General

The Israeli cloud seeding project was based on the static cloud seeding conceptual model. Rangno and Hobbs questioned the validity of this conceptual model for the Israeli clouds. They made their own survey on the microstructure of the Israeli clouds, which led them to conclude that “Contrary to previous reports, clouds in Israel contain large cloud droplets, precipitation-sized drops, and considerable concentrations of natural ice particles at quite high temperatures, all of which should obviate attempts to increase rainfall by artificial seeding in wintertime air masses.”

However, RH acknowledged that there is a large variability in clouds’ microstructure in Israel. Even if their description is accepted as is, it pertains only to a fraction of the clouds, which, according to the quoted concept of RH, are not suitable for rain enhancement. It was never claimed that all clouds in Israel are suitable for rain enhancement.

Furthermore, RH consider clouds that contain active warm rain processes and ice multiplication as unsuitable for rain enhancement. But glaciogenic seeding of such clouds has demonstrated positive seeding effects, at least on the convective cell scale (e.g., Simpson and Woodley 1971; Gagin et al. 1986; Rosenfeld and Woodley 1989, 1993). This does not imply that clouds in Israel have to be similar to those in Texas or Florida, or that the static seeding hypothesis is exchanged with the dynamic seeding hypothesis. All that is said is that existence of warm rain processes and ice multiplication does not mean that realization of rain enhancement is less likely than from “textbook” continental clouds.

Therefore, the conclusion that cloud seeding is not likely to enhance rainfall in Israel is not a necessarily logical outcome, even if RH’s new microphysical description was correct. In addition, RH’s description of the Israeli clouds is based on too few measurements to serve as a basis for any generalizations, as will be shown in the next sections.

Rangno and Hobbs based their new microphysical description on

• subjective observations of clouds during a visit to Israel,

• comparison of cloud-top temperature with rainfall, and

• in situ measurements, as published by Levin (1992).

Subjective observations of clouds

Rangno and Hobbs provided, in their Fig. 9, pictures of a convective cloud to the north-northwest of Tel Aviv. This picture is used to suggest a fast natural glaciation of cloud tops at −14°C in the “polar maritime” air mass in Israel and the existence of haze in a polar maritime air mass.

Rangno and Hobbs neglected in their calculations the distortion of the vertical scale in high-elevation viewing angles. Taking this into consideration should add 235 m to their calculations and lower the cloud-top temperature to −15°C to −16°C. Convective clouds with such top temperatures always produce some precipitation particles, at least aloft.

The cloud in question is documented by Electrical Mechanical Services weather radar at Ben-Gurion airport. The cloud was well defined, as seen in Fig. 1. The echo top of the cloud reached 5200 m at the time that RH’s Fig. 9b was taken and exceeded 6000 m 5 min later. The temperature at 5200 m was −19°C and not −14°C, as claimed by RH. Classic continental clouds can be expected to already glaciate at −19°C. The remaining discrepancy of 400 m between the radar and RH’s estimation can be attributed to errors in either estimate.

According to the radar, the cloud is at horizontal range of 20 km from the point where RH photographed the cloud (Tel Aviv beach). Since cloud base is seen very clearly at that distance, the picture serves as evidence for lack of significant haze in this situation, in contrast to the claim made by RH.

Conclusions opposite to RH’s can be reached by carefully inspecting their evidence.

Comparison of cloud-top temperature with rainfall

Rangno (1988) showed that light rainfall occurs in Israel from clouds with tops warmer than −10°C. Rangno and Hobbs used these observations to support their suggestion that clouds in Israel are not susceptible to seeding. However, only about 4% of the precipitation depth in Israel is contributed from such clouds (Rosenfeld and Gutman 1994). Since rain intensities are typically very light in these conditions, they can persist for a considerable portion of the rainy periods. These conditions typically occur in Israel under strong midlevel inversions, where long-living cumulus and stratocumulus clouds are formed. These clouds live long enough for the development of some ice and warm rain processes in continental clouds. But much of the rainfall in Israel occurs from more convective and shorter-lived clouds, where time is a major limiting factor in precipitation efficiency (Rosenfeld and Gagin 1989). Seeding such clouds can make a significant difference, by reducing the time needed to initiate the precipitation formation processes.

Recent in situ measurements

Rangno and Hobbs rely on in situ measurements (Levin 1992) to conclude that large concentrations of ice and large cloud droplets are typical to Israeli clouds. The available microphysical measurements are insufficient for any generalization of the properties of clouds in Israel, much less for supporting RH’s specific conclusions. The measurements with FSSP 1D and 2DC probes were conducted in five research flights during 1990. Data from a very small number of clouds were obtained. Not all instruments were operational in all flights. The cloud-base drop size distributions varied widely, some of them being narrow and some of them wide. The quality and amount of data do not warrant any generalization such as RH made. Subsequent cloud physics measurements made by us show similar variability.

Case study analyses are meaningless without having a minimal dataset (FSSP 1D and 2DC, vertical sounding, temperature, vertical wind, and cloud-base temperature) and without knowing the context of the cloud in the rain cloud system and within its life cycle. No such data were available to RH.

Gagin’s in situ measurements

Rangno and Hobbs pointed out the existence of profound differences between ice concentrations that were found by Gagin and his colleagues in Israel (Gagin 1975; Gagin and Neumann 1974, 1981) and the much larger ice concentrations in similar clouds compiled by Rangno and Hobbs (1988) and measured by Levin (1992). This apparent contradiction can be simply resolved by recalling that Gagin was interested mainly in the initiation of precipitation in clouds, since late initiation could mean less supercooled water being available for the mixed phase precipitation formation processes in the later stages of the cloud’s life cycle. In contrast, Rangno and Hobbs (1988) and Levin (1992) counted the maximal concentrations of ice particles in mature and aging clouds.

The added speculations of RH about the way Gagin et al. conducted their measurements are not necessary to explain the differences between the Israeli clouds and the other clouds, as shown in point 8 of RH’s Fig. 12. In the same Fig. 12, point 10 (northeast Colorado) is even more out of line with the rest of the clouds. Rangno and Hobbs (1988) explained this difference as a result of measurements being taken only in fresh, vigorously growing, convective towers, without ever questioning the validity of the measurement methods. Why should RH then question the validity of Gagin et al.’s microphysical measurements in the Israeli clouds?

Microphysical summary

The available microphysical data in the relevant meteorological and cloud evolution context do not support RH’s suggestion of the dominance of “maritime” clouds in Israel. If anything, they support the suggestion of Levi and Rosenfeld (1996) that clouds in southwesterly flow and near the southern margins of the rain cloud system (Rosenfeld and Nirel 1996) are affected by the desert dust, mostly by its glaciogenic activity (see section 7).

A trivial error

Before discussing the substance of RH’s interpretations of the Israeli I experiment, it is necessary to point out that the north–south dimension of the Israeli I experimental area was 160 km and not 200 km, as erroneously stated by RH.

The fraction of clouds that are seeded

Rangno and Hobbs calculate that seeding was done only during about 30% of the “showery periods.” They defined showery periods as time intervals in which even the lightest rain was recorded. Rangno (1988) has already shown that Israel is frequently covered with shallow clouds that produce light rainfall and that these clouds are not suitable for glaciogenic seeding. These conditions are responsible for roughly half of the duration of the showery periods, as defined by RH. This is evident from Fig. 4 of RH, which indicates that during half of the rainy days cloud tops are lower than 750 mb in central Israel (the temperature at 750 mb reaches −5°C in the colder rainy days). The seeding criteria were designed to avoid wasting flight hours and resources on such clouds. Therefore, based solely on information provided by RH themselves, showery periods with suitable clouds should be reduced to half, and the seeding fraction should be doubled to nearly 60% (i.e., seeding was done during 60% of the time in which suitable clouds existed).

Dispersion of the seeding material

Ignoring the previous work of Gagin and Aroyo (1985), RH made their own calculations of the dispersion of the seeding material. Gagin and Aroyo have calculated the concentration of AgI particles released from a single seeding aircraft along the seeding line under typical conditions. According to their dispersion model, concentrations of 10 L⁻¹ are reached 50% of the time at the distance of 25 km downwind from the seeding line. The calculations of RH are not more realistic than those of Gagin and Arroyo (1985) because if 99% of the material remains within 2.1 km of its release level (cloud-base height), it follows that 99% of the air at cloud-base level would remain at the same height interval. Had this been the case, no convective rain would have been possible because the convection is fed through cloud bases. In reality, seeding material is convected vertically very fast to both the clouds and the boundary layer. This description is supported by recent modeling of Levin et al. (1997). Very often the orography induces moderately deep convective clouds, and thus the seeding agent can be convected efficiently into the orographic cloud tops.

Rangno and Hobbs claim that seeding material is not likely to be ingested efficiently into convective clouds, quoting the observations of Stith et al. (1986, 1990) that cloud-base seeding results in thin filaments of seeding material in the clouds. However, there is a fundamental difference between direct cloud-base seeding and seeding along a line upwind the target clouds. In the latter case, the seeding material has time to spread in the turbulent sub-cloud-base layer and even recirculate with the low-level return flow (See Khain et al. 1993). This way the seeding agent can be ingested into a large number of clouds that grow from the sub-cloud-base level, and the material is not likely to be confined to thin filaments.

Effectiveness of the seeding material

Rangno and Hobbs calculated that about 10⁶ m³ of water were attributed to rain enhancement due to each gram of released AgI. They consider this ratio as unreasonable.

Rangno and Hobbs used 15% as the factor of rain enhancement, but the overall added rainfall was only half of that amount because only half of the days were seeded randomly. They assumed that only 1% of the seeding material finds its way to the clouds. But a more realistic number is 10%, considering the convective nature of the atmosphere and the large downwind area that is considered in this calculation. Therefore, the revised calculation would be 50 000 added cubic meters of water per gram of AgI. Dividing the 50 000 m³ into the 5 × 10¹³ particles produced by each gram of activated AgI yields 5 × 10¹³ water droplets with radii of only 0.2 mm—barely the size of a raindrop. Even when taking RH’s assumption that only 1% of the seeding material finds its way to the clouds, the radius of an AgI-induced rain droplet would grow to only 0.43 mm. The calculation leads to the conclusion that the added rain amount is compatible with the amount of the seeding agent, which is the opposite of RH’s conclusion.

The correlation between the north and center

Rangno and Hobbs quoted Gabriel (1966) that the correlation between daily rainfall in the northern and central target areas was about 0.80. They used this correlation as evidence for the similarity in the rain regimes in the two target areas. However, 4 years later Gabriel (1970) published the final analyses of Israeli I, where the correlations are 0.61 for the northern seeded days, 0.69 for the central seeded days, and 0.72 for the period 1949–60. The latter quote of Gabriel should be considered the definitive one. Rangno and Hobbs quote Gabriel (1970) elsewhere in their paper.

Wasting seeding on dry days

Rangno and Hobbs questioned the efficacy of the seeding based on operational considerations. They quoted LeCam and Neyman (1967) as evidence that much of the operational effort was wasted on days with very light rain. However, according to Table 3 of Gabriel (1970), the daily mean precipitation in the northern target area on actually seeded days was 13.4 mm day⁻¹, while the rainfall on days allocated to seeding but not actually seeded was only 2.4 mm day⁻¹. The respective means on days allocated for seeding in the center were 11.4 and 3.2 mm day⁻¹. These numbers show that most of the seeding effort was conducted during the rainiest situations (rainiest naturally or due to seeding).

The question of lucky draw

Rangno and Hobbs suggest that a type I statistical error (a lucky draw) is the best explanation for the positive results of the Israeli I experiment. They quoted Wurtele (1971), as if she suggested the same. However, Wurtele (1971) ascribed the lucky draw possibility only to the hypothesis that rainfall in the buffer zone (BZ, separating the northern and central target areas) was enhanced when the BZ seeded on central seeded days was considered. Since it was not postulated a priori that rainfall should be enhanced in the BZ on central seeded days, Wurtele (1971) is irrelevant to the subject.

Even according to the explanations of RH themselves, the crossover design should take care for a systematic bias or a lucky draw, as it did, according to their suggestion, for the Israeli II experiment. Therefore, the crossover overall result of 15% for Israeli I is also unlikely to be a type I error (according to the logic of RH). This leads to a logical contradiction in the arguments of RH and the invalidation of their conclusion that both the results of the Israeli I and Israeli II experiments were merely lucky draws.

The “enigma” of the buffer zone

Gagin and Neumann (1974) and Wurtele (1971) pointed out that the large (+31%) indicated SAR in the western part of the BZ was contributed mainly from days with southwesterly wind. Rangno and Hobbs stated that “inadvertent seeding of the BZ was extremely unlikely.” This means that rain in the BZ was not really affected by the seeding, in spite of the large SAR on central target area seeded days. In such a case, it is very likely that the indicated positive effect in the center was also due to a lucky draw. Rangno and Hobbs also contend that in this case, due to the nature of the crossover design, the seeding effect in the north would be underestimated. If this is true and if seeding did not affect the BZ, then we should subtract 31% from the seeding effect in the center and add 31% to the seeding effect in the north, obtaining a corrected SAR of 1.08 × 1.31=1.41 for the north and an SAR of 1.22/1.31 = 0.93 for the center. The overall effect (RDR, root double ratio) remains unchanged: (0.93 × 1.41)^0.5 = 1.15. Interestingly, this is similar to the H₊₋ hypothesis that was obtained for Israeli II, where large significant positive effects were obtained for Israeli II in the north and smaller insignificant negative effects in the south (Gabriel and Rosenfeld 1990).

Having reached this conclusion, RH contradict themselves in the next sentence, writing, “However, this was not the case; instead there were strong gradients in rainfall. Brier et al. (1973) reported that on center-seeded days rainfall was unusually heavy in western Jordan relative to that in Southern Lebanon; . . . On north target area-seeded days, Brier et al. found the opposite pattern. . .” But such changes in the rainfall distribution were the exact objective of the cloud seeding in a crossover design. Brier et al. (1973) analyzed only areas that were downwind of the seeding lines. Israel is small enough (then even smaller) so that areas in Lebanon, Syria, and Jordan were not far downwind of the seeding lines. According to the map of the effects made by Brier et al. (Brier et al. 1973, their Fig. 6), the areas that they analyzed extended from about 20 to 120 km downwind of the seeding lines. There is no obvious reason for the seeding effect to diminish within such a distance. If anything, the orographic regeneration of relatively short living clouds to the east of the Jordan Valley (about 60–70 km to the east of the coastline) provides a plausible physical reason for the regeneration of the seeding effect there as well. Therefore, Brier et al. (1973) attributed these changes in the rainfall distribution to the seeding effect. Figure 6 of Brier et al. is reproduced by RH as their Fig. 14. However, the reproduced figure is distorted, such that the indicated seeding effects are posted farther downwind than in the original figure. This distortion may create the incorrect impression that positive downwind effects extend to greater distances downwind than those analyzed by Brier et al. (up to 70 miles, according to Brier et al. 1973).

Seeding effects in the coastal areas

Rangno and Hobbs wrote about the Israeli I experiment that “it is extremely unlikely that rainfall at the surface in the coastal regions, which were so close (≤15 km) to the line of seeding at cloud base, could have been affected by seeding,” because of the strong westerly winds between cloud base and the −15°C level. However, they were unaware of the following observations, which are yet not published elsewhere.

In the Israeli coastal plain the heavy rainfall is typically associated with the following.

• There is weak inversion or stable layer near the surface.

• There is a markedly increased relative humidity in the coastal boundary layer air.

• There is a component of return flow to the sea near the surface.

• At the same time, there are strong westerly winds at the level of the hills. One can drive from Tel Aviv to Jerusalem and experience reversal of the winds. I experience it in more than half of the times that I travel between Tel Aviv and Jerusalem on rainy days.

• The interaction of the return flow with the westerly flow over the open sea is manifested by a marked lowering of cloud bases some distance west of the coastline. This is observed nicely during the research flights that I conducted. The typical lowering of cloud base occurs from about 3000 ft over the open sea to about 1500 ft at a distance of about 5–20 km to the west of the coastline. The lower clouds are not connected to the higher bases except under active deep convective areas.

• In such conditions, radar analyses show a marked invigoration of clouds toward landfall, peaking at the coastline.

Such a typical situation was the subject of a modeling study (Khain et al. 1993) that replicates the main observed features above. The modeling also suggests that much of the cooling of the coastal boundary layer is due to the downdraft and evaporative cooling from the precipitation, even during daytime. The precipitation has a positive feedback to maintain this dynamic structure in such situations. Seeding material could easily be washed down with the downdraft and recirculated into the clouds to the west of the coastline.

Rangno and Hobbs further doubted the seeding effect in the coastal plain, relying on the ice particle measurements of Gagin and Neumann (1981). But they questioned these measurements in the previous section of the same manuscript. They cannot have it both ways.

Seeding effects in northwesterly and southwesterly winds

Rangno and Hobbs acknowledged the existence of seeding effect on days with northwesterly wind, as apparent in their Table 1. However, they dismissed its importance by writing that “the occurrence of northwest flow with suitable clouds does not appear to be frequent enough (81 days in the six and a half seasons of Israeli I), and the rainfall is too little and often too spotty, to provide a practical target. . .” Even if these arguments of RH were correct, they would be irrelevant. The issue of their paper is whether seeding did enhance rainfall in Israel or not. Only after this is established should it be considered whether the actual seeding is worthwhile.

This situation is apparent in the way that RH discuss the seeding effect in northwesterly wind days. After accepting that statistically rain seems to be enhanced during northwesterly wind days and dismissing its practical importance, RH dismissed the possibility that rain was actually enhanced in northwesterly winds by mentioning again the high ice particle concentrations that were reported by Levin in northwesterly midlevel flow. In addition to these inconsistencies, RH’s arguments can be challenged for the following reasons.

• The number of rainy days with northwesterly wind is not negligible at all. According to Gagin and Neumann (1974), who RH quote in their Table 1, there were 81 northwesterly wind days out of 324 days with wind data, which is 25% of the days.

• According to Fig. 3 of RH, more than half of the rainy days have northwesterly wind at 850 mb.

• Significant rain amounts occur in Israel under these conditions. In fact, the average daily rainfall in northwesterly winds (7.8 mm day⁻¹) was greater than in southwesterly winds (7.5 mm day⁻¹) in the central target area during the experimental days of the Israeli I experiment (Rosenfeld 1989).

• The ice particle measurements were already discussed in the microphysics section. These measurements are too few to warrant any general conclusion. Moreover, southwesterly surface winds blew when large ice particle concentrations were observed. It should be kept in mind that it is the surface air that is ingested to convective clouds, which are formed by surface heating.

The selection of the control area

Rangno and Hobbs state that “if the rainfall on the same day in the alternate target area had been used to evaluate the result of seeding in each target area, this misperception [the actually indicated seeding effects] would not have occurred.”

In fact, RH suggest analyzing the north with the south as its control area and vice versa. But this can be done only if the control area is left unseeded in all days, which is not the case in the crossover component of the Israeli II experiment. This suggestion shows that RH incorrectly interpreted the principles of the experimental design of the experiment that they criticize.

The completeness of the analyses

Rangno and Hobbs state that “Chief among these [the problems with Israeli II] is that a complete analysis of Israeli II that is consistent with the original design of the experiment (i.e., crossover design), has not yet been carried out.”

According to the design document of Israeli II, this statement is incorrect. Israeli II was conducted in a dual-mode crossover with the south and a single area with the upwind control. To comply with the analysis requirements of both experiments with the same dataset, the BZ rainfall had to be used as a qualifier for both aspects of Israeli II.

Gagin and Neumann carried out the full target–control analysis, and Gabriel and Rosenfeld (1990) completed the full crossover component of the analysis. All days that had measurable rain in the BZ in the experimental season were included in the analyses, as required by the design document. We checked the sensitivity of the results to the definition of experimental days. Inclusion of all the 614 days in which any rainfall was recorded in any of the rain gauges in the north or south targets, the buffer and control areas, resulted in DR=1.123 at the north target area. This indicated seeding effect is virtually identical to that calculated by Gagin and Neumann (1981) and by Rosenfeld and Farbstein (1992). This shows that the results are insensitive to the added nonexperimental days.

Rangno and Hobbs mention that “only a portion of the available IMS gauges has been used in statistical evaluation of Israeli II.” They imply that results could have been affected by the selection of the gauges. In fact, the criteria for selection of the rain gauges were

• the historical continuity and technical reliability of the rain gauges and

• an even areal coverage density within the availability constraints of such rain gauges.

The sensitivity to the selection of rain gauges was tested by reanalysis of the Israeli II north experiment with all of the IMS operational rain gauge data. The inclusion of all gauges, without any quality control, introduced much noise due to broken gauge records, uneven distribution of rain gauges, and daily rainfall at different days obtained by different gauges. In spite of the random noise, which by nature reduces the detectability of an effect, the DR for the north was still 1.101.

The question of natural bias with respect to orographic rainfall

Rangno and Hobbs suggested that the statistical results of the Israeli II experiment are a result of natural bias in favor of orographically enhanced rain inland during the northern seeded days. This hypothesis was already investigated earlier by Gabriel and Rosenfeld (1990) and Rosenfeld and Farbstein (1992, hereafter RF92), who used parameters of Bet Dagan radiosondes as covariates in a regression model. The model was used to separate the orographic component from the overall indicated seeding effect, which still remained substantial. Following are RH’s arguments against the validity of this analysis and the discussion of the validity of RH’s arguments.

• “On many occasions, rawinsondes from Bet Dagan are unlikely to provide moisture profiles relevant to the north target area.”

Rangno and Hobbs base this claim on a comparison with the soundings of Beirut, although the northern target area is halfway between Bet Dagan and Beirut, with the average distance of only 1° latitude to the north. Therefore, the Bet Dagan sounding is still very useful to represent the conditions in the northern target area.

• “The time scale of the rain cloud systems is finer than the 12 hours between consecutive rawinsondes used in the model of Rosenfeld and Farbstein.”

The predicted value was a 24-h integrated rainfall, so that two soundings were available for each rainfall integration period. This problem also concerned Rosenfeld and Farbstein when they formulated their model. They tested the sensitivity of the time resolution using data from several years for which 6-h radiosonde data were available and found negligible differences in the model’s skill to predict the daily rainfall in the target areas.

• “We doubt that this model can distinguish between rainfall in either the two target areas of Israeli II sufficiently well to elucidate any effects of cloud seeding . . . The difficulty of accurately predicting the locations and amounts of rainfall in convective situations, even in the most sophisticated numerical models is testimony to the difficulty of the task.”

Orographic effects were suspected already by Gabriel and Rosenfeld (1990), Rosenfeld and Farbstein (1992), and now also by RH, as the cause for the indicated seeding effects. However, the model of Rosenfeld and Farbstein was aimed mainly at the orographic component of the rainfall. The use of radiosonde-derived parameters for evaluation of rain enhancement in mountainous terrain was first used by Elliott and Shaffer (1962). They found remarkably high predictability of the rainfall in the coastal ranges of California. This area shares many geographical and meteorological similarities with the experimental areas in Israel. Alpert and Shafir (1989) have shown that a similar model is very skillful in predicting rainfall in the hilly areas of Israel. Shafir and Alpert (1990) used such a model to detect urban effects on rainfall in Jerusalem.

Two radiosondes per day are adequate because the decorrelation time of the precipitation echo-top height, as was measured by radar in Israel during 11 full rainy seasons, is 12 h. This means that each radiosonde can represent at most a time interval of 24 h. With two radiosondes per day, each is used for half the decorrelation interval only.

At the bottom line, the accuracy of the predictions is judged by the correlation between the prediction and the actual rainfall. The regression coefficients ranged between 0.72 and 0.77 (Rosenfeld and Farbstein 1992), which are greater than the correlation between the northern and the southern target areas. Any systematic error of the model prediction due to inaccuracies would be compensated for by the fact that only the ratios of the model predictions are used.

Seeding effects beyond the borders of Israel

Rangno and Hobbs provide in their Fig. 17 the seeded/no-seeded ratios (SAR) for several stations in Israel, west Jordan, southwest Syria, and south Lebanon. According to the distribution of the SARs, RH show that rainfall was heavier on northern seeded days in the whole region, except for the upwind control area of the northern target area. Based on that, they conclude that the rainfall was also heavier naturally in the seeded days in the northern target area. However, the SAR was never regarded as an appropriate measure of systematic fluctuations in rainfall amounts. Furthermore, under seeding operations the SAR cannot be used to represent natural fluctuations. The only such measure is a valid double ratio.

In fact, RH performed in this questionable fashion an alternative exploratory a posteriori analysis of the Israeli II experiment, which provides results opposite to the a priori designed analysis for the north. Then, they reject the results of the a priori formal analysis in favor of the results of their a posteriori analyses. They do so while not accepting Gagin and Neumann’s (1981) analyses for the north, which were, according to RH’s erroneous perception, exploratory a posteriori analyses. It is the opinion of the author of this comment that one cannot reject in principle a posteriori analyses of others and then suggest that his a posteriori analysis is the correct one.

Rangno and Hobbs used very few stations for each of their areas, as shown in their Fig. 17. They used only three unspecified stations for the northern target area. They did not specify the number of stations selected for the other areas. Here again RH do something that they criticized in the Israeli experiments. They questioned the selection of the rain gauges for analyses of the Israeli experiments, implying that this selection could have changed the resultant indicated seeding effect. This double standard does not add credibility to RH’s interpretation of the Israeli II experiment.

The above remarks suggest that RH’s analyses, based on their Fig. 17, may not be correct. Additionally, these analyses can be questioned, as follows.

Rangno and Hobbs compared the SARs in target areas with those in potential control areas, which is equivalent to a double ratio analysis. In fact, RH provided a series of double ratio analyses for the Israeli II northern target area with alternative control areas. But there are several requirements for a potential control area to be acceptable.

• The rainfall in the control and target areas should be highly correlated. This condition is fulfilled only with respect to the northern coast and southern Lebanon.

• The control area should be unseeded in all days. But the areas in central and southern Israel and Jordan were seeded during the no-seed days in the north. The single station used in southern Lebanon (Marjayun) is downwind of the seeding line during conditions of southwesterly winds, which are the dominant low-level winds during rainy conditions there.

These requirements leave no alternative control area for the Israeli II north, except for its formal upwind control area at the northern coastal plain. The south target area does not have such a control area, so that its separate DR cannot be assessed.

Summary of the Israeli II statistical analyses

The Israeli II experiment was fully analyzed in accordance with its experimental design. The results are in line with those of the Israeli I experiment, in spite of Israeli II not being strictly confirmatory for Israeli I. This strengthens confidence that the indicated positive seeding effects in the north are not due to chance beyond the formal significance level.

Effects of dust on clouds microstructure

“There is no evidence that there are two microstructurally very different cloud types in Israel. . .”

Rosenfeld and Gutman (1994) developed a method to obtain information about cloud-top microphysical structure from multispectral analyses satellite data, using the Advanced Very High Resolution Radiometer onboard National Oceanographic and Atmospheric Administration satellites. They have shown that cloud-top particles’ effective radii (r_eff) greater than 14 μm were a good criterion for delineating precipitating clouds.

Levi and Rosenfeld (1997, manuscript submitted to J. Appl. Meteor.) applied this methodology to clouds in Israel and showed fundamental differences between the microphysical structure of cloud tops occurring in desert dust and in clean air. Tops of clouds that grew in dusty air developed large particles and started to precipitate at much warmer cloud-top temperatures (−10°C on the average, according to Fig. 2) than clouds growing in clean air (−16°C). Furthermore, Levi and Rosenfeld (1997, manuscript submitted to J. Appl. Meteor.) have shown that the same differences existed between clouds growing in southwesterly and northwesterly 850-mb flow (see Fig. 2).

An alternative explanation to the difference in cloud-top r_eff could have been differences in cloud-base temperature. Clouds with warmer bases have larger particles for a given cloud-top temperature when all other parameters are held constant. However, the averaged cloud-base temperatures (T_b) for the southwesterly flow cases were colder (T_b = 6.01°C; SD = 4.05°C) than T_b at the northwesterly flow cases (T_b = 7.37°C; SD = 4.31°C). This leaves the difference in aerosols in northwesterly and southwesterly flow as the most plausible explanation for the differences between cloud particle size and precipitation properties.

Interactions of dust with cloud systems

“RF92 present no evidence that dust and/or haze from desert was transported into the cloud systems . . . on days in which dust and/or haze affected the clouds . . .”

Levi et al. (1996) have shown a significant difference between the amount of materials of continental origin washed down with the rain on “dust” and “no dust” days. The chemical composition of the continental material indicates its source from the deserts. Rosenfeld and Nirel (1996) have shown that the “coastal front” mechanism can cause desert dust to be imported at low levels into the southern margins of the rain cloud systems, even when the back trajectory of the flow at 850 mb and above does not backtrack to the deserts. In fact, some of the microphysical measurements quoted from Levin (1992) took place in such conditions.

The effect of dust on rain enhancement

“The hypothesis that dust and/or haze negated seeding effects in the south target area of Israeli II . . . runs counter to the apparent results of Israeli I.”

It has been shown already in section 4d that the most likely explanation to the BZ indicated positive seeding effect is that seeding, in fact, enhanced rainfall only in the north, with no enhancement in the center. This makes the Israeli I and Israeli II experiments consistent with each other with respect to the desert dust hypothesis.

The usefulness of SAR analyses

“RF92 do not show point values of SAR’s in Israel on days when the north target area was seeded and for days which they claim large increases of rain due to seeding.”

The necessity of an unaffected control area for a credible evaluation is well established. The SAR analyses of RH have been shown already in section 6d to be invalid.

Why was the dust effect on clouds not detected sooner?

“Why did so many earlier studies in Israel . . . fail to detect the profound effect of dust and/or dust haze on the microstructure of clouds and on rainfall in Israel?”

Ice nuclei concentration is only one of many factors determining the concentrations of ice particles in clouds (Rangno and Hobbs 1988). Perhaps the most important factor is the time that is available for glaciation to progress in the cloud. This and other factors may well have masked the microphysical signature of the desert dust in past research.

Even so, these effects did not go completely undetected in the past. Gagin had shown already in 1965 that enhanced concentrations of ice nuclei are associated with intrusions of desert dust, which is imported with southwesterly winds. Gagin and Neumann (1974) had already noticed the effects, related here to dust, on the results of the Israeli I experiment (see section 4d).