Michael Douglas*

National Severe Storms Laboratory

Norman, Oklahoma, USA


Walter Fernández

University of Costa Rica

San Jose, Costa Rica


Malaquías Peña

CIMMS/Univ. of Oklahoma

Norman, Oklahoma, USA





The Pan American Climate Study Program - Sounding Network (hereafter called PACS-SONET) is one of a number of research projects funded by the National Oceanic and Oceanic Administration (NOAA) Office of Global Programs. This research activity (Co-Principal Investigators M. Douglas and W. Fernandez) was approved in December 1996, with a primary objective of determining the atmospheric circulations associated with wet and dry periods in Central America during the rainy season. Another important objective was to help determine whether the current National Center for Environmental Prediction (NCEP) operational analyses over Central America and the eastern tropical Pacific are accurate. To help answer these questions required establishing a special atmospheric sounding network from southern Mexico to northern Peru, to describe the windfield with higher spatial density than is possible from the routine radiosonde network in the region. This paper discusses the logistics of this project, summarizes some of the lessons learned, and discusses requirements for maintaining such an observing network indefinitely.

Because many wind sounding stations were desired, inexpensive pilot balloon stations, rather than radiosonde stations, were established. The suitability of this relatively low-technology approach for the region was based on previous field program experience in Mexico. Only on Cocos Island were special radiosonde observations planned. The following section addresses some of the concerns regarding this type of observation.


2.1 Pilot balloon observations

The use of pilot balloons for determining winds above the surface has been carried out for more than 100 years, and is still widespread in some regions today. In principle it is straightforward, though many details must be considered to make the highest quality observations.

The main advantage of pilot balloon observations is that they are inexpensive. Typically, a 30 gm pilot balloon launch, excluding labor costs, is only 5-10% the cost of a radiosonde launch. In addition, they require a relatively simple instrument (a meteorological theodolite), require no electric power, and the observer does not need to be computer-literate. The theodolites for PACS-SONET were obtained without charge from a US National Weather Service storage and reconditioning facility; this greatly reduced the cost of the PACS-SONET project, since new theodolites cost ~$8,000 each.

The main uncertainty in the winds calculated from pilot balloon observations does not result from errors in measuring the angles - such errors can usually be detected and corrected by data editing procedures. Rather it is a consequence of turbulence in the atmosphere which causes the balloon's ascent rate to deviate from the assumed value. On a hot, windy afternoon the vertical motions associated with turbulence one km above the surface can be 1-2m/s or even more, introducing departures of 25-50% from the assumed ascent rate. This source of error was recognized, and the PACS-SONET observations were scheduled for early morning or late afternoon, when boundary layer turbulence is generally weak. For monthly mean values, this source of error should be fairly small, since the turbulence fluctuations are random, and the average ascent rate should be close to the assumed rate for an average composed of approximately 30 observations.

2.2 The original network configuration

The original PACS-SONET configuration (Fig. 1), consisted of 12 pilot balloon sites. The stations were established during April and May of 1997 in Mexico, Nicaragua, Costa Rica, Panama, Colombia, Ecuador and Peru, together with one radiosonde station on Cocos Island in the eastern Pacific Ocean. The intended duration of the observations was 6 months, ideally extending through October, and close to the end of the rainy season in much of Central America. The pilot balloon soundings were made twice-daily at all sites. This permits an assessment of the uncertainties inherent in climatological upper-air wind analyses that are generally based upon only once-daily soundings.


Fig. 1 The initial PACS-SONET station distribution



Shortly after the first stations were established in April 1997, it became apparent that a significant warm event in the eastern and central Pacific was developing. By mid-summer, the extreme magnitude of the event had become apparent. This led to two conclusions: 1) the observations being made during the summer of 1997 would be very anomalous over Central America, thus putting into question the extrapolation of potential scientific findings to other years, and 2) there existed the possibility of obtaining an unprecedented data set to describe the evolution of the windfield associated with the developing El Niño. Therefore, the observational plan was modified to: 1) extend the observational period at many stations through the end of October 1998, 2) establish additional pilot balloon stations in Ecuador and Peru to better sample conditions in the region of anticipated heavy rainfall, and 3) include the installation of several hundred inexpensive raingauges in coastal Ecuador and Peru. The last activity was necessary to assure adequate measurement of daily rainfall for comparison with pilot balloon and satellite observations. The financial, logistical, and technical difficulties involved in establishing and expanding the network (ultimately to 17 pilot balloon sites) is described in this section.

3.1 Extension of the observing period

Because the warm season rainfall over Central America was observed to be well below normal, and it was apparent that El Niño was modifying the normal conditions, it was considered desirable to make observations during a more normal warm season. This would not only more closely satisfy the original objectives of the project, but also provide an excellent comparison with the 1997 El Niño summer. Thus, plans were made to extend the observation program until the end of October 1998, which is close to the end of the rainy season over Central America. Thus, two complete rainy seasons will have been observed by PACS-SONET. The principal difficulty in extending the observations for another year was related to the cost of the additional observations. To manage this, we had to reduce the observations from twice- to once-daily. However, even this created considerable budgetary difficulties, due to the short notice that was given to the funding agency.

After about 45 mostly successful launches, we experienced difficulties with the operation of the radiosonde system on Cocos Island. Based on these difficulties, it was decided to eliminate the radiosonde observations there; a step that saved some $20-25,000. The elimination was deemed justifiable, since the loss of winds from the majority of the radiosonde sites in the Caribbean, Central America and northern South America (due to the shutdown of the Omega system on Sept 30, 1997) had already left huge gaps in the radiosonde network. Thus, the Cocos observations would not have been especially useful to estimate typical NCEP analysis uncertainty over this data void, since any difference between the analyses and the Cocos observations could be ascribed to the lack of neighboring stations to help initialize the analyses. Also, given the anomalous year, the Cocos soundings would not describe a "normal" boundary layer downwind of the zone of cold ocean temperatures along the Equator (the oceanic "cold tongue"), which was an additional PACS-SONET objective.

3.2 Increasing the number of pilot balloon stations

By October 1997 it was apparent that the El Niño event was comparable to one during 1982-3, and that heavy rains might again affect the northern region of Peru. During the 1982-3 event there were no regular radiosonde or other wind sounding systems anywhere in the coastal region of Peru (north of Lima), or in any part of Ecuador. This lack of atmospheric soundings has limited the meteorological interpretation and explanation of the distribution of heavy rainfall associated with the 1982-3 event. Since such strong El Niño events are quite rare, we deemed it wise not to pass up this opportunity for sampling a rare phenomenon, especially since a basic sounding network was already in place in Ecuador and Peru.

During the 1982-3 El Niño there was a large north-south gradient in the rainfall along the Peruvian coast, with Piura, Peru receiving more than 2000 mm over the 6 month period of rains, while Trujillo, only 300 km to the south, received no rainfall. To measure the north-south variations in the windfield we decided to establish additional pilot balloon stations at Tumbes, Trujillo, Chiclayo and Ancon2. (Ancon2 was to partially replace the lost winds from the Lima Omega-based radiosonde station.) A site was also established at Iquitos, in the Amazon Basin, which when coupled with the wind observations along the coast, and at San Cristobal in the Galapagos Islands, would better describe east-west windfield variations. In Ecuador, additional sites were established on the coast at Ancon (near Salinas), and at Portoviejo.


Fig. 2 The pilot balloon station distribution during Jan-June 1998 (January-March only for Santa Cruz).

Although the region of Ecuador and northern Peru might have appeared to have been over-sampled compared with the rest of the PACS-SONET domain, there is considerable cloudiness in the former region during major El Niño events , and most balloons cannot be tracked to as high a level as in other localities. To obtain a reasonable number of wind observations at higher levels, more stations were necessary.

For financial reasons, most of the special El Niño sites stopped operation at the end of June 1998, following the end of heavy rains in Ecuador and Peru.

3.3 Establishment of a special raingauge network in Ecuador and Peru

Once the emphasis on explaining the variations of El Niño rainfall along the coast of Ecuador and Peru was accepted as an additional scientific objective of PACS-SONET, it was necessary to consider how the rainfall would be measured in this region. Although a basic climatological rainfall network exists in the region, some questions, such as the detailed mechanisms responsible for organizing the rainfall, require a knowledge of mesoscale rainfall variations. It was considered impractical and excessively expensive to establish automated raingauge networks in the region, or to bring a meteorological radar to the region. In any case, such activities could not be carried out on short notice. Based on experience with simple raingauges during experiments in Mexico in 1993, 200 plastic wedge-shaped raingauges (@$6 each) were purchased, and 100 were shipped to each country. Despite customs delays in both countries, the majority of the raingauges were placed in the field by Jan 1, when rain started to fall in earnest in northern Peru. Because of widespread interest in El Niño rainfall, all of the rainfall observations (recorded daily) were made by volunteers, which has minimized the cost of the activity.



The PACS-SONET has shown that a large network of inexpensive sounding stations can be established and operated for reasonably low costs (about $100,000/year for a network of 10 stations). Some of the positive aspects of the network operation and observations include:

• stations can be established easily, and training takes little time.

• quality and quantity of observations is acceptable for producing reliable monthly mean wind statistics for lower troposphere at most sites.

• Internet-based distribution of data is feasible for research activities and to monitor many aspects of the network.

Some lessons learned include:

• Although observers can be trained quickly, quality observations require practice and quality control at the local level.

• Price and availability of inflation gas varies widely throughout the region complicating precise budgeting.

• Cloudiness at some coastal sites is very high, resulting in few high soundings.

• Internet data distribution is not fast enough for real-time weather forecasting use of data in most countries.

• Near real-time communications via fax or phone to observing sites and data processing centers are needed as a backup to Internet access.

• Local financial resources are not sufficient to operate most of the network stations without outside support.

• Use of the pilot balloon data is very limited within each country, mostly due to lack of familiarity with how to interpret the observations.

• Operating the network requires nearly a full-time manager; this was not anticipated, nor budgeted for in the original project. One or more regional managers (located in one of the participant countries) would be highly desirable for following logistical details at each of the sounding stations.


Although the PACS-SONET was established to satisfy basic research objectives, it was recognized from the outset that the network could also serve as a model for a climate monitoring network in the region. As such, it would have to satisfy several requirements, including being inexpensive to operate, simple to maintain, and being capable of providing information needed for research studies and monitoring of short-term climate variability. However, it was also recognized that it was not realistic to expect a research program such as PACS to provide all of the operating expenses for such a long-term monitoring activity. How then could such a network be maintained indefinitely? The answer probably lies in demonstrating that the network observations will be useful for daily weather forecasting activities in the region. If this value can be demonstrated to each of the participating countries, then such a network might be supported jointly, with contributions from both the meteorological services and scientific institutions of the region, and from NOAA.

In order to establish a permanent network like PACS-SONET to serve the needs of both operational weather forecasting, climate monitoring, and research studies, a number of developments must be undertaken in the host countries and in the meteorological community both in the US and Latin America. Some of these conditions are stated below:

• the value of the PACS-SONET observations for daily weather forecasting must be demonstrated to the meteorological services in each of the participant countries. This has not been successfully done to date.

• the observations must be transmitted in real-time, so that they are available at every weather forecast center in time for local forecasting activities.

• the network must be financially or logistically supported by each participating country or institution. It cannot be financially sustained strictly as a research project.

• there must be individual researchers throughout the region who are willing to work with the observations to improve the understanding of local and regional meteorology and climate, and to develop better forecasting techniques.

• Closer communication between all weather services in the region on a daily basis would certainly improve analysis and interpretation of observations. This might reduce duplicity of activities, something that can be ill-afforded by the meteorological services in the region.

A smoothly functioning atmospheric sounding network in Latin America that satisfies the needs of local, regional and global weather forecasting will require considerable attention to the weather forecasting needs in each country. Without input from the public and private sector of each country, it is not possible to design an observing system that can satisfy most forecasting needs effectively and economically. Yet although national input is critical to the design of a national network, a sounding network must be regional in perspective, and the network must be designed with input from the entire region, as weather recognizes no national boundaries.



The PACS-SONET activities were funded through the support of the NOAA PACS program. In addition, for their support in obtaining funds on short notice for the special El Niño observations, personnel from the PACS program (Andrea Ray, Mike Patterson) and numerous others from the NOAA Office of Global Programs are sincerely thanked. The support of many members of NSSL was essential for managing the numerous financial transactions, shipping activities, and paperwork required to complete the project. Rosario Douglas arranged countless details associated with the project both in the field and throughout the project. The support of almost innumerable observers, administrators and other meteorological service personnel in the various countries of the SONET domain is sincerely appreciated.

A more thorough list of acknowledgments (as well as other information and all of the soundings made during SONET) can be found on PACS-SONET Homepage at:



* Corresponding author address: Michael W. Douglas, NSSL,

1313 Halley Circle, Norman, OK 73069,