• Bauer, D. F., 1972: Constructing confidence sets using rank statistics. J. Amer. Stat. Assoc., 67, 687690.

  • Bruintjes, R. T., 1999: A review of cloud seeding experiments to enhance precipitation and some new prospects. Bull. Amer. Meteor. Soc., 80, 805820.

    • Search Google Scholar
    • Export Citation
  • Chai, S. K., , W. G. Finnegan, , and R. L. Pitter, 1993: An interpretation of the mechanism of ice-crystal formation operative in the Lake Almanor cloud-seeding program. J. Appl. Meteor., 32, 17261732.

    • Search Google Scholar
    • Export Citation
  • Cotton, W. R., , and R. A. Pielke, 2006: Human Impacts on Weather and Climate. 2nd ed. Cambridge University Press, 332 pp.

  • Dennis, A. S., 1980: Weather Modification by Cloud Seeding. Academic, 275 pp.

  • Deshler, T., , D. W. Reynolds, , and A. W. Huggins, 1990: Physical response of winter orographic clouds over the Sierra Nevada to airborne seeding using dry ice or silver iodide. J. Appl. Meteor., 29, 288330.

    • Search Google Scholar
    • Export Citation
  • Efron, B., 1982: The Jackknife, the Bootstrap and Other Resampling Plans. Capital City, 93 pp.

  • Heimbach, J. A., , and A. B. Super, 1996: Simulating the influence of type II error on the outcome of past statistical experiments. J. Appl. Meteor., 35, 15511567.

    • Search Google Scholar
    • Export Citation
  • Huggins, A. W., 1995: Mobile microwave radiometer observations: Spatial characteristics of supercooled cloud water and cloud seeding implications. J. Appl. Meteor., 34, 432446.

    • Search Google Scholar
    • Export Citation
  • Huggins, A. W., , S. L. Kenyon, , L. Warren, , A. D. Peace, , S. P. Bilish, , and M. J. Manton, 2008: The Snowy Precipitation Enhancement Research Project: A description and preliminary results. J. Wea. Modif., 40, 2851.

    • Search Google Scholar
    • Export Citation
  • Long, A. B., , and A. W. Huggins, 1992: Australian Winter Storms Experiment (AWSE) I: supercooled liquid water and precipitation-enhancement opportunities. J. Appl. Meteor., 31, 10411055.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., , and L. Warren, 2011: A confirmatory snowfall enhancement project in the Snowy Mountains of Australia. Part II: Primary and associated analyses. J. Appl. Meteor. Climatol., 50, 14481458.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., , J. Dai, , and L. Warren, 2009: Evaluation plan for a snow enhancement experiment in Australia. J. Wea. Modif., 41, 5974.

    • Search Google Scholar
    • Export Citation
  • Mielke, P. W., , K. J. Berry, , and J. G. Medina, 1982: Climax I and II: Distortion resistant residual analyses. J. Appl. Meteor., 21, 788792.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , M. J. Manton, , and A. Nazarov, 2009: On the analysis of a cloud seeding dataset over Tasmania. J. Appl. Meteor. Climatol., 48, 12671280.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , M. J. Manton, , and A. Nazarov, 2010: A modeling case study of mixed-phase clouds over the Southern Ocean and Tasmania. Mon. Wea. Rev., 138, 839862.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , and M. J. Manton, 2011: A three-year climatology of cloud-top phase over the Southern Ocean and North Pacific. J. Climate, 24, 24052418.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 2003: Critical Issues in Weather Modification Research. National Academies Press, 144 pp.

  • Rauber, R. M., , R. D. Elliott, , J. O. Rhea, , A. W. Huggins, , and D. W. Reynolds, 1988: A diagnostic technique for targeting during airborne seeding experiments in wintertime storms over the Sierra Nevada. J. Appl. Meteor., 27, 811828.

    • Search Google Scholar
    • Export Citation
  • Reynolds, D. W., , and A. S. Dennis, 1986: A review of the Sierra Cooperative Pilot Project. Bull. Amer. Meteor. Soc., 67, 513523.

  • Ristovski, Z., , and R. Jayaratne, 2005: Measurement of the size distribution of cloud seeder (AgI) and tracer (InO2) aerosols produced by an LPG burner. International Laboratory for Air Quality and Health, Queensland University of Technology, 22 pp.

    • Search Google Scholar
    • Export Citation
  • Ryan, B. F., , and W. D. King, 1997: A critical review of the Australian experience in cloud seeding. Bull. Amer. Meteor. Soc., 78, 239254.

    • Search Google Scholar
    • Export Citation
  • Smith, E. J., , E. E. Adderley, , and D. T. Walsh, 1963: A cloud-seeding experiment in the Snowy Mountains, Australia. J. Appl. Meteor., 2, 324332.

    • Search Google Scholar
    • Export Citation
  • Smith, E. J., , L. G. Veitch, , D. E. Shaw, , and A. J. Miller, 1979: A cloud-seeding experiment in Tasmania. J. Appl. Meteor., 18, 804815.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., 1986: Further exploratory analysis of the Bridger Range winter cloud seeding experiment. J. Climate Appl. Meteor., 25, 19261933.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., , and J. A. Heimbach, 1983: Evaluation of the Bridger Range winter cloud seeding experiment using control gages. J. Climate Appl. Meteor., 22, 19892011.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., , and J. A. Heimbach, 1988: Microphysical effects of wintertime cloud seeding with silver iodide over the Rocky Mountains. Part II: Observations over the Bridger Range, Montana. J. Appl. Meteor., 27, 11521165.

    • Search Google Scholar
    • Export Citation
  • Twomey, S., , and I. Robertson, 1973: Numerical simulation of cloud seeding experiments in selected Australian areas. J. Appl. Meteor., 12, 473478.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , and M. A. Wetzel, 1992: Field study of the potential for winter precipitation enhancement in the Australian Snowy Mountains. Atmos. Res., 28, 327363.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , R. H. Stone, , and B. L. Marler, 1995a: How the transport and dispersion of AgI aerosols may affect detectability of seeding effects by statistical methods. J. Appl. Meteor., 34, 19291941.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , L. G. Young, , and R. H. Stone, 1995b: Assessment of seeding effects in snowpack augmentation programs: Ice nucleation and scavenging of seeding aerosols. J. Appl. Meteor., 34, 121130.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Map of southeastern Australia showing locations of cloud seeding areas for Tasmania (circle), Baw Baw Plateau (triangle), and the Snowy Mountains (square); contours are at 500, 1000, and 1500 m.

  • View in gallery

    Map of the Snowy Mountains showing generator sites (black diamonds) against the orography, with contours at 500, 1000, and 1500 m; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; upper-air soundings are taken at Khancoban (blue dot); supercooled liquid water is measured at Blue Cow (red dot); the distance between Khancoban and Blue Cow is 29 km.

  • View in gallery

    Mean sea level pressure chart at 2200 LT 18 Jul 2008 during a seeding campaign in which 5 EUs were carried out (from the Australian Bureau of Meteorology).

  • View in gallery

    Map of the SPERP area showing the plume of seeding material from a generator at Khancoban on 17 Jul 2008 calculated from the GUIDE model; plus sign indicates where nucleation of a particle occurs and an asterisk indicates fallout of the particle in the primary target area (red line); tick marks represent 1 km.

  • View in gallery

    Cross section of the trajectory of a seeding particle emitted from a generator on 17 Jul 2008 at Khancoban calculated from the GUIDE model (within the plume shown in Fig. 4); the times sign indicates where the particle is nucleated.

  • View in gallery

    Location of precipitation gauges (black dots) used for SPERP analysis; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; Khancoban is denoted by a blue diamond and Blue Cow is shown by a red diamond; the distance between Khancoban and Blue Cow is 29 km; background colors show topography, with contours at 500, 1000, and 1500 m.

  • View in gallery

    Histograms of precipitation in primary target, overall target, control, and extended areas over all EUs.

  • View in gallery

    Correlation of precipitation at each site across all EUs used in the SPERP analysis with the mean precipitation in the control area; sites are indicated by black dots; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; contours are at 0.1 intervals.

  • View in gallery

    Histograms of wind (top) speed and (bottom) direction at the −5°C level at the start of EUs.

  • View in gallery

    Histograms of (top) the freezing level and (bottom) the height of −5°C level at the start of EUs.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 86 86 26
PDF Downloads 54 54 14

A Confirmatory Snowfall Enhancement Project in the Snowy Mountains of Australia. Part I: Project Design and Response Variables

View More View Less
  • 1 Monash University, Melbourne, Victoria, Australia
  • | 2 Snowy Hydro Limited, Sydney, New South Wales, Australia
  • | 3 Snowy Hydro Limited, Cooma, New South Wales, Australia
  • | 4 Snowy Hydro Limited, Sydney, New South Wales, Australia
© Get Permissions
Full access

Abstract

The Snowy Precipitation Enhancement Research Project (SPERP) was undertaken from May 2005 to June 2009 in the Snowy Mountains of southeastern Australia with the aim of enhancing snowfall in westerly flows associated with winter cold fronts. Building on earlier field studies in the region, SPERP was developed as a confirmatory experiment of glaciogenic static seeding using a silver-chloroiodide material dispersed from ground-based generators. Seeding of 5-h experimental units (EUs) was randomized with a seeding ratio of 2:1. A total of 107 EUs were undertaken at suitable times, based on surface and upper-air observations. Indium (III) oxide was released during all EUs for comparison of indium and silver concentrations in snow in seeded and unseeded EUs to test the targeting of seeding material. A network of gauges was deployed at 44 sites across the region to detect whether precipitation was enhanced in a fixed target area of 832 km2, using observations from a fixed control area to estimate the natural precipitation in the target. Additional measurements included integrated supercooled liquid water at a site in the target area and upper-air data from a site upwind of the target.

Corresponding author address: Michael Manton, School of Mathematical Sciences, Monash University, Melbourne, VIC 3800, Australia. E-mail: michael.manton@monash.edu

Abstract

The Snowy Precipitation Enhancement Research Project (SPERP) was undertaken from May 2005 to June 2009 in the Snowy Mountains of southeastern Australia with the aim of enhancing snowfall in westerly flows associated with winter cold fronts. Building on earlier field studies in the region, SPERP was developed as a confirmatory experiment of glaciogenic static seeding using a silver-chloroiodide material dispersed from ground-based generators. Seeding of 5-h experimental units (EUs) was randomized with a seeding ratio of 2:1. A total of 107 EUs were undertaken at suitable times, based on surface and upper-air observations. Indium (III) oxide was released during all EUs for comparison of indium and silver concentrations in snow in seeded and unseeded EUs to test the targeting of seeding material. A network of gauges was deployed at 44 sites across the region to detect whether precipitation was enhanced in a fixed target area of 832 km2, using observations from a fixed control area to estimate the natural precipitation in the target. Additional measurements included integrated supercooled liquid water at a site in the target area and upper-air data from a site upwind of the target.

Corresponding author address: Michael Manton, School of Mathematical Sciences, Monash University, Melbourne, VIC 3800, Australia. E-mail: michael.manton@monash.edu

1. Introduction

Since the mid-1940s, Australia has made significant investments in cloud seeding experiments. A comprehensive review of this research by Ryan and King (1997) noted that two separate experiments in Tasmania (Fig. 1) in the 1960s and 1970s showed statistically significant increases in rainfall when clouds in a southwesterly stream off the Southern Ocean are lifted by the (relatively low) mountains on the west coast of the island. On the basis of those results, operational cloud seeding has been undertaken in Tasmania since that time.

Fig. 1.
Fig. 1.

Map of southeastern Australia showing locations of cloud seeding areas for Tasmania (circle), Baw Baw Plateau (triangle), and the Snowy Mountains (square); contours are at 500, 1000, and 1500 m.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

Using both experimental and operational cloud seeding data covering a 46-yr period from that program, Morrison et al. (2009) find statistically significant increases in monthly rainfall on the order of 5%–10%. Analysis of aircraft data and model simulations of the passage of cold fronts across Tasmania by Morrison et al. (2010) also show some physical evidence that significant amounts of supercooled liquid water (SLW) occur in those weather systems, implying that there are regular and physically based conditions for glaciogenic cloud seeding.

We show below that the cold fronts studied in Tasmania are similar to or the same as those that pass over the Great Dividing Range of the mainland of southeastern Australia. Long and Huggins (1992) report observations of substantial quantities of SLW associated with fronts passing over the Baw Baw Plateau in the southern parts of the range (Fig. 1). That research was part of a cloud seeding project with results that were not published. Nevertheless, Ryan and King (1997) note that, while the results of the primary analysis were not statistically significant, analysis of precipitation in the buffer area between the target and control showed a statistically significant increase.

The Snowy Mountains region of New South Wales lies north of the Baw Baw Plateau (Fig. 1). Smith et al. (1963) investigated cloud seeding potential over the area during the late 1950s. While the results of that experiment were considered to be controversial, the evaluation analysis reported a statistically significant increase in precipitation of 19% when potential seeding events extended over several days during March–December. That research was followed by several more studies sponsored by the Snowy Mountains Hydro-Electric Authority (SMHEA) (e.g., Warburton and Wetzel 1992). In 2003 Snowy Hydro Limited (the corporatized SMHEA) proposed a further field trial, and the Snowy Precipitation Enhancement Research Project (SPERP) subsequently commenced in 2004 as a confirmatory experiment of glaciogenic static seeding using silver iodide from ground-based generators. The design of SPERP was based on knowledge gained from the many cloud seeding experiments carried out over the last five decades, and on exploratory studies carried out in the Snowy Mountains. One difference between SPERP and the earlier experiments in southeastern Australia is that seeding material is dispersed from ground-based generators rather than from aircraft.

A summary of the historical record of cloud seeding, including glaciogenic static seeding, is given by Bruintjes (1999), with an update by the U.S. National Research Council (2003). A number of these experiments are especially relevant to the design of SPERP. Reynolds and Dennis (1986) describe the Sierra Cooperative Pilot Project (SCPP), in which glaciogenic seeding was used to enhance precipitation from winter storms passing over the barrier range of the Sierra Nevada extending about 2000 m MSL. A primary conclusion of SCPP is that the best opportunities for seeding arose from widespread orographic cloud with tops warmer than about −15°C. Moreover, Deshler et al. (1990) provide observational evidence of consistent microphysical changes in some cases where seeding effects were detected at the surface.

The results of an exploratory randomized glaciogenic seeding experiment in the Bridger Range of Montana are evaluated by Super and Heimbach (1983). Statistical analysis suggests that precipitation was enhanced on days when the seeding plume temperature (estimated by the temperature at the mountain ridge) was cooler than about −9°C. Further analysis by Super (1986) finds that the effects of seeding are greatest for cloud-top temperatures greater than −25°C, that is, somewhat colder than the lower limit suggested by Reynolds and Dennis (1986) from SCPP. The statistical results from the Bridger Range Experiment, which indicated that about 15% more precipitation fell on seeded than unseeded days, are supported by aircraft observations that show the dispersion of seeding material across the target area and the transformation of SLW to ice particles in the target area (Super and Heimbach 1988). The apparent differences in the optimal conditions for seeding between SCPP and the Bridger Range Experiment may be due to differences in the detailed microphysics of the clouds.

From the earlier cloud seeding experiments, a reasonable body of evidence has accumulated that, at least on some occasions, the dispersion of silver iodide into orographically modified cloud containing SLW can enhance precipitation at the surface through the transformation of SLW into ice particles. Further evidence of microphysical changes can be deduced from snow chemistry measurements through the dispersion of a passive tracer [indium (III) oxide] in addition to silver iodide (Chai et al. 1993). The concentration of indium (In) in the snow represents the effects of scavenging of seeding material, and so (with similar mass release rates) any greater concentrations of silver (Ag) indicate that the seeding material has promoted ice crystal formation through either contact nucleation or condensation freezing. Ratios of Ag:In of up to 17 have been found in snow impacted by cloud seeding. Similar studies by Warburton et al. (1995b) provide further support for the strategy of using high values of Ag:In in snow to indicate that silver iodide has affected the microphysical processes in the cloud from which the snow was formed.

Using the experience from the earlier studies in the United States, Warburton and Wetzel (1992) carried out an observational program in the Snowy Mountains of southeastern Australia to assess the potential for precipitation enhancement in that region based on glaciogenic static seeding from ground-based generators. They found that during the passage of winter storms integrated paths of SLW of between 0.2 and 0.4 mm are not uncommon, with the peak values of SLW occurring when the cloud-top temperature is around −9°C. Moreover, isotopic studies of the snow indicate that snow formation generally occurs at temperatures higher than −15°C. In addition to ensuring that the microphysical conditions are suitable for glaciogenic cloud seeding, Warburton and Wetzel use the GUIDE dispersion model (Rauber et al. 1988) to demonstrate that seeding material from ground-based generators to the west of the mountain range should be able to convert the SLW in the cloud to ice particles, which in turn should fall as snow on the range.

The observational studies by Warburton and Wetzel (1992) are generally consistent with the results of Long and Huggins (1992), who find significant fluxes of SLW following the passage of winter cold fronts near Baw Baw to the south of the Snowy Mountains (Fig. 1). Similarly, Ryan and King (1997) and Morrison et al. (2010) both report high concentrations of SLW in cloud in Tasmania where seeding operations are routinely carried out. An indication of the commonality of the synoptic conditions associated with cloud seeding across southeastern Australia is given by the fact that one of the case studies used by Morrison et al. in Tasmania (9 August 2006) corresponds to the date of an experimental campaign in SPERP during the passage of a winter front, associated with a low pressure system to the south of Tasmania. (The other case study of Morrison et al. was in October, which is outside the season for seeding during SPERP.) Further evidence of similar cloud properties in Tasmania, Baw Baw, and Snowy is given by Morrison et al. (2011), who use Moderate Resolution Imaging Spectroradiometer (MODIS) data to show consistent occurrences of SLW across these regions in winter. Nevertheless, similarity in the synoptic-scale forcing, in the local orographic forcing, and in the frequency of occurrence of SLW does not guarantee similarity in all microphysical properties.

This paper summarizes the design and implementation of the SPERP, which are detailed in Huggins et al. (2008). To ensure transparency in the evaluation of the experiment, the formal evaluation plan for SPERP (Manton et al. 2009) was published before the operational phase ended. Moreover, a preliminary analysis of the data collected over the operational phase was carried out before the seeding sequence was revealed. This part of this two-part paper describes the evaluation procedures for SPERP, as well as the results of the preliminary analysis. Manton and Warren (2011, hereinafter Part II) gives the results of the evaluation of the project after the seeding sequence was revealed.

2. Design and implementation

a. Seeding strategy

The objectives of the SPERP are to determine the technical, economic, and environmental feasibility of enhancing snowfall over the Snowy Mountains region (Fig. 1) during the winter season from May through September; this paper is focused on the technical aspects of the overall program. A confirmatory randomized cloud seeding experiment has been designed to demonstrate that glaciogenic static seeding from ground-based generators can enhance winter precipitation in a target area centered on the highest terrain. The results of earlier experimentation (Smith et al. 1963; Warburton and Wetzel 1992) suggest that silver iodide (AgI) released to the west of the mountain ridge should disperse into clouds associated with winter fronts containing SLW. Under the appropriate conditions, the AgI can transform the SLW into ice particles, which fall as enhanced snowfall on the mountain target area.

The SPERP has a fixed target–control design (Dennis 1980), with a primary target area lying along the high ridges (elevation from 1400 to 2200 m) where snow is the predominant form of precipitation in winter, a secondary target extending to the east and west of the primary target (elevation from 400 to 2000 m), and a control area to the west and north of the target (Fig. 2). The primary and secondary target areas together form the overall target with an area of 832 km2. The control area is selected to provide precipitation measurements unaffected by seeding but well correlated with that of the target. There is a focus on the primary target as that is the area where snow is required to fall during seeding conditions (section 2c).

Fig. 2.
Fig. 2.

Map of the Snowy Mountains showing generator sites (black diamonds) against the orography, with contours at 500, 1000, and 1500 m; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; upper-air soundings are taken at Khancoban (blue dot); supercooled liquid water is measured at Blue Cow (red dot); the distance between Khancoban and Blue Cow is 29 km.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

When the environmental conditions (section 2c) are suitable for seeding, an experimental unit (EU) is commenced and it is expected to run for 5 h. Analysis of historical data (section 3a) shows that about 100 five-hour EUs should be obtained over a 5-yr period. The selection of a relatively short period for an EU is consistent with the observations of Super (1986) that the impacts of seeding tend to be limited to short periods of time. Seeding is randomized with a seeding ratio of 2:1, so that precipitation in unseeded EUs can be used to estimate the “natural” precipitation in the target area in seeded EUs. The seeding sequence has been prepared by an independent authority, and the seeding decision is only disclosed to the operators of the seeding generators immediately before the commencement of each EU. The decision is not communicated to any personnel involved with the scientific aspects (such as the decision to start an EU) of SPERP.

b. Seeding material

The seeding material is dispersed from 13 generator sites located to the west of the primary target area (Fig. 2) at altitudes from 438 to 1661 m, with nine above 800 m. The distance between the generators and a central site in the target area (Blue Cow, also known locally as Blue Calf) varies from 14 to 36 km. For a wind speed at the −5°C level of 10 m s−1, 30 min would be required for material to travel 20 km from a generator to Blue Cow; this is consistent with travel times of 15–60 min, which are typically used for ground-based generators in mountainous areas in the United States (Dennis 1980). We see from Fig. 2 that, although the operation of the generators is very carefully controlled, it is not impossible for seeding material to drift into the control area. In the unlikely situation that precipitation in the control area is affected (i.e., increased) by seeding, the net effect would be to reduce rather than increase the estimated impacts of seeding in the target area.

Two generators are located at each site: one disperses the seeding material and the other disperses indium (III) oxide as a passive tracer with physical properties similar to those of the seeding material. The seeding material is AgCl0.22I0.78 · 0.5NaCl (hereinafter referred to as AgI or silver iodide), which has an activity of 1.2 × 1014 nuclei per gram at −10°C and 1012 nuclei per gram at −6°C (Huggins et al. 2008). Laboratory tests (Ristovski and Jayaratne 2005) have confirmed that the tracer and seeding particles have similar size distributions, with a median diameter of about 70 nm. The mass release rate of elemental silver (Ag) and indium (In) are both about 9.4 g h−1. The mass release rate of AgI is 20.4 g h−1 and of indium (III) oxide is 11.4 g h−1.

c. Seeding start and suspension criteria

Conditions suitable for seeding are generally associated with the passage of winter cold fronts from the west; for example, Fig. 3 shows the mean sea level chart during a seeding campaign in which five EUs were declared over the period 17–19 July 2008 during the passage of a weak front. Soundings are taken every 3 h from Khancoban, which is situated west of the target area (Fig. 2). Data from the soundings are used to drive the GUIDE dispersion model (Rauber et al. 1988) to estimate the development of plumes of seeding material from the generators and the development of ice particles nucleated by the seeding material. Measurements from earlier field experiments in the Snowy Mountains and snow chemistry results from trials in 2004 have been used to optimize the parameters in GUIDE for conditions in the Snowy Mountains (Warburton and Wetzel 1992; Huggins et al. 2008). Figure 4 shows the plume calculated from GUIDE for seeding material from a generator at Khancoban during an event on 17 July 2008, with a plus sign marking the point where nucleation occurs and an asterisk indicating the fallout of the nucleated particle in the primary target area. A cross-sectional view of the particle trajectory is shown in Fig. 5. The field studies of Huggins et al. show evidence of the ratio of silver to indium being larger than one in snow samples in the target area during seeding periods, indicating that silver iodide has played a role in the microphysical development of the snow in the target area (Chai et al. 1993).

Fig. 3.
Fig. 3.

Mean sea level pressure chart at 2200 LT 18 Jul 2008 during a seeding campaign in which 5 EUs were carried out (from the Australian Bureau of Meteorology).

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

Fig. 4.
Fig. 4.

Map of the SPERP area showing the plume of seeding material from a generator at Khancoban on 17 Jul 2008 calculated from the GUIDE model; plus sign indicates where nucleation of a particle occurs and an asterisk indicates fallout of the particle in the primary target area (red line); tick marks represent 1 km.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

Fig. 5.
Fig. 5.

Cross section of the trajectory of a seeding particle emitted from a generator on 17 Jul 2008 at Khancoban calculated from the GUIDE model (within the plume shown in Fig. 4); the times sign indicates where the particle is nucleated.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

An EU can commence only if seeding material from at least one generator will pass over the target and ice particles formed from the seeding material will fall into the target area. These conditions on trajectories mean that the wind must have a westerly component and the wind speed must allow time for ice particles to form and fall in the target. Further conditions are that the cloud-top temperature must be less than −7°C and that there must be at least 400 m of cloud above the −5°C level. A dual-channel radiometer is located at Blue Cow (Fig. 2) to measure integrated liquid water and water vapor (Huggins 1995; Huggins et al. 2008), and at least 0.05 mm of SLW averaged over 0.5 h must be available before seeding can commence. [In the event of instrument failure, this condition may be met if ice trips are recorded on a Goodrich (Model 0871LH1) icing rate detector.] A government-legislated condition for seeding to commence is that precipitation must fall as snow in the primary target area, and so the freezing level must be 1600 m or lower. A final condition is that the scientific controller must be confident that suitable conditions will continue for at least 3 h.

When the conditions for seeding are satisfied, an EU commences and targeting generators operate for 5 h unless specified suspension criteria are reached. Operations must be suspended if the freezing level rises above 1600 m, if the reservoir storage becomes excessive, or if relevant severe weather is declared in the area by the Bureau of Meteorology. Each generator may be switched on or off as new sounding information is fed into the GUIDE model during an EU; that is, the total number of generator hours for an EU varies with the prevailing meteorological conditions.

To account for the travel and microphysical response times of the seeding material, the 5-h evaluation period for an EU commences 1 h after the generators are first switched on, and so it extends for 1 h after all the generators are switched off. The GUIDE model is used to estimate the purge time required for all seeding material to be transported out of the target area after the generators are switched off. A minimum purge time of 1 h is used, so that there can be no overlap between the end of the evaluation period of one EU and the commencement of generators for the next.

d. Observing systems

The main observations for SPERP are precipitation measurements at 44 sites across the region of interest (Fig. 6). There are 8 sites in the primary target area, 8 in the secondary target, 12 in the control area, and 16 in the extended area downwind of the target and control areas.

Fig. 6.
Fig. 6.

Location of precipitation gauges (black dots) used for SPERP analysis; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; Khancoban is denoted by a blue diamond and Blue Cow is shown by a red diamond; the distance between Khancoban and Blue Cow is 29 km; background colors show topography, with contours at 500, 1000, and 1500 m.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

In the early years of SPERP, it became clear that the measurement of snow in winter storms requires considerable care because of the problem of undercatch in high-wind conditions at exposed high-elevation sites. To address this issue, half-size (6 m) wind fences (double fence intercomparison references, or DFIRs) were installed around 11 ETI Instrument Systems (Model NOAH II) gauges in the observing network. A full-sized DFIR was operated at one site along with a number of other gauges. A further 12 ETI Instrument Systems gauges were located at unfenced sites in the ranges. Australian Hydrological Services tipping-bucket gauges (Model TB3) were used at sites below the snowline. The efforts to improve the quality of the precipitation measurements resulted in 62 instruments located at 44 sites. There were 10 sites (5 in the target, 2 in the control, and 3 in the extended area) that were not operational at the start of SPERP, and they were commissioned over 2005 and 2006. All precipitation data were quality controlled and archived by Snowy Hydro Limited, along with associated metadata; in particular, great care was taken to remove false tips from the ETI gauges.

Snow chemistry observations of Ag and In are used to verify the fallout of seeded precipitation in the target area by delineating seeded from unseeded events, and to test the microphysical hypothesis that silver iodide particles are acting as ice nuclei using the method of Chai et al. (1993). Snow samples for trace chemistry analysis are collected using two separate methods. Real-time snow samples are collected during seeding campaigns at Blue Cow (Fig. 2). Snow profile samples are also collected from 16 sites across the study area (14 in the target, and 1 each in the control and extended areas) after each major experimental campaign. Each sample represents a 20-mm-depth increment of the snowpack at each of the locations sampled. All snow samples are analyzed in a laboratory at Melbourne University where the Ag109 and In115 concentrations are measured using an inductively coupled plasma mass spectrometer (Huggins et al. 2008).

e. Experiment duration

The 2004 winter was used to establish most of the SPERP infrastructure, and to our refine seeding techniques and observing methods. The experimental phase of the SPERP extended from May 2005 through June 2009 when the anticipated target of 100 EUs was met; in fact 107 EUs were undertaken over that time. The conclusion of SPERP was essentially determined by the constraints of funding and the mandated duration of the project.

3. Evaluation plan

The analysis of cloud seeding experiments is often controversial, with claims that knowledge of the seeding sequence may influence the nature of the analysis and hence the overall results. Substantial effort was taken during the SPERP to ensure transparency in the analysis process. The evaluation plan was prepared well before the conclusion of the experiment and was published in the scientific literature (Manton et al. 2009).

An internal report was prepared by the analyst (M. J. Manton) before the seeding sequence was revealed describing the nature of the main datasets, and confirming that the data should be adequate for the primary analysis. The report and the datasets were provided to Snowy Hydro Limited and an independent referee (D. E. Shaw). Upon receipt of the report and data, the seeding sequence was sent by the referee to the analyst. The primary analysis was then undertaken independently by the analyst without reference to Snowy Hydro Limited officers. Following completion of the primary analysis and tabling of the associated internal report with the referee, the seeding sequence was revealed to Snowy Hydro Limited officers, and a number of secondary analyses were then undertaken.

a. Historical precipitation analysis

A detailed analysis of historical precipitation data was carried out in order to develop an appropriate evaluation methodology for SPERP (Manton et al. 2009). Because most of the sites used for SPERP were commissioned at the start of the project, only 11 of the 28 precipitation sites in the target and control areas have at least a decade of hourly data before 2004. Sounding data from the Bureau of Meteorology site at Wagga Wagga (130 km north east of Khancoban) were used in the historical analysis. Using the sounding and surface data, a historical EU dataset was created by approximating the seeding start conditions specified in section 2c; no conditions on the availability of supercooled liquid water could be applied. A total of 214 five-hour EUs were identified over the 9-yr period from 1995, that is, about 24 EUs per year. The technique was successfully checked by applying it to 2004 when SPERP trials were carried out. Analysis of the seasonal and interannual variability of EUs supported the basic assumption that a consistent set of about 100 EUs should be obtainable over a 5-yr period.

The key aspect of a historical analysis is the statistical simulation used to estimate the probability that a seeding impact will be detected over the duration of the experiment (Twomey and Robertson 1973; Heimbach and Super 1996). Precipitation in the control area is used to estimate natural precipitation in the target area, through linear regression (Dennis 1980). Assuming that 99 EUs are obtained in SPERP and with a seeding ratio of 2:1, we randomly select (with replacement so that all the draws are identical) 33 EUs from the historical set of 214 as the unseeded events (Efron 1982). With TU and CU as the precipitation in the target and control areas, respectively, for the unseeded EUs, we use linear regression to calculate the coefficients a and b, where
e1
and t is the EU number. We then randomly select (with replacement) 66 EUs from the historical set as the seeded EUs. To simulate the impacts of seeding, we increase the actual precipitation by s, where s = 0.05, 0.10, 0.20, or 0.40; that is, we set
eq1
where T and C are the actual precipitation in the target and control areas, and TS and CS are the simulated precipitation for the seeded EUs. The impacts of seeding can be estimated by the regression residual for the seeded EUs:
e2
The mean residual for the unseeded EUs is clearly zero. The overall precipitation increase (PI) is given by
eq2
and the total fractional increase in precipitation (TFI) is given by
e3
The regression approach used here is similar to that of Super (1986), who used residuals to identify the relationships between the seeding impact and external variables such as cloud-top temperature and wind direction. However, it is different from the approach of Smith et al. (1979), who use a seeding indicator variable in the regression equation, and that of Mielke et al. (1982), who recommend the use of median regression to reduce the impacts of outliers. The present approach is taken because of its physical simplicity and its capability to readily analyze links between seeding impact and external variables, as proposed by Super.

Stable results are found by repeating the simulation 600 times, and it is found that there is a 97% probability of detecting a positive impact when the actual impact is 0.20, but only a 72% probability when the actual impact is 0.05 (Manton et al. 2009). Embedded within each simulation is another bootstrap analysis (Efron 1982) to estimate the statistical significance of the computed seeding impact. For this test 66 of the 99 EUs are selected (with replacement) as “seeded” and 33 as “unseeded.” The regression analysis is used to estimate the natural precipitation in the target area, from which the seeded residual (RS) and total fractional increase (TFI) are computed. This resampling is repeated 400 times to build up the distribution of TFI, and the one-sided significance of TFI is estimated by the fraction of the distribution with values greater than the “observed” value (Smith et al. 1979). It is found that there is only a 63% probability that a 0.20 seeding impact will be detected at the 5% significance level. On the other hand, this probability increases to 77% if the significance level is relaxed to 10%. A 0.10 seeding impact has only a 42% chance of being detected at the 10% significance level.

The results of a target–control simulation are expected to be sensitive to the correlation between the target and the control precipitation. For the historical data, we find that the correlation between the primary target and the control is 0.71, and between the overall target and the control is 0.79. Although the correlation with the overall target is substantially larger than with the primary target, we find that the probability of detecting a TFI greater than zero is only marginally increased when the overall rather than primary target is used in the simulations. Similarly, the probability of detecting a 0.20 seeding impact at the 10% significance level only increases from 77% to 81% (Manton et al. 2009).

The bootstrap analyses described above can be used to estimate the double ratio (Smith et al. 1979) of the seeding impact (as well as the TFI). Both the probability of a positive result and the probability of detecting an impact at the 5% or 10% significance levels are found to be a little lower than for the TFI (Manton et al. 2009).

b. Snow chemistry

The chemical techniques of Chai et al. (1993) were fundamental in the initial design of SPERP, but two important factors need to be clarified in order to obtain quantitative estimates from the snow chemistry data. The first factor is the estimation of the time associated with each 20-mm slice from a snow profile (section 2d), and the second is the selection of the most appropriate variable for identifying a seeding impact.

The timing of real-time snow samples from Blue Cow is clear, but profiles from the other 16 sites are sampled only after campaigns when several EUs may have occurred (section 2d). The timing of snow slices from these sites is estimated by aligning each slice in a profile by weight with the timing tips of a collocated precipitation gauge. The number of slices in a 5-h period is determined by the precipitation rate, and there may be no well-defined snow slice contained within a specific EU. Moreover, a snow profile can be affected by warming and refreezing periods over the time that snow is collected. Thus, snow chemistry data may not be available for all EUs, and there can be considerable uncertainty in the timing of a specific slice.

While Chai et al. (1993) suggest that the ratio of silver to indium (Ag:In) is an indicator of the microphysical impacts of seeding, the ratio is found to vary greatly in both seeded and unseeded cases in the Snowy Mountains. One cause of uncertainties in Ag:In is variability in the background level of Ag. On the other hand, Huggins et al. (2008) suggest that the background levels of both Ag and In are less than 3 ppt. Nonetheless, it would seem that uncertainties in the value of Ag in snow can arise from uncertainties in the timing of snow slices, from background sources from mining operations to the west of the Snowy Mountains, or from wind-blown snow during winter storms. Field studies in 2004 suggest that the presence of In at concentrations above 1 ppt is indicative of the presence of tracer material from a seeding generator. While some uncertainties may remain about the causes of variability of Ag:In in the Snowy Mountains, the use of Ag to identify the targeting of seeding material is well established (Warburton et al. 1995a). In unseeded EUs, both Ag and Ag:In should be essentially zero; that is, Ag should be at background levels that are expected to be very small compared with levels associated with seeding.

c. Primary analysis

Mielke et al. (1982) point out that, as there is a substantial random aspect to cloud seeding experiments, the application of many different statistical tests to the observed data leads to the problem of multiplicity; that is, using many tests on the same data is likely to lead to false positive results. To minimize the risk of multiplicity, the evaluation of SPERP is separated into primary and secondary analyses. The primary analysis is the key test of whether there has been an impact of seeding in the target area. For SPERP the primary analysis has two components: one test is to use the precipitation data to determine whether the amount of precipitation in the target area has been enhanced in seeded EUs, and the second is to use the snow chemistry data to determine whether the seeding material has reached the target area.

The primary test of the impacts of seeding on precipitation uses the primary target rather than the overall target area. The reason for this decision is that precipitation over the primary target area is required to fall as snow (not rain) during seeding events (section 2a). Moreover, historical simulations (section 3a) suggest that the probability of detecting an impact is only marginally increased by using the overall target. Based on the historical analysis, the primary test is taken to be that the TFI, given by Eq. (3), in the primary target is greater than that at the 10% significance level. The 10% rather than 5% significance level is adopted because it is found in section 3a that the probability of detecting a positive result with only 100 EUs is relatively low. If the primary test is satisfied, then we would expect that more detailed secondary analyses should confirm the physical basis of the result. The analysis is carried out using the arithmetic mean of the precipitation at each site in the target and control areas. The significance test is the bootstrap method described in section 3a.

Because of the uncertainties associated with the snow chemistry (section 3b), the primary test of the targeting of seeding material is simply that the peak concentration of Ag in the primary target is greater when averaged over seeded EUs than over unseeded EUs at the 5% significance level. Only samples that can be unequivocally associated with one EU and for which In is greater than 1 ppt are used in the analysis. A Wilcoxon rank-sum test (Bauer 1972) is used to demonstrate the differences in the means. Using the peak value of Ag only when In is greater than 1 ppt should provide a large and robust signal, with the robustness being assured by the conditions on In ensuring that material from generators has reached each valid site. Consideration of the Ag:In ratio is left to secondary analyses.

d. Secondary analyses

A fully successful outcome of SPERP would be realized if both components of the primary analysis are achieved. The two tests of section 3c are physically independent, as the first test seeks evidence of a macroscale impact of seeding across the primary target area while the second seeks evidence of targeting of seeding material. However, the limited duration of SPERP means that there is a 23% chance that a 0.20 seeding impact may not be detected at the 10% significance level; that chance increases to 58% for a 0.10 seeding impact (section 3a). The second component of the analysis aims to consolidate the first result by confirming that seeding material is reaching the target area.

The purpose of secondary analyses is to support the results of the primary analysis if it yields a positive result or to help explain the sources of uncertainty if the primary analysis results are negative or uncertain. As just noted, the natural variability of precipitation means that there is no guarantee of a positive result from the primary analysis, and so the secondary analyses are vital elements in the overall assessment of a cloud seeding experiment. The secondary analyses aim to identify links between the seeding impacts and independent variables, such as the amount of supercooled liquid water.

4. Observed datasets

An extensive range of data was collected over the course of the SPERP, with procedures implemented to quality control and archive the experimental data to optimize accuracy and accessibility. In particular, three special datasets were produced: a basic dataset, consisting of 0.5-h values of all the variables observed during the experiment, to provide the foundation for a range of detailed studies (including case studies) on the physical processes associated with cloud seeding and general cloud processes; a second dataset, containing values of the main variables for each EU, to support specific studies on the characteristics of events that are suitable for seeding; and a third core dataset, containing the key precipitation and snow chemistry data, to allow the primary analysis to be carried out.

The primary analysis is based on both precipitation and snow chemistry data. A range of secondary analyses must be carried out to explain the results of the primary analysis and to develop strategies to improve the methodology for cloud seeding. The secondary analyses use additional data from radiosondes released from Khancoban and from instruments operated at Blue Cow (Fig. 6).

a. Precipitation

The primary analysis for SPERP is based on the arithmetic mean of the precipitation at 44 sites across the target, control, and extended areas (Fig. 6). There are 8 sites in the primary target area, 8 in the secondary target area, 12 in the control area, and 16 in the extended area downwind of the target and control areas. It is noted in section 2d that efforts to improve the quality of precipitation measurements resulted in 62 instruments being located at the 44 sites. To remove any bias from having several observations from the one site, only one measurement is taken from each site. A priority sequence was developed to determine the instrument to be used at each site for each EU, with priority given to the highest quality gauge. A number of checks were made to ensure that the instrument selection process provided an accurate estimate of precipitation and that the area-average results are statistically robust; for example, the difference between the precipitation using the priority instrument and using all available instruments was computed and found to be generally small.

The representative precipitation in the primary analysis is given by the arithmetic mean of all valid sites in each area for each 5-h EU. It is noted in section 2d that not all precipitation gauges were commissioned at the start of SPERP. However, no apparent trends or biases are found in the mean values of the target and control precipitation. As expected, the precipitation in the primary target area, which covers the high mountain range, has the highest precipitation levels, while the extended area covering the lower levels to the east of the target and control areas tends to have much lower precipitation (Fig. 7). A statistical summary of the precipitation (mm) in each area over all EUs is given in Table 1. It is apparent that there are EUs with no precipitation in either the target or control areas; in particular, EU88 has no precipitation in the control area and EU91 has none in the target.

Fig. 7.
Fig. 7.

Histograms of precipitation in primary target, overall target, control, and extended areas over all EUs.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

Table 1.

Statistics on precipitation (mm) over all EUs in the primary target, the overall target, the control, and the extended areas for SPERP.

Table 1.

The spatial variability of precipitation across the region is indicated by Fig. 8, which shows the correlation of the precipitation at each site across all EUs with the mean precipitation in the control area. It is clear that the target and control areas are highly correlated, especially along the mountain ridges. The rain shadow to the southeast in the lee of the mountains results in considerably lower correlations. The correlations between the target areas and the control have values of 0.82 for the primary and 0.90 for the overall targets. The correlations are significantly higher than those found using the limited historical data (section 3c). The decrease in correlation to 0.76 between the control and extended areas emphasizes the difference in the basic rainfall climate of the two regions.

Fig. 8.
Fig. 8.

Correlation of precipitation at each site across all EUs used in the SPERP analysis with the mean precipitation in the control area; sites are indicated by black dots; the red line shows the primary target area, blue shows the overall target area, and green shows the control area; contours are at 0.1 intervals.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

b. Snow chemistry

The primary analysis for SPERP is based on measurements of Ag and In in the primary target area. Real-time snow samples were collected during seeding campaigns at Blue Cow, and snow profiles were collected from 16 sites across the study area after each major experimental campaign. Profile depths varied from 20 to 700 mm, with a median of 180 mm. The challenge associated with snow profile samples is the accurate assignment of a time period (and hence an EU) to each sample. The weight of each sample is used to estimate the slice sample time by comparison with collocated precipitation measurements; that is, the accumulated weight of snow in a profile is aligned with the accumulated depth of precipitation in a collocated gauge. Uncertainties arise not only from difficulties with interpolation along the snow profile but also from the variable weather of the region: snowfalls may be interspersed with warm days and periods of rain, which subsequently freezes. Inspection of collocated temperature measurements is used to identify and reject profiles affected by melting. Quality control and calibration of the raw Ag and In data were carried out before the primary analysis commenced, and the processed Ag and In data were part of a data package sent to the independent referee prior to the execution of the primary analysis.

The median EU precipitation in the primary target area is only 2.5 mm, with the first quartile at 1.1 mm. Thus, a 20-mm snow slice may cover a longer time period than a single 5-h EU. Indeed, it is found that slices can cover several EUs during an intense seeding campaign. Care is taken in the primary analysis to include only slices that intersect a single EU. A total of 167 profiles were collected, yielding 1746 raw slices. There were 311 samples from Blue Cow and 1018 slices from the profiles that were in the primary target and passed the quality control checks. Of the total of 1329 samples, 568 aligned unambiguously with one EU, and 441 (78%) of those had In > 1 ppt. This result implies that about 80% of the samples over all EUs and all sites in the primary target area contained material from the generators. To separate the variability between and within EUs, we compute the number of samples aligned unambiguously with one EU (n1) for each EU and the number of those with In > 1 ppt (nin) for each EU. It is found that 85 EUs have values of n1 greater than 0, but only 3 of those EUs have nin equal to 0. This result suggests that there is a more than 95% chance of some material from the generators being found somewhere in the target area at some time during an EU. Thus, most of the variability in targeting seems to occur within an EU, and this is confirmed by examination of the ratio nin:n1, which has a median value of 0.8. Valid chemistry data are not available for 25 EUs, owing largely to the scarcity of samples covering one EU unambiguously.

c. Generator hours

The basic assumption of cloud seeding is that the seeding material leads to an increase in the precipitation that would develop and fall through natural processes. It follows that there should be some relationship between the amount of seeding material in the target area and the increase in precipitation. Careful records of the times of operation of the generators that release the seeding material and the tracer material [indium (III) oxide] were maintained during the course of the SPERP. The seeding records were not made available to the personnel involved with conducting the experiments.

With 13 generators in operation and an EU length of 5 h, the maximum cover over the target is 65 generator hours per EU. The number and timing of generators for each EU are determined in real time as a function of wind speed and direction from the Khancoban upper-air soundings (section 2c). The generator hours for each EU over the course of the SPERP ranged from 13 to 61 h, with a median value of 53 h. It is found that about 20% of EUs were subject to fewer than 45 generator hours. Eight EUs were suspended (generally due to the mandatory freezing-level criterion of 1600 m), and these cases also had relatively low generator hours. Low generator hours also occurred when the wind direction limited the number of generators that could be used while ensuring that seeding material passed over only the target area.

d. Supercooled liquid water

The physical hypothesis for SPERP is that silver iodide acts as an ice nucleus allowing ice to form in cloud regions with SLW, which can naturally persist without ice formation in some locations and some periods of a storm. The artificially formed ice particles grow rapidly at the expense of the SLW and become large enough to fall as precipitation (Cotton and Pielke 2006). The hypothesis is based on the observation that the concentration of natural ice nuclei is generally very low, so that regions of SLW can form in some locations such as the mountains of southeastern Australia (Long and Huggins 1992).

For analysis purposes, the value of SLW is set equal to 0 if the temperature at Blue Cow (at 1921 m) is above 0°C; this ensures that there is a conservative estimate of the presence of SLW. As part of the EU dataset for SPERP, two SLW measurements are archived and analyzed: the start SLW over the half-hour before an EU commences and the mean SLW over the duration of the EU. The start SLW ranged from 0 to 0.62 mm, with a median value of 0.11 mm. The mean SLW over an EU varied from 0.001 to 0.47 mm, with a median of 0.07 mm. The levels of SLW observed during SPERP are somewhat lower than the 1-h values of 0.2–0.4 mm recorded by Warburton and Wetzel (1992) during storms in the Snowy region in 1989. However, their observations were not subject to all the cloud seeding conditions of SPERP. It may be expected that the amount of available SLW should relate to the effectiveness of seeding, which is investigated as a secondary analysis in Part II of this paper. Owing to instrument failure, observations of SLW were not obtained for three EUs.

e. Wind

Wind has two roles in cloud seeding. First, it is the wind that disperses the seeding material horizontally and vertically into the clouds across the region of interest. The wind speed also determines the time available for the seeding material to affect the microphysics of the clouds, leading to precipitation development. It may happen that at very high wind speeds there is insufficient time for the microphysical effects to occur before the seeding material is transported out of the seeding area and out of the region of SLW.

The wind speed and direction may also be seen as representing the nature of the synoptic weather system bringing the precipitation to the region. The wind direction in particular can indicate the state of the synoptic system as it passes the seeding region.

The vector wind generally varies with altitude and so a representative value must be selected. The seeding hypothesis assumes that the seeding effect of silver iodide becomes significant above the −5°C level. Such a level is also likely to be a reasonable representation of the synoptic features of the weather system. The wind speed and direction from the Khancoban radiosonde at the −5°C level at the start of an EU are therefore taken as the representative wind vector for SPERP. The wind speed at the −5°C level varied from 3 to 27 m s−1 with a median value of 12 m s−1, and the wind direction ranged from 220° to 341° with a median of 277°. Histograms of the wind speed and direction are given in Fig. 9.

Fig. 9.
Fig. 9.

Histograms of wind (top) speed and (bottom) direction at the −5°C level at the start of EUs.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

f. Cloud properties

The Khancoban soundings at the start of an EU provide a significant amount of information on the cloud properties of the weather system bringing precipitation to the region. The cloud-top temperature (CTT) is a significant indicator of the depth and often intensity of the synoptic system. Moreover, it is known that natural atmospheric processes tend to cause naturally occurring particles to act as ice nuclei at temperatures lower than about −20°C (Dennis 1980). We may therefore expect to find some relationship between CTT and seeding impact (Super 1986). The values of CTT ranged from −46.7° to −6.2°C with a median of −14.0°C, indicating the great variation in CTT. At the warmer end of this range, the depth of cloud above the −5°C level is likely to limit the amount of nucleated seeding material. At the cold end, homogeneous ice nucleation can occur and so natural processes are likely to dominate at cloud top.

The −5°C level is important for the microphysical processes of cloud seeding with silver iodide (Cotton and Pielke 2006), and so the height of this level is taken as a variable for analysis. The height of the −5°C level varied from 1320 to 2440 m with a median value of 2160 m (Fig. 10), and so it can lie below the high mountain ridges at 1900 m. Another important height is the freezing level, which must be at or lower than 1600 m so the risk of precipitation falling as rain (rather than snow) on the mountain ridges is minimized. The imposed upper limit on the freezing level produces a cutoff in the distribution such that the mode occurs at the maximum value (Fig. 10).

Fig. 10.
Fig. 10.

Histograms of (top) the freezing level and (bottom) the height of −5°C level at the start of EUs.

Citation: Journal of Applied Meteorology and Climatology 50, 7; 10.1175/2011JAMC2659.1

For completeness, we also consider the distributions of the heights of cloud top and cloud base. As cloud base often comes to ground level, it is not a particularly effective variable for analysis. The cloud-top height varied from 2260 to 7780 m with a median of 3580 m. As reflected in CTT, cloud-top height can be very high for some EUs. A starting condition for seeding (section 2c) is that the depth of the cloud above the −5°C level is greater than 400 m. This depth is found to be generally between 500 and 2000 m with a median of 1440 m.

5. Conclusions

A confirmatory glaciogenic cloud seeding experiment has been carried out in the Snowy Mountains of southeastern Australia, using ground-based generators dispersing material into the westerly streams associated with the passage of winter synoptic fronts. Seeding is randomized with a 2:1 seeding ratio. The experimental design is based on past experience of orographic seeding in the United States and Australia, as well as earlier field studies in the Snowy Mountains. The evaluation methodology was published well before the field experiment ended. An initial evaluation of the observed data has been carried out before the seeding sequence was known, to ensure that the datasets are adequate for the primary analysis. The observed data are found to be consistent with the historical observations and analyses, and they support the assumptions used in the design of SPERP. The design process should provide a sound basis for the formal evaluation of the impact of seeding on precipitation in the Snowy Mountains. The evaluation is considered in Part II of this paper.

Acknowledgments

The thoughtful comments of the project referees, Dr. Warren King and Dr. Doug Shaw, helped to refine the analysis. The manuscript benefited greatly from the insights of three anonymous reviewers. The project was wisely managed by John Denholm of Snowy Hydro Limited.

REFERENCES

  • Bauer, D. F., 1972: Constructing confidence sets using rank statistics. J. Amer. Stat. Assoc., 67, 687690.

  • Bruintjes, R. T., 1999: A review of cloud seeding experiments to enhance precipitation and some new prospects. Bull. Amer. Meteor. Soc., 80, 805820.

    • Search Google Scholar
    • Export Citation
  • Chai, S. K., , W. G. Finnegan, , and R. L. Pitter, 1993: An interpretation of the mechanism of ice-crystal formation operative in the Lake Almanor cloud-seeding program. J. Appl. Meteor., 32, 17261732.

    • Search Google Scholar
    • Export Citation
  • Cotton, W. R., , and R. A. Pielke, 2006: Human Impacts on Weather and Climate. 2nd ed. Cambridge University Press, 332 pp.

  • Dennis, A. S., 1980: Weather Modification by Cloud Seeding. Academic, 275 pp.

  • Deshler, T., , D. W. Reynolds, , and A. W. Huggins, 1990: Physical response of winter orographic clouds over the Sierra Nevada to airborne seeding using dry ice or silver iodide. J. Appl. Meteor., 29, 288330.

    • Search Google Scholar
    • Export Citation
  • Efron, B., 1982: The Jackknife, the Bootstrap and Other Resampling Plans. Capital City, 93 pp.

  • Heimbach, J. A., , and A. B. Super, 1996: Simulating the influence of type II error on the outcome of past statistical experiments. J. Appl. Meteor., 35, 15511567.

    • Search Google Scholar
    • Export Citation
  • Huggins, A. W., 1995: Mobile microwave radiometer observations: Spatial characteristics of supercooled cloud water and cloud seeding implications. J. Appl. Meteor., 34, 432446.

    • Search Google Scholar
    • Export Citation
  • Huggins, A. W., , S. L. Kenyon, , L. Warren, , A. D. Peace, , S. P. Bilish, , and M. J. Manton, 2008: The Snowy Precipitation Enhancement Research Project: A description and preliminary results. J. Wea. Modif., 40, 2851.

    • Search Google Scholar
    • Export Citation
  • Long, A. B., , and A. W. Huggins, 1992: Australian Winter Storms Experiment (AWSE) I: supercooled liquid water and precipitation-enhancement opportunities. J. Appl. Meteor., 31, 10411055.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., , and L. Warren, 2011: A confirmatory snowfall enhancement project in the Snowy Mountains of Australia. Part II: Primary and associated analyses. J. Appl. Meteor. Climatol., 50, 14481458.

    • Search Google Scholar
    • Export Citation
  • Manton, M. J., , J. Dai, , and L. Warren, 2009: Evaluation plan for a snow enhancement experiment in Australia. J. Wea. Modif., 41, 5974.

    • Search Google Scholar
    • Export Citation
  • Mielke, P. W., , K. J. Berry, , and J. G. Medina, 1982: Climax I and II: Distortion resistant residual analyses. J. Appl. Meteor., 21, 788792.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , M. J. Manton, , and A. Nazarov, 2009: On the analysis of a cloud seeding dataset over Tasmania. J. Appl. Meteor. Climatol., 48, 12671280.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , M. J. Manton, , and A. Nazarov, 2010: A modeling case study of mixed-phase clouds over the Southern Ocean and Tasmania. Mon. Wea. Rev., 138, 839862.

    • Search Google Scholar
    • Export Citation
  • Morrison, A. E., , S. T. Siems, , and M. J. Manton, 2011: A three-year climatology of cloud-top phase over the Southern Ocean and North Pacific. J. Climate, 24, 24052418.

    • Search Google Scholar
    • Export Citation
  • National Research Council, 2003: Critical Issues in Weather Modification Research. National Academies Press, 144 pp.

  • Rauber, R. M., , R. D. Elliott, , J. O. Rhea, , A. W. Huggins, , and D. W. Reynolds, 1988: A diagnostic technique for targeting during airborne seeding experiments in wintertime storms over the Sierra Nevada. J. Appl. Meteor., 27, 811828.

    • Search Google Scholar
    • Export Citation
  • Reynolds, D. W., , and A. S. Dennis, 1986: A review of the Sierra Cooperative Pilot Project. Bull. Amer. Meteor. Soc., 67, 513523.

  • Ristovski, Z., , and R. Jayaratne, 2005: Measurement of the size distribution of cloud seeder (AgI) and tracer (InO2) aerosols produced by an LPG burner. International Laboratory for Air Quality and Health, Queensland University of Technology, 22 pp.

    • Search Google Scholar
    • Export Citation
  • Ryan, B. F., , and W. D. King, 1997: A critical review of the Australian experience in cloud seeding. Bull. Amer. Meteor. Soc., 78, 239254.

    • Search Google Scholar
    • Export Citation
  • Smith, E. J., , E. E. Adderley, , and D. T. Walsh, 1963: A cloud-seeding experiment in the Snowy Mountains, Australia. J. Appl. Meteor., 2, 324332.

    • Search Google Scholar
    • Export Citation
  • Smith, E. J., , L. G. Veitch, , D. E. Shaw, , and A. J. Miller, 1979: A cloud-seeding experiment in Tasmania. J. Appl. Meteor., 18, 804815.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., 1986: Further exploratory analysis of the Bridger Range winter cloud seeding experiment. J. Climate Appl. Meteor., 25, 19261933.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., , and J. A. Heimbach, 1983: Evaluation of the Bridger Range winter cloud seeding experiment using control gages. J. Climate Appl. Meteor., 22, 19892011.

    • Search Google Scholar
    • Export Citation
  • Super, A. B., , and J. A. Heimbach, 1988: Microphysical effects of wintertime cloud seeding with silver iodide over the Rocky Mountains. Part II: Observations over the Bridger Range, Montana. J. Appl. Meteor., 27, 11521165.

    • Search Google Scholar
    • Export Citation
  • Twomey, S., , and I. Robertson, 1973: Numerical simulation of cloud seeding experiments in selected Australian areas. J. Appl. Meteor., 12, 473478.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , and M. A. Wetzel, 1992: Field study of the potential for winter precipitation enhancement in the Australian Snowy Mountains. Atmos. Res., 28, 327363.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , R. H. Stone, , and B. L. Marler, 1995a: How the transport and dispersion of AgI aerosols may affect detectability of seeding effects by statistical methods. J. Appl. Meteor., 34, 19291941.

    • Search Google Scholar
    • Export Citation
  • Warburton, J. A., , L. G. Young, , and R. H. Stone, 1995b: Assessment of seeding effects in snowpack augmentation programs: Ice nucleation and scavenging of seeding aerosols. J. Appl. Meteor., 34, 121130.

    • Search Google Scholar
    • Export Citation
Save