticated physical and statistical analysis techniques as they become available. The most recent assessments have shown statistically significant increases in annual streamflow of 5-7% using ratio regression techniques. They are based on the cumulative effects of seeding the target area for over 50 years (Farber, et. al., 2010, Silverman, 2008). However, these methods offer no temporal insight about when and where seeding is having its greatest impacts.
For these reasons we have chosen to use a local version of the WRF model coupled with the other databases previously mentioned to produce quantitative precipitation forecasts (QPF's) as formal non-seeded precipitation controls. HydroWRF was developed in April 2008 and is based on version 3.0.1 of WRF-ARW (Shaw, 2008). It was designed to provide additional high resolution forecast tools for the operational program and runs continuously producing a 36 hour forecast updated every six hours. The spatial resolution of the model over the project area covers a 100 x 100 domain with a 1 km grid spacing. No special seeding modules have been added to the code so by running the model continuously we develop an unseeded QPF control database. Modifications to the code were added to allow six hour precipitation amounts and other forecast response variables to be routinely recorded and displayed for use by the project.
The San Joaquin program is operational and any opportunity that meets project seeding criteria is seeded; no randomization scheme is employed. However, by running HydroWRF continuously an expanding control data set is generated. To refine the QPF's and QPE's for the project area, the watershed has been subdivided into sub-basins. HydroWRF uses bulk microphysics routines to conserve computational resources, however, potential biases in the target control sub-basin relationships exist because of their use, particularly under southerly airflow regimes.
QPF's for the backcountry are formed in the same manner as they are over the entire model domain. Comparisons of QPE's and QPF's west of Kaiser Ridge are used to establish the viability of using the same method to produce QPF's in the backcountry as they are on the windward sides of the target area.
For the purposes of this paper we have defined fixed target and control basins, although a more sophisticated approach would use a floating target design. This paper will present the concepts involved in making a preliminary assessment of the effects of seeding the Upper San Joaquin during 2009 and 2010 winter seasons..
Comparisons of sub-basin QPF's and radar derived QPE's will be made for portions of the target area and control areas upwind of Kaiser Ridge and visible to KHNX. For those portions of the target area not visible to KHNX, comparisons of modeled sub-basin QPE's with QPF's using the methods described herein will be presented for both seeded and unseeded periods. The results of these preliminary analyses will be compared and contrasted with results previously found using traditional streamflow assessments.