Evaluation of seeding technologies

   Methods of evaluation

The evidence that is required to establish that a cloud seeding methodology is “scientifically proven” can be divided into two aspects, namely statistical and physical evidence. Statistical evidence is usually obtained by an experiment based on a seeding conceptual model that is conducted and evaluated in accordance with its original design using accepted statistical principles and procedures, and results in the rejection of the null hypothesis at an appropriate level of statistical significance and power of detection. The statistical evaluation enables the detection, in an as unbiased manner as possible, of a change (seeding signal) in a response variable, as specified by the seeding conceptual model, which is usually relatively small compared to its natural variability.
Physical evidence constitutes the measurement of key links in the chain of events associated with the seeding conceptual model that establishes the physical plausibility that the positive effects of seeding suggested by the results of a statistical experiment could have been caused by seeding intervention. The physical evidence enables the establishment of a cause-and-effect relationship between the seeding intervention and the changes in the response variables  as documented in the statistical evaluation. This is usually accomplished by means of case studies of the behavior of seeded and unseeded clouds that are conducted on a sample of clouds involved in the statistical experiment or separate from it, and/or as an integral part of the statistical experiment through the identification of a series of response variables associated with the seeding conceptual model. Such variables must be capable of being measured to the degree necessary to discern the anticipated changes due to the seeding intervention.

So the goal of a randomized experiment is to objetively evaluate the seeding hypotesis, which is designed to determine the effects of seeding. This section includes PRELIMINARY results from campaings in Mexico 1997-1998.

The hypotesis tested for each time series response variable is that the value of the variable is larger for seeded storms than for
non-seeded storms during the time period between 10 and 60 minutes after decision time. For variables that are total quantities the value of the response variable is hypothesized to be larger overall for seeded storms than for non-seeded storms.
Time series variables include precipitation flux, total storm mass, storm mass above 6 km, storm area and a measure of the location of maximum rada reflectivity. The total response variables include total precipitation mass, the area-time integral and the storm duration.

The time series response variables will be evaluated using the 0.25 th, 0.5th, and 0.75th quantiles of the distributions of the values of the variables for seeded and unseeded cases.

The preliminary results presented here provide an indication of a positice effect of seeding, in terms of precipitation flux. These results are very similar to those found un the South African experiment. This fact is encouraging, especially because the timing as well as the magnitude of the seeding effects corresponds well to the South African results.
 

   Re-randomization test indicate that most of the observed differences are not statistically significant at the 95% level. Thus, a small posibility exist that the apparent seeding effects may be the result of chance. Data from further field seasons are necessary to extend these results and establish statistical significance.