Introduction

Drought is a natural condition in many regions ofthe world, and should not necessarily be regarded as a problem but as a normal situation produced mainly by the absence or irregularity of rainfall and by excessive evaporation. Due to its geographical location, Mexico is very vulnerable to droughts, which have had to be endured from the earliest settlement of pre-Hispanic civilizations. A drought season started in Mexico during October 2010 and became more critical during the summer of 2011 due to summer rainfall anomalies. The state of Zacatecas, located in the semiarid region of Mexico where droughts are known to be frequent (^{Medina et al, 2006}), was declared inApril 2011 to be under severe drought conditions. These conditions continued through May and extended to cover 86% of the country. During the months of June, July, and August 2011 this condition of extreme drought persisted in Zacatecas (^{Giner et al, 2011}).

Weather-modification activities have frequently been undertaken during droughts, when there is a desperate need for water (^{Bruintjes, 1999}). Modern weather modification began with the discoveries of Schaefer in 1946 and Vonnegut in 1947, who found that supercooled liquid water clouds could be changed into ice crystals using silver iodide (Agí) or dry ice (Pendick, 2000). By the 1960's and 1970's, artificial stimulation ofrain had become widespread around the world, and Mexico was no exception (^{Rangno and Hobbs, 1995}). Cloud-seeding aims to enhance a natural process already existing in the atmosphere. Such seeding can increase rainfall from clouds about to precipitate or already precipitating. However, it can not be used to form new clouds nor to transform nonprecipitating clouds into rain clouds. No effect will be achieved by applying chemicals if the moisture conditions are inadequate, if no drops are present in the clouds, or if there are no ascending currents. Nevertheless, even slight increases in rainfall can be important in subtropical regions where rainfall is scarce (^{Rosengaus and Bruintjes, 2002}; ^{Levin, 2009}). Not all clouds are susceptible to seeding, being limited to supercooled continental clouds with temperatures of -10 to - 20 °C and with significant amounts ofsupercooled water (^{Cooper and Lawson, 1984}).

The concept behind artificially stimulating rain through cloud-seeding is to emplace an adequate proportion of condensation nuclei (hygroscopic seeding) or ice crystals (glaciogenic seeding) to render conversion mechanisms more efficient for transforming water vapor into liquid and solid phases, respectively. Possible increases in rainfall clearly depend on the natural existence ofthese nuclei in the atmosphere (^{Rosengaus and Bruintjes, 2002}). Traditional techniques of artificial stimulation with Agí or dry ice, developed in mid-latitude countries and for winter rainfall, consider as an ideal target those storms found mainly at atmospheric levels with temperatures above freezing, since these levels are found relatively close to the ground in such countries during such seasons. In tropical or even subtropical regions during the summer, most developing rain storms are located in areas below the freezing level, because this level is found several kilometers above the sea and therefore these clouds do not present significant ice-crystal fractions. The indiscriminate application ofglaciogenic techniques in these regions has not produced strong positive results (^{Rosengaus and Bruintjes, 2002}).

Glaciogenic seeding disseminates ice-producing materials into clouds containing water drops and with temperatures below 0 °C, stimulating rainfall (^{Levin, 2009}). Many other experiments releasing dry ice among clouds from an aircraft have been conducted with varied results since the first glaciogenic seeding experiments in 1946 (^{Edwards and Evans, 1968}; ^{Gagin and Neuman, 1981}; ^{Braham 1986}; ^{Gabriel and Rosenfeld, 1990}; ^{Laiguang and Yangang, 1995}). Several seeding agents have been used, but the most common have beenAgí and dry ice. Both ofthese chemicals increase the concentrations of ice crystals in clouds, either as new crystal nuclei or as frozen cloud droplets (^{Hydro Tasmania, 2008}).

Hygroscopic seeding disperses large hygroscopic particles (salt particles) into a cloud. The object of artificially introducing condensation nuclei is to condense water vapor in order to increase the size of water particles into drops (^{Murty 1989}; ^{Levin, 2009}). The disadvantage ofthis method is the large amounts of salt that are necessary in the process (^{Murty 1989}). The technique is widely used in Southeast Asia, although many experiments have also been conducted in other countries, yielding varied results (^{Braham 1986}; Murty 1989; ^{Rosengaus and Bruintjes, 2002}; ^{Rosengaus and Calderon, 2004}). Although some of the experiments have suggested positive effects, the results have generally been inconclusive. Hygroscopic cloud-seeding studies have also been implemented in South Africa (1991-1996), Thailand (1995-1998), and Mexico (1996-1998), and even though their initial findings were promising, more recent experiments have not supported them. Nevertheless, results from these new studies suggest that rainfall can be increased on some occasions (^{Levin, 2009}).

In Mexico, cloud-seeding programs have made use of glaciogenic seeding techniques, even though they are not particularly favored by summer storm characteristics (^{Rosengaus and Calderon, 2004}). The Mexican Light and Power Company (Compañia Mexicana de Luz y Fuerza) conducted cloud-seeding experiments with Agí from 1949 to 1962 over the Necaxa watershed and the upper Lerma River valley. Lower-precipitation categories registered rainfall increases, but higher-precipitation categories recorded decreases (^{Pérez and Ahumada, 1963}).

Between 1978 and 1982, a cloud-seeding program was conducted in the area of the Nazas River basin in the La Laguna region in northeastern Mexico, aimed at increasing the contribution of rainfall to the Francisco Zarco Dam reservoir (^{El Siglo de Torreón, 1978}). After several years, the program ended, having led to a 15-20% volume increase in the reservoirs of the El Palmito and Francisco Zarco dams. Similar seeding programs were also conducted in the states of Durango (1999) and Zacatecas (2001), but their findings have not been published.At the beginning of 1990, in response to severe drought conditions in the Monclova (Coahuila) region in northern Mexico, a scientific program was launched to assess the feasibility of increasing rainfall through hygroscopic cloud-seeding techniques. The seeding program was implemented from 1996 to 1998, and it was found that precipitation amounts were higher for seeded rain storms ( ^{Brant and Bruintjes, 1999}; ^{Rosengaus and Calderón, 2004}).

Although several international studies have reported positive results on enhancing rainfall by cloud-seeding, no scientific consensus has been reached with regard to the effectiveness of cloud-seeding in modifying weather (U.S. Weather Modification Research and Operations, 2003; ^{Orville et al., 2004}). An overall view of existing research programs indicates that even though cloud-seeding influences the microphysical processes occurring in clouds, there is very limited evidence that it affects the amount of rain falling on the ground (^{Levin, 2009}). ín this context, the objective ofthe present study was to assess the efficacy of Agí applications in the Mexican states of Zacatecas and Aguascalientes during the 2012 wet season.

Materials and methods

This work was developed during the 2012 rainy season in the major agricultural areas ofthe Mexican states of Zacatecas andAguascalientes. Forty-six seeding flights withAgí were made, totaling 90 hours. Data registered for each event were the flight date, flight start and end times, seeding start and end times, and GPS-recorded flight routes. The study period of interest was 21 June to 10 September 2012.

The website http://weather.rap.ucar.edu/satellite/ was consulted daily to monitor how clouds formed and traveled as shown by images reported by the satellite service every 15 min. Concurrently with satellite-image observations, visual inspections were made at a local level to identify cloud types, and cumulus clouds in particular as they are favorable for seeding. No flight was started until sufficient cumulus-type cloud formations were observed in the target area both in satellite images and through on-site inspections. Also, the probability of precipitation occurrence at the target site was determined; if the probability lay within the range of 30-70%, then a cloud-seeding flight was conducted.

Seeding was performed using a twin-engined airplane equipped with a generator mounted under each wing. A mixture of Agí and acetone was carried inside a 26-liter tank in each generator. When the plane was positioned directly under the clouds, the generators were started so that Agí could be released through combustion and thereby forced to travel upward into the clouds. According to the company performing the applications, depending on weather and cloud conditions, precipitation occurred 15 to 30 minutes after seeding had begun.

As a means for verifying rainfall, data were obtained from the agroclimatic monitoring networks of the Mexican National ínstitute for Agricultural, Livestock, and Forestry Research (íNíFAP, Instituto Nacional de investigaciones Forestales, Agricolas y Pecuarias) and the Zacatecas and Aguascalientes Produce Foundations (FP, Fundacion Produce). Data were also obtained from the Zacatecas and Aguascalientes National Water Commission (CONAGUA, Comisión Nacional del Agua) weather station network in Zacatecas and Aguascalientes.

Figure 1A presents all 46 flight routes as well as the locations of the CONAGUA and INIFAP-FP weather stations located in Zacatecas and Aguascalientes. Figure 1B shows only the automatic stations that were eventually selected within the area of influence of each route.

Figure 1 (A) The 46 flight routes of the cloud-seeding program and locations of weather stations in the states of Zacatecas and Aguascalientes. (B) Weather stations selected within the areas of influence of flight routes. 

Definition of measurement criteria

The INIFAP-FP automatic station networks record data every 15 min and, in addition to rainfall amounts, they also provide the hour in which precipitation occurs. Therefore, we established a first criterion that rainfall would be considered as having been due to seeding when it took place between the time the flight started and one hour after cloud-seeding ended, representing about a three-hour period (Figure 2). A second criterion was to consider whether or not rain had fallen over both the Agí seeding route and other areas simultaneously. ín other words, if the precipitation was widespread, then it was not regarded as having been caused by cloud-seeding.

Figure 2 Precipitation registered by the Rancho Grande Fresnillo weather station on 24 June 2012. The interval over which rainfall was considered to have been caused by seeding is shown in red. 

As a third criterion, after locating the GPS-registered flight route, a 30- km AgI buffer or influence zone was established 15 km to each side of the recorded route. This allowed only those weather stations located within such an area to be selected for data, having taken into account the possible effect of the wind on the dispersion of the seeding chemical. According to the average wind velocity in the zones, it was considered that a buffer area established 15 km on either side of a recorded flight route would cover most eventualities. A fourth criterion considered the rainfall effective for agricultural purposes to be only those precipitation events over 5 mm, because lower amounts of rainfall provide negligible amounts of moisture for crop roots due to the evaporation that occurs when rainwater reaches the ground (^{Serna et al., 2011}).

Statistical analyses (chi-square tests and t-tests) were performed on the data to determine whether results were significantly different from normal precipitation. Average historical precipitation data were obtained (^{Medina et al, 2004}) from 21 CONAGUA stations for the wet seasons (21 June to 10 September) for years 1996-2009, as well as the average precipitation data for the 2012 season. Maps were made for the 2012 wet season, including of percentages computed with respect to historical averages.

Results and discussion

In most cases, depending on environmental and cloud conditions, some clouds started to produce rain 15 to 30 minutes after seeding had begun in the study area. Increases greater than radar precipitation estimates occur 30 to 60 min after seeding (^{WMO, 1999}). However, an inspection reveals that these events in the present study were light, highly localized rainfall and, in many cases, were so light that they were not registered by weather stations. This agrees with the observation that only a small part of available moisture in clouds transforms into precipitation that reaches the surface (^{National Academy of Sciences, 1973}).

One of the criteria established for the present study was that no precipitation would be considered as having been caused by seeding if it occurred during a period of widespread precipitation. Four out of the 46 cloud-seeding dates experienced widespread precipitation, and consequently those data were not included in the analysis.

Due to the fact that CONAGUA weather stations record 24-hour rainfall amounts, compared with the 15-minute reports of the automatic INIFAP-FP stations, it was not possible to distinguish the amount of rainfall occurring during a seeding period. Rainfall percentage during seeding periods with respect to the 24- h precipitation was 29.2% as recorded by INIFAP-FP stations, whereas CONAGUA stations registered 100%. Therefore, the latter data were not included in the analysis, because there was no certainty that precipitation had actually occurred during the rainfall stimulation period and it might have happened at any moment due solely to natural causes.

Table 1 presents statistics for the INIFAP-FP weather station network. The total number ofstations within the flight routes was 275 considering all dates, whereas the number ofstations for which rainfall was recorded was 45, meaning that only 16.4% ofall stations recorded precipitation. This difference between the total number of stations (275) and the number of rainfall-recording stations (45) was statistically significant (Chi-square, p< 0.05), indicating that the number ofstations recording rainfall was indeed smaller than the total.

Table 1 Statistics for Zacatecas and Aguascalientes INIFAP-FP weather stations located within flight routes on all seeding dates (40 stations). 

Data with precipitation values above 5 mm were analyzed given that rainfall amounts in this range have agricultural significance. During the study period, such rainfall occurred at only 9 of 275 stations (3.3%), which drastically reduces the percentage of stations with rainfall and indicates that in many cases, precipitation occurring due to seeding was light. This difference was confirmed to be statistically significant (Chi-square, p< 0.05), indicating that very few stations recorded precipitation deemed effective for agricultural purposes.

The average 24-h rainfall on all seeding dates was 11.9 mm, and the average precipitation average during seeding-time was 3.5 mm, representing 29.4% of the precipitation recorded in 24 h (t-test, p< 0.05). This difference indicates that rainfall due to seeding was significantly lower than the total amount of precipitation recorded for the corresponding day. The average 24- h rainfall during the seeding periods, computed using only those days for which rainfall exceeded 5 mm (i.e., considering the dates with rain < 5 mm as zero), was 2.6 mm, and consequently the percentage with respect to 24- h precipitation decreased to 21.8%, similar to seeding results obtained in Tasmania (^{Morrison et al, 2009}).

INIFAP-FP network stations within the target area in the states of Aguascalientes and Zacatecas (Figure 3b) supplied the average rainfall data registered by all stations located under the seeding routes on each of the 42 dates for rainfall occurring during the 24- h periods as well as during the seeding periods. The total 24- h precipitation on all dates was 81 mm, where as the total precipitation during the seeding period was 26 mm, or 32.1% ofthe total.

Figure 3 (A) Precipitation during the 2012 rainy season (21 June to 10 September). (B) Percentages for the 2012 rainy season with respect to rainy season historical averages (1961-2009). 

These data agree with those reported by ^{Sharon et al. (2008)} , who concluded that seeding in Israel enhanced the amount of precipitation by 30% in storms producing less than 5 mm per day. However, other authors have obtained different re sults. For example, ^{Rosengau and Calderón (2004)} reported a 66.8% increase in total water volume precipitated by isolated summer convective storms in Monclova, Coahuila, in semiarid northeastern Mexico. Several cloud-seeding studies in Tasmania found that seeding was effective, withprecipitation being increased by as much as 8% per seeded month (^{Hydro Tasmania, 2008}). A scientific experiment in South Africa reported that cloud-seeding produced an average 20% increase in the amount of rainfall (^{Rosengaus and Bruintjes, 2002}). Results of a 1960-2005 seeding of a hydroelectric basin located in central Tasmania, Australia, indicated a monthly precipitation increase of between 5% and 14% (^{Morrison et al, 2009}).

In the study region, increases in precipitation should be expected to occur during cloud-seeding programs. Rainfall data were compared to verify if such an increase had taken place during the study period of 21 June to 10 September 2012, corresponding to the rainy season of the study area. Average rainfall data were obtained for the same seasonal period for the years 1961 to 2009 from 21 CONAGUA weather stations (Table 2). From these data, the mean precipitation during the wet season is 249.7 mm and it generally fluctuates between 155 and 345 mm (s= 95 mm, 38% of the mean), with the rainfall lying within this range for 38 out of the 49 rainy seasons.

Table 2 Basic total precipitation statistics for the rainy season (21 June to 10 September) during the years 1961-2009 and during 2012 (the year of the cloud-seeding study). 

The total precipitation during the 2012 rainy season, averaged over the seeded area, was 213.1 mm, an amount that lies within the normal precipitation range (Table 2) but which happens to be below average. This indicates that in spite of seeding, precipitation did not improve sufficiently to exceed the average. The high natural variability in rainfall renders it extremely difficult to discern the effects of cloud-seeding. Even the best experiments have needed more than 100 seeding days to detect with any confidence a 10% increase due to seeding (^{Bigg and Turton, 1988}).

The total amount of rainfall recorded, including both natural rainfall and that due to seeding effects, was between 87 and 360 mm for the 2012 seeded area, depending on location within the seeded area (Figure 3A). Comparing this precipitation to the historical average, it is noted that rainfall was 10-50% below normal in most of the seeded area (Figure 3B).

The historical average precipitation in the seeded area during the rainy season (21 June to 10 September) is 249.7 mm. The precipitation that accumulated during the study period due to seeding effects and in events of > 5 mm was 20 mm, which represents 8% of the historical average.

Overall, it may be stated that in spite ofthe seeding program, precipitation recorded by the INIFAP-FPZ weather stations during the June to September 2012 wet season was not widespread and it turned out to be less than normal. We conclude that the cloud-seeding had no significant impact on increasing the amount of precipitation, at least not for agricultural purposes.

Conclusions

A cloud-seeding program aimed at enhancing precipitation during drought was implemented during the 2012 wet season in the Mexican states of Zacatecas and Aguascalientes. For most of the 46 seeding flights, it was found that some clouds began to precipitate 15 to 30 min after seeding started, although these events constituted light and highly localized rainfall. Rainfall recorded by all stations during 24 h and accumulated for all seeding dates was 81 mm, but only 26 mm of this total corresponded to seeding periods, representing 32.1%.

For all weather stations on all flight routes, the percentage of stations with rainfall recorded during the seeding period was 16.4%; however, this percentage decreased to 3.3% when only those rainfall events with amounts over 5 mm were considered. The historical average in the seeded area during the wet season (June to September) is 249.7 mm. The accumulated precipitation during the study period within the seeded area due to seeding effects and in events over 5 mm was 20 mm, which represents 8% ofthe historical average.

Overall, in spite ofthe cloud-seeding program, precipitation recorded during the June-September season by the INIFAP-FPZ network stations was not widespread and in fact was less than normal, indicating that the cloud-seeding did not have a significant impact on increasing the amount ofprecipitation, at least not for agricultural purposes.