Seeding and precipitation processes.

Weather modification research requires the involvement of a large range of expertise due to both the multifaceted nature of the problem and the large range of scales that are addressed. The large and meso scale dynamics determining the characteristics of the cloud systems down to the small-scale microphysics determining the nucleation and growth characteristics of water droplets and ice particles all form part of the chain of events of precipitation development  that is shown next. Although our knowledge of the individual aspects in the chain has significantly increased in the past twenty years there still exist major gaps about certain physical processes.
 
 

Weather modification saw the inception of its modern era with the discoveries of Schaefer (1946) and Vonnegut (1947) showing that supercooled liquid water could be converted to ice crystals using either dry ice or silver iodide. The motivation and conception of these projects were based on conceptual models developed from past experience defining conditions that are conducive to positive seeding effects.

These conceptual models were based on

visual observations of clouds that did not precipitate,
the presence of supercooled water,
the similarity with clouds in other regions which responded positively to seeding, and
data collected with aircraft and radars, among others (Vali et al., 1988).

Since the discovery of glaciogenic materials more than forty years ago, both silver iodide and dry ice are still the most widely used cloud-seeding materials in the world. Both materials enhance the ice crystal concentrations in clouds by either nucleating new crystals or freezing cloud droplets. Based on their ice-nucleating capabilities two seeding concepts have been proposed in the past, namely the static and dynamic seeding concepts (Braham, 1986).

   Convective clouds

Some of the initial steps in this chain of events have been demonstrated in field measurements, laboratory and modeling studies including increased concentrations of ice crystals and the more rapid production of precipitation particles in cumulus clouds. In the HIPLEX-1 experiment a detailed seeding hypothesis (Smith et al., 1984), together with a well-designed field program which monitored each step in the physical hypothesis, was conducted. Although the experiment failed to demonstrate statistically all the hypothesized steps, the reasons for the failures could be traced to the physical data set (Cooper and Lawson, 1984). This in itself is a significant result that indicates the importance of the ability of physical measurements and studies to provide understanding of the underlying physical processes in each experiment.

It is interesting to note that although the HIPLEX seeding hypothesis called for the ice crystals produced by seeding to develop into graupel, measurements at -6oC in HIPLEX clouds did indicate large numbers of aggregates (Cooper and Lawson, 1984). This was due to the short-lived nature of HIPLEX-1 clouds and the rapid decrease in supercooled liquid water, which is the primary growth source for graupel. The aggregates on the other hand, had much lower fall velocities than graupel particles and stayed aloft and evaporated, resulting in no increase in precipitation. This result indicates that not all clouds may be amenable to seeding and that there exists a certain window of opportunity. For the static seeding concept this opportunity appears to be limited to continental cold-based clouds with top temperatures approximately between -10 and -20oC, and limited to the time when significant amounts of supercooled liquid water is available for growth by riming of the seeded produced ice crystals (Cooper and Lawson, 1984).

Experiments which used a combination of the statistical and physical approach are the Canadian studies by Isaac et al. (1977, 1982) and English and Marwitz (1981), the World Meteorological Organization (WMO) Precipitation Enhancement Project (PEP) (WMO PEP Report No. 34, 1986; Vali et al., 1988), the Australian experiments (Ryan and King, 1997), the South African studies by Krauss et al. (1987), and others (e.g., Dye et al., 1976; Holroyd et al., 1978; Sax et al., 1979; Hobbs and Politovich, 1980; Orville, 1996).
Most of these experiments were conducted on semi-isolated cumulus congestus clouds to provide a relatively simple cloud dynamics framework to confirm fundamental cause and effect relations of cloud microphysical processes. However, such clouds do not contribute significantly to rainfall at the ground. Convective complexes contribute significantly more than semi-isolated cumulus congestus clouds to the rainfall at the surface in most regions where a major part of the annual precipitation is the result of convective activity (Bureau of Reclamation, 1979). However, convective complexes are much more complex dynamically than the smaller clouds because they are to a large extent manifestations of mesoscale and large-scale dynamical processes.

The increases in precipitation at the ground due to the static seeding concept in convective cumulus clouds have in general been inconclusive and the initial optimism has been replaced by a more cautious approach. Braham's (1986) list of factors that have limited research progress can be summarized in two points: the large natural variability, and an incomplete understanding of the physical processes involved.

The Israeli experiments (Gagin and Neumann, 1981) have provided the strongest evidence to date that static seeding of convective cold-based continental clouds can cause significant increases in precipitation on the ground. However, Rangno and Hobbs (1995) have questioned the validity of the conclusions of the Israeli experiments. From their re-analyses of the Israeli I and II experiments they claim that that seeding-induced increases in Israeli I were contaminated by a type I statistical error (i.e. a lucky draw). In addition, they claimed that naturally higher precipitation in the north target area on seeded days during Israeli II could have been mistaken for a seeding induced change in precipitation.

The apparent decrease in rainfall in the south target area in Israeli II was linked to the incursion of desert dust by Rosenfeld and Farbstein (1992). They suggest that desert dust contain more ice nuclei and also can provide coalescence embryos that can enhance the collision-coalescence process in clouds providing for more efficient precipitation processes in the clouds.

The original thought that clouds in Israel were continental in nature and that ice particle concentrations in these clouds were generally small for cloud tops warmer than -12oC with neither coalescence nor ice multiplication process operating, have also been questioned. Rangno and Hobbs (1995) and Levin (1994) presented evidence for the existence of large supercooled droplets and high ice concentrations at relatively warm temperatures in these clouds. Although the measurements represented a limited number of cases it somewhat erodes the earlier perception that clouds over Israel were highly susceptible to seeding (Gagin, 1986; Gagin and Neuman, 1974).
The criticisms of Rangno and Hobbs (1995) have generated a significant number of responses in the scientific literature (Rosenfeld, 1997; Rangno and Hobbs, 1997a; Woodley, 1997; Rangno and Hobbs, 1997b; Ben-Zvi, 1997; Rangno and Hobbs, 1997c; Dennis and Orville, 1997; Rangno and Hobbs, 1997d). Although many of the issues were clarified by these comments, the perception that the Israeli experiments were the most successful example of precipitation enhancement has been weakened.

Mather et al., (1996) reported results from randomized cloud seeding experiments using dry ice in South Africa. They hypothesize that their results do not totally fit the static seeding hypothesis but also include the freezing of large drops that grow much faster than graupel particles (Johnson, 1987). The results from 127 storms analyzed using radar data indicate that radar-measured rain flux and storm area from seeded clouds were significantly larger than for the unseeded clouds. In their analyses a floating target defined by the storm track from the radar was used. Although these results indicate increases in rain from specific storms, they do not address the issue of rainfall increases over a target area on the ground. Results of this study also indicated that clouds in which the coalescence process was active seem to be more amenable to seeding (Mather et al., 1986).
 
 

   Winter orographic cloud seeding

Since the first conceptual models, (Bergeron, 1949; Ludlum, 1955) attempts began to increase winter snowpack on mountain ranges by seeding clouds with silver iodide or dry ice, and several operational and research winter orographic cloud-seeding programs have been conducted worldwide. Many of the steps in the physical chain of events associated with the static seeding concept have also been documented in these experiments (Elliott (1986; Reynolds, 1988; Reynolds and Dennis, 1986; Super, 1990; and Reinking and Meitin 1989). These studies have shown that seeding does increase precipitation under certain favorable conditions (American Meteorological Society, 1992), and can result in increases in snowpack. However, there are still many unanswered questions. In particular, the variability of clouds in complex terrain and associated temporal and spatial changes in wind flow and regions of supercooled liquid water lead to difficulties in the targeting and dispersion of seeding material and the identification of suitable seeding situations.

A review of the relevant literature immediately highlights the correlation between the temporal and spatial evolution of cloud liquid water (CLW) and the complexity of the terrain (Rauber et al., 1986; Rauber and Grant, 1986; Marwitz, 1986; Deshler et al., 1990; and Huggins and Sassen, 1990). Rangno (1986) notes that the cloud variability encountered in several mountainous areas in the United States poses severe challenges for forecasting seeding opportunities and determining a treatment strategy, especially when seeding opportunities are short-lived. This conclusion was re-iterated by Super and Holroyd (1989) based on studies in Arizona where they specifically noted that, although CLW was present in all storm systems, it was highly variable in time. However, in most of these studies, measurements of CLW with microwave radiometers indicated many hours of CLW that could potentially be seeded (Huggins, 1995).

The temporal and spatial variability of CLW also poses a severe problem for targeting regions of CLW with seeding material. This was especially highlighted in seeding experiments over the Sierra Nevada (Deshler et al., 1990) where the complete chain of events from seeding to precipitation could be documented in only two of 36 experiments. The authors ascribed the failures to difficult technical and logistic limitations, and to the variability of even simple cloud systems, and not necessarily to the seeding conceptual model. This was particularly evident in the spatial and temporal distributions of CLW and in the natural fluctuations in ice crystal concentrations. Huggins and Sassen (1990) were also unable to document the physical chain of events from seeding to precipitation at the surface in seeding experiments in the Tushar Mountains in Utah. Once again, insufficient knowledge about the transport and dispersion of the seeding material was quoted as one of the primary reasons for failure. In the Utah experiment as well as in many other orographic seeding experiments, seeding generators were located in fixed positions upwind from the target while the measurement facilities were concentrated in a single location downwind from the seeding generators. These locations were chosen assuming a mean wind direction and assuming that no changes in wind direction occur between the seeding generator and the target position. One would expect that with fixed seeding and target locations, seeding effects would only be detected when the flow is parallel to a line connecting the seeding generator and the target. Thus, the opportunities to document the chain of physical events are limited. This approach also assumes that CLW regions will always be present in the same location and that sufficient amounts are present for seeding material to interact with, and to produce precipitation in the target area.
 

Due to the problem of insufficient knowledge about the wind-flow patterns and associated CLW regions, some investigators proposed conducting continuous seeding during the entire storm duration in the hope that seeding would have positive effects when CLW was present and no effects when CLW was not present (Super and Holroyd, 1989; Super, 1990). However, this approach may further mask seeding effects and may even have negative effects in certain instances. This may explain why some seeding experiments in winter orographic regions have produced either inconclusive or negative results.
 

The timely identification of regions of supercooled liquid water and the efficient targeting and dispersion of seeding material in mountainous terrain remains a difficult problem. This will have to be considered when developing guidelines and strategies for dispersing seeding material. An important aspect emphasized in nearly all past experiments was the need for more wind measurements in time and space between the seeding release site and the target area. Although this requires a dense network of sensors, recent studies (Bruintjes et al., 1994, 1995; Heimbach and Hall, 1994, 1996) have shown the utility of using state-of-the-art models for guiding and understanding the flow patterns and associated CLW regions in complex terrain.
 

During the last ten years there has been a thorough scrutiny and evaluation of cloud seeding projects involving the static seeding concept. Although there still are indications that seeding can increase precipitation, a number of recent studies have questioned many of the positive results, weakening the scientific credibility of some of these experiments. As a result considerable skepticism exists whether this method provides a cost-effective means for increasing precipitation for water resources.