1. Method
The main claim of Alpert et al. 2008 (hereinafter AHL08) is that Givati and Rosenfeld (2004, 2005, hereinafter GR04 and GR05, respectively) used rain gauges selectively to obtain a decreasing trend of the ratio between hilly (called “mountain” in AHL08) and plains (called “shore” and “inland” in AHL08) rain gauges. AHL08 used rain gauges different from those used by GR04 and GR05 to show an increasing trend for central and northern Israel (in AHL08’s Figs. 3 and 4, respectively). AHL08 showed that using different gauge selection methods gave opposite results, and so questioned the validity of the results of GR04 and GR05. AHL08 used all available records, whereas GR04 and GR05 used pairs of gauges or gauge clusters that have long enough records and maintain a high correlation between them. To resolve this selectivity question, we did the following:
We composed combined tables, separately for central (Table 1) and northern (Table 2) Israel. Each of these combined tables include all of the gauges used by AHL08 for their Figs. 3 and 4 that indicated increasing tends of Ro and the gauges used by GR04 and GR05 for their figures that indicated decreasing trend of Ro.
We paired all possible combinations between these hill and plains gauges, separately for the north and for the center, and retained only the pairs for which at least 30 yr of data from both rain gauges are available. All of the possible pairs, their correlation, Ro, and the slope of Ro are tabulated in Tables 3 and 4 for central and northern Israel, respectively.
We classified the paired rain gauges according to the correlation coefficient R between their annual rainfall into three groups: R ≥ 0.9, 0.9 > R ≥ 0.8, and R < 0.8.
According to Fig. 2 of AHL08 the probabilities for the trends in the orographic enhancement factor Ro were random in central Israel. Recalculating the probabilities with the same data when applying the objective selection criteria of correlation and duration of measurements showed that decreasing slopes of Ro dominate the highly correlated pairs of gauges that recorded data for long periods.
The trends between the hill and lowland rainfalls were calculated for the three correlation classes, separately for north and central Israel. Additional classification was done as a function of distance eastward from the coastline, to account for the decay of the convectiveness of the clouds with distance from the sea inland or for other possible factors that may depend on the distance from the sea. The distributions of the trends from all of the paired gauges are displayed the same way as in AHL08’s Fig. 2.
2. Results for central Israel
a. Is the probability for trends in Ro over Judea and Samaria random?
Figure 2b of AHL08 suggests a random probability for Ro when all gauges are paired. However, valid pairs of rain gauges require that they will be well correlated and also comeasured for a sufficiently long period. AHL08 did not apply any such test. However, where should we put the threshold for correlation between the pairs of gauges and for the number of years that they cover? To avoid an arbitrary cutoff, a range of these thresholds was applied for the trends of Ro, as shown in Fig. 1.
Figure 1a is composed of 181 pairs of gauges that had at least 20 yr of common measurements. According to Fig. 1a, 94% of the pairs with correlations R > 0.90 had negative slopes of the correlation of Ro with time. For 0.9 < R ≤ 0.8, 62% of the slopes were negative. For R < 0.80, this number falls to 52%, which means practically a random sign for the slopes. To test the impact of observational period, Fig. 1b selects the 107 pairs with at least 30 common years. This reduced the spread of the values of the slopes mainly for 0.9 < R ≤ 0.8, and increased to 78% the fraction of negative slopes. The slopes of the pairs with R < 0.8 remained random.
In summary, according to Fig. 1 the pairs with the lowest correlation and measuring period replicated the results of AHL08 and showed a random distribution of trend in Ro. However, the pairs with higher correlations and longer periods had smaller scatter of Ro and converged to negative slopes as found by GR04 and GR05.
b. Reevaluating the trends in Ro over Judea and Samaria
Figure 3 of AHL08 and Fig. 6b of GR04 present opposite trends of Ro for the same geographic region. Which one of these figures, if any, presents the correct trend? To examine this question, all of the rain gauges used for constructing the trends of Ro in central Israel by both AHL08 and GR04 were combined into Table 3. These pairs of rain gauges were analyzed and displayed in Figs. 2a,b the same way as was done for Fig. 1.
Figure 2 shows the accumulated probabilities of all trends from the most negative slope upward, for pairs of stations used by AHL08 and GR04 in central Israel. There were in all 350 pairs, out of which 342 and 222 pairs exceeded 20 and 30 common measurement years, respectively. As in Fig. 1, it can be seen that negative trends occur for 90% of the hilly–plains pairs with correlation of R ≥ 0.90 and common measuring period of at least 30 yr. Upon lowering the correlation range to 0.90 > R ≥ 0.80, only 75% of the pairs have decreasing slopes. For the cases with R < 0.80, the slopes appear to be completely random.
The trend in Ro was recalculated using all the pairs that had R ≥ 0.90 and period ≥30 yr. The Ro for each year was calculated as the sum of the hill rainfall divided by the sum of the plains rainfall, taken from all pairs that passed the selection criteria. The result is presented in Fig. 3, which replicates the indicated decreasing trend in Ro as reported by GR04.
c. The role of distance of the plains gauges from the coastline
AHL08 suggested that any indicated decreasing trend in Ro would be an artifact due to a relative increase of the rainfall in the inner plains in comparison with the coastline, possibly caused by effects of the coastal urbanization. They further suggested that the trend of hill/plains rainfall should be increasing, in contrast with the indications of GR04. This question is addressed here by classification of the pairs according to the distance D of the plains rain gauge eastward from the coastline in kilometers.
The pairs that passed the criteria of R ≥ 0.90 and at least 30 yr (used for composing Fig. 3) were further classified according to D, as shown in Fig. 4a. This figure shows that almost all of the pairs produced negative trends, regardless of the value of D. However, the requirement of R ≥ 0.90 left too-few pairs for a robust conclusion. The number of pairs can be seen by the number of dots on the lines in Figs. 1, 2, and 4. Therefore, it was necessary to relax the correlation to R ≥ 0.85, shown in Fig. 4b.
According to Fig. 4, especially Fig. 4b, the probability for more negative trends was obtained when the hill gauges were paired with plains gauges that were farther eastward from the coastline. This is consistent with the suggestion of AHL08 that rainfall has increased in the inner plains with respect to the coastline. However, Fig. 5a shows that there is still a decreasing trend of Ro even when calculated against the shore rain gauges alone (i.e., for the class of D < 10 km in Fig. 4b). This trend (Fig. 5a) is half of its magnitude when Ro is computed with respect to the rain gauges at D > 20 km from the coastline (Fig. 5c).
3. Results for northern Israel
a. Is the probability for trends in Ro over the upper Galilee random?
The method that was applied to central Israel was applied also to the rain gauge pairs that were used by AHL08 and GR05 for northern Israel. The rain gauges are the same as specified in the legend of Fig. 4 of AHL08. All of the possible pairs between plains and hill gauges are shown in Table 4. As for central Israel, we selected also for the north only pairs with measurements of the same years for at least 30 yr. According to Fig. 6a more than 80% of the pairs had negative trends of Ro.
b. The meteorological significance of distance from the coastline
According to AHL08, a major issue in central Israel was the possible changes in the plains rainfall with distance inland from the coastline, due to possible anthropogenic-induced changes in the convective clouds that moved inland from the sea. This could not be an issue in northern Israel, because all of the plains rain gauges were at D < 10 km from the coastline. Instead, the short distance of part of the hill rain gauges to the coastline (five pairs were at D < 20 km) presented another issue of convective clouds that formed over the sea contributing significantly to the hill rainfall. This was not as much an issue for the hill gauges in central Israel, because D was larger than 30 km for all of them.
The distance of the hill gauges from the coastline matters because the convective clouds of the winter rainstorms in Israel are energized by the heat flux from the sea surface to the colder air mass arriving from Europe (Shay-El and Alpert 1991). The convective clouds mature when they move inland and lose their energy source. This is evident by the rain intensities decreasing from the coastline inland, along a distance of about 30 km eastward from the coastline (Sharon and Kutiel 1986). The frequency of thunderstorms also behaves similarly (Yair and Levin 1994). Rosenfeld (1986) showed that the radar-detected echo-top heights average, in Israel, about 5 km MSL. The top heights decrease from 5.4 km over the sea by about 400 m when moving from sea to the coastal plains and decrease by an additional 400 m over the hills. Such deep convective clouds have smaller susceptibility to aerosol effects on rainfall amounts than do the shallower orographic clouds (Phillips et al. 2002) that form when the air ascends over the hills inland.
The western parts of the hilly areas, which are close to the sea, are likely rained over by the maturing convective clouds whose response to changing meteorological and aerosol conditions is different from that of the orographic clouds, for which the rainfall is more significant over the farther inland hills. Therefore, the stratification of the effects by distance from coastline is important for gaining insights to the possible causes of the indicated trends.
c. The dependence of trends on distance from the coastline
There were only four pairs of gauges in Table 3 with R > 0.90 that had a record of at least 30 yr. Relaxing R to 0.85 still left too-few pairs for additional classification by distance of the hill gauge eastward from the coastline. According to Fig. 6b, the number of pairs that survived the selection criteria was 15 for D ≤ 15, 7 for 15 < D ≤ 30, and 6 for D > 30. Relaxing the threshold further to R > 0.80 allowed 22, 13, and 29 pairs to enter the three D classes, respectively. Comparing Figs. 6b and 6c shows that the relaxation of R was at the expense of increasing scatter of the probabilities of Ro. Using either threshold did not change the essence of the results, and therefore we used the lower threshold for making them representative for the larger number of gauges.
The pairs of rain gauges were combined for the three D intervals, as was done for the center, shown in Fig. 7. The results show a slight, statistically insignificant increase in Ro for the hills at D ≤ 15 km from the sea, a similar decreasing trend in Ro for 15 < D ≤ 30, and a significant decreasing trend for D > 30 km. GR05 showed that farther east of the upper Galilee, in the northern Jordan Valley where the precipitation is not orographic, Ro no longer decreases.
4. Discussion and conclusions
Reanalysis of the trends in Ro using an objective selection method of the pairs of rain gauges showed that the probabilities of trends in Ro are not random, in contrast to the assertion in Fig. 2 of AHL08, but rather are clearly negative over the hills in both central and northern Israel. A complicating factor is the weak trend of increased rainfall some distance from the sea inland, regardless of the topography. However, beyond the range of that effect, at D > 30 km inland, Ro decreased with respect to the plains rain gauges, in both central and northern Israel. The trends in the orographic precipitation over the hills at D < 30 km from the coastline are partially masked by the convective clouds that move inland from the sea. At greater distance, the decreasing trend in Ro is evident even when comparing with the coastal rain gauges (D < 10 km from the shore line), especially in the north. The strong decreasing trend of Ro at the eastern half of the upper Galilee is tied to the hills and disappears farther east in the Jordan Valley.
AHL08 ascribed the trend of increased rainfall some distance from the sea inland to the hypothesis of an urban heat island effect. However, there is greater support for the hypothesis that a trend of increasing aerosols that are imported from eastern Europe with the rain-bearing air mass has caused the convective clouds over sea to delay their precipitation and redistribute it farther eastward of the coastline. Teller and Levin (2006), based on their cloud simulations, showed that adding cloud condensation nuclei (CCN) to clouds that are typical Israeli winter storms decreases their rainfall amounts and delays their peak intensity by about 20 min. According to their Fig. 4, this redistributes the rainfall and causes a large enhancement of the rainfall a distance of about 20 min downwind. Khain et al. (1999) simulated the effects of added CCN aerosols on clouds near the coastline of Israel. Their simulations showed that the main cause of the convection is the contrast between the cold air and the warm seawater. They showed that the added aerosols delayed the rainfall and redistributed it from the sea to a few tens of kilometers inland.
A likely cause of the decreasing trend of orographic precipitation in Israel remains the aerosols, as speculated by GR04 and GR05. However, the possible impacts of other meteorological factors, such as changes in the instability profiles, cannot be excluded. This important question should be investigated further.
Acknowledgments
This study was partially funded by CIRCE (Climate Change and Impact Research: The Mediterranean Environment) of the Commission of the European Union (http://www.circeproject.eu/).
REFERENCES
Alpert, P., N. Halfon, and Z. Levin, 2008: Does air pollution really suppress precipitation in Israel? J. Appl. Meteor. Climatol., 47 , 933–943.
Givati, A., and D. Rosenfeld, 2004: Quantifying precipitation suppression due to air pollution. J. Appl. Meteor., 43 , 1038–1056.
Givati, A., and D. Rosenfeld, 2005: Separation between cloud-seeding and air-pollution effects. J. Appl. Meteor., 44 , 1298–1315.
Khain, A. P., A. Pokrovsky, and I. Sednev, 1999: Effects of cloud-aerosol interaction on cloud microphysics, precipitation formation and size distribution of atmospheric aerosol particles: Numerical experiments with a spectral microphysics cloud model. Atmos. Res., 52 , 195–220.
Phillips, V. T. J., T. W. Choularton, and A. M. Blyth, 2002: The influence of aerosol concentrations on the glaciation and precipitation of a cumulus cloud. Quart. J. Roy. Meteor. Soc., 128 , 951–971.
Rosenfeld, D., 1986: The dynamic characteristics of cumuliform clouds and cloud systems and their effect on the rainfall precipitated by them. Ph.D. thesis, The Hebrew University of Jerusalem, 142 pp.
Sharon, D., and H. Kutiel, 1986: The distribution of rainfall intensity in Israel, its regional and seasonal variations and its climatological evaluation. Int. J. Climatol., 6 , 277–291.
Shay-El, Y., and P. Alpert, 1991: A diagnostic study of winter diabatic heating in the Mediterranean in relation to cyclones. Quart. J. Roy. Meteor. Soc., 117 , 715–747.
Teller, A., and Z. Levin, 2006: The effects of aerosols on precipitation and dimensions of subtropical clouds; a sensitivity study using a numerical cloud model. Atmos. Chem. Phys., 6 , 67–80.
Yair, Y., and Z. Levin, 1994: Lightning disintegration in clouds and into the ground in thunderstorms in Israel. Meteor. Isr., 3 , 20–28.
The cumulative probabilities of trends in Ro exceeding the value in the abscissa, for all pairs of Judea and Samaria rain gauges with the plains gauges, for the gauges used to compose Fig. 2 of AHL08. The gauges are classified according to three groups according to their correlation coefficient R, as indicated in the legend. Shown are the results for pairs of gauges with comeasured rainfall for at least (a) 20 and (b) 30 yr.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The cumulative probabilities of trends in Ro exceeding the value in the abscissa, for all pairs of Judea and Samaria rain gauges with the plains gauges, for the gauges used to compose Fig. 2 of AHL08. The gauges are classified according to three groups according to their correlation coefficient R, as indicated in the legend. Shown are the results for pairs of gauges with comeasured rainfall for at least (a) 20 and (b) 30 yr.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The cumulative probabilities of trends in Ro exceeding the value in the abscissa, for all pairs of Judea and Samaria rain gauges with the plains gauges, for the gauges used to compose Fig. 2 of AHL08. The gauges are classified according to three groups according to their correlation coefficient R, as indicated in the legend. Shown are the results for pairs of gauges with comeasured rainfall for at least (a) 20 and (b) 30 yr.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for all pairs of rain gauges used by AHL08 and GR04 for their calculations of trends in Ro in central Israel. The individual values are provided in Table 3. Note the decreasing random scatter and focusing on negative values for greater R and the longer measuring period.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for all pairs of rain gauges used by AHL08 and GR04 for their calculations of trends in Ro in central Israel. The individual values are provided in Table 3. Note the decreasing random scatter and focusing on negative values for greater R and the longer measuring period.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for all pairs of rain gauges used by AHL08 and GR04 for their calculations of trends in Ro in central Israel. The individual values are provided in Table 3. Note the decreasing random scatter and focusing on negative values for greater R and the longer measuring period.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R > 0.90 and at least 30 yr of paired rainfall data, from Table 3. The inset P value is the probability that the slope of the regression line is not different from zero. The decreasing linear trend is statistically significant at P = 0.001 using the Kendall nonparametric two-tail test.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R > 0.90 and at least 30 yr of paired rainfall data, from Table 3. The inset P value is the probability that the slope of the regression line is not different from zero. The decreasing linear trend is statistically significant at P = 0.001 using the Kendall nonparametric two-tail test.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R > 0.90 and at least 30 yr of paired rainfall data, from Table 3. The inset P value is the probability that the slope of the regression line is not different from zero. The decreasing linear trend is statistically significant at P = 0.001 using the Kendall nonparametric two-tail test.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the data from Table 3 that passed the criteria of at least 30 yr and (a) R ≥ 0.9 or (b) R ≥ 0.85. The rain gauge pairs are further classified according to the distance D of the plains rain gauge eastward from the coastline, in kilometers.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the data from Table 3 that passed the criteria of at least 30 yr and (a) R ≥ 0.9 or (b) R ≥ 0.85. The rain gauge pairs are further classified according to the distance D of the plains rain gauge eastward from the coastline, in kilometers.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the data from Table 3 that passed the criteria of at least 30 yr and (a) R ≥ 0.9 or (b) R ≥ 0.85. The rain gauge pairs are further classified according to the distance D of the plains rain gauge eastward from the coastline, in kilometers.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R ≥ 0.85 and at least 30 yr of paired rainfall data (the same as used for Fig. 4b). The pairs are partitioned by the distance of the plains rain gauge eastward from the coastline: (a) D < 10 km, (b) 10 ≤ D < 20 km, and (c) D ≥ 20 km. The Kendall two-tail nonparametric P value for (c) is 0.002.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R ≥ 0.85 and at least 30 yr of paired rainfall data (the same as used for Fig. 4b). The pairs are partitioned by the distance of the plains rain gauge eastward from the coastline: (a) D < 10 km, (b) 10 ≤ D < 20 km, and (c) D ≥ 20 km. The Kendall two-tail nonparametric P value for (c) is 0.002.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Judean and Samaria hills, based on the pairs of rain gauges that had correlation of R ≥ 0.85 and at least 30 yr of paired rainfall data (the same as used for Fig. 4b). The pairs are partitioned by the distance of the plains rain gauge eastward from the coastline: (a) D < 10 km, (b) 10 ≤ D < 20 km, and (c) D ≥ 20 km. The Kendall two-tail nonparametric P value for (c) is 0.002.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the probabilities of trends in Ro in northern Israel, based on the data from Table 4 for gauge pairs that had at least 30 yr of coincident rainfall data. (a) Pairs classified by their correlation coefficients R; (b) all pairs with R > 0.85 classified according to the distance of the hill gauges eastward from the coastline. (c) As in (b), but for R > 0.80.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the probabilities of trends in Ro in northern Israel, based on the data from Table 4 for gauge pairs that had at least 30 yr of coincident rainfall data. (a) Pairs classified by their correlation coefficients R; (b) all pairs with R > 0.85 classified according to the distance of the hill gauges eastward from the coastline. (c) As in (b), but for R > 0.80.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
As in Fig. 1, but for the probabilities of trends in Ro in northern Israel, based on the data from Table 4 for gauge pairs that had at least 30 yr of coincident rainfall data. (a) Pairs classified by their correlation coefficients R; (b) all pairs with R > 0.85 classified according to the distance of the hill gauges eastward from the coastline. (c) As in (b), but for R > 0.80.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Galilee hills, based on the pairs of rain gauges that had correlation of R ≥ 0.80 and at least 30 yr of paired rainfall data (the same as used for Fig. 6c). The pairs are partitioned by the distance of the hill rain gauge eastward from the coastline: (a) D ≤ 15 km, (b) 15 < D ≤ 30 km, and (c) D > 30 km. The Kendall two-tail nonparametric P value for (c) is 0.006.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Galilee hills, based on the pairs of rain gauges that had correlation of R ≥ 0.80 and at least 30 yr of paired rainfall data (the same as used for Fig. 6c). The pairs are partitioned by the distance of the hill rain gauge eastward from the coastline: (a) D ≤ 15 km, (b) 15 < D ≤ 30 km, and (c) D > 30 km. The Kendall two-tail nonparametric P value for (c) is 0.006.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
The trend of Ro for the Galilee hills, based on the pairs of rain gauges that had correlation of R ≥ 0.80 and at least 30 yr of paired rainfall data (the same as used for Fig. 6c). The pairs are partitioned by the distance of the hill rain gauge eastward from the coastline: (a) D ≤ 15 km, (b) 15 < D ≤ 30 km, and (c) D > 30 km. The Kendall two-tail nonparametric P value for (c) is 0.006.
Citation: Journal of Applied Meteorology and Climatology 48, 8; 10.1175/2009JAMC1902.1
Center Israel: station name, designation, distance from the sea, and station height. The stations that were used by AHL08 are marked as A. Those used by GR04 or GR05 are marked by G.
Northern Israel: station name, designation, distance from the sea and station height.
All possible pairs between the rain gauges used for calculating trends in orographic precipitation enhancement factor in central Israel by AHL08 flagged with A, by GR04 and GR05 only flagged with G, and by both flagged with AG. Only pairs that had at least 20 yr of paired rainfall data were retained. For each pair, the following are shown: name of the plains rain gauge, its AG flag, and its average rainfall (mm yr−1), name of the hill rain gauge, its AG flag, and its average rainfall (mm yr−1), the ratio Ro of the average hill/plains rainfall amounts, number of years with paired rain gauge data, correlation between the annual rainfalls, and slope of the least squares fit to the annual ratios of hill/plains rainfalls. The pairs are sorted by descending order of their correlation.
