Evaluation of the Effectiveness of Cloud Seeding in Texas from 2002 through 2006

A method for the objective evaluation of short-term, non-randomized, operational, convective cloud-seeding projects on a floating-target area basis has been developed and tested in the context of the operational cloud seeding projects of Texas. The computer-based method makes use of NEXRAD, 15-min, mosaic, radar data to define fields of circular (25-km radius) floating-target analysis units with lifetimes from first echo to the disappearance of all echoes, and then superimposes the track and seeding actions of the project seeder aircraft onto the unit fields to define seeded (S) and non-seeded (NS) analysis units. Objective criteria are used to identify “control” (C) matches for each of the seed units from the archive of NS units. To minimize potential contamination by seeding no matching is allowed for any control unit if its perimeter came within 25 km of the perimeter of a seed unit during its lifetime.

The methodology was used to evaluate seeding effects in the Texas seeding projects existing in the period April through September in 2002 through 2006. Objective unit matches were selected from within and outside each operational target within 12, 6, and 3h of the time on a given day that seeding of a particular unit took place. These were done to determine whether selection biases and the diurnal convective cycle confounded the results. Matches were also drawn from within and outside each target using the entire archive of days on which seeding was done. Although the results of all analyses are subjected to statistical testing, the resulting P-values are used solely to determine the relative strength of the various findings. In the absence of treatment randomization P-values cannot be used as proof of seeding efficacy.

The results are presented for all seasons (2002-2006) combined instead of for individual seasons because of the limited sample sizes for some projects. The evidence for seeding-induced rain increases is strongest for the Panhandle (PH), SOAR (SR), CRMWD (CR), West Texas (WT), and South Texas (ST) projects. The Southwest (SW) and High Plains (HP) projects are also quite positive, although their P-value support is a little weaker than the first five projects. Typically, the sizes of the effects range between 20 and 25%, although the apparent effects for the CR and HP projects are greater. The volumetric rain increment per seeded unit ranges between 1,600 and 2,400 acre-feet. Again, the increments for the CR and HP projects are larger. The results for the Abilene (AB) and Pecos (PC) projects appear negative although they are based on a small sample and have very weak P-value support.

A major component of the analysis involved an examination of the project seeding effects as a function of the age of the unit when it was first seeded. In doing the analysis it was required that the prospective control match be of the same age at the time in its history when it was matched with the seed unit. A young unit was defined as one that had existed no longer than 60 minutes at its time of first seeding while a middle-age unit had an age of 75 to 120 minutes at its first seeding. An old unit was one that had existed on radar 135 minutes or longer. In all ten projects the percentage seeding effect was greatest for the young seeded units and smallest, even negative for the AB project, for the old seeded units. This consistency suggests that seeding was operative even in the two weak projects (AB and PC) even though the overall seeding effects appeared negative with weak P-value support in both projects. The percentage seeding effects in young units exceeded 100% in 6 of the 10 projects studied even exceeding 200% in 2 (HP and SW) of these 6 projects.

The rain increments (mean S - mean C in acre feet) for the units as a function of age are also of intense interest. As one would expect, the volumetric rain increments are greatest for the young seeded units, reaching a huge 20,000 acre feet per young unit for the old HP program. In 4 of the other 9 projects the seeding increment exceeded 8,000 acre-feet. Even when seeding older units, however, there was still a positive payoff except for the AB and PC programs that had small negative increments. The obvious lesson here is that a project is better off seeding convective units early in their lifetimes.

The unit rainfall results were partitioned further by unit rain-volume rate (RVR) at the time of initial seeding (RVR0) into three RVR0 categories (light; 0 to 284 acre-feet/h, medium; 284 to 810 acre-feet/h and heavy; > 810 acre-feet/h at the time of first seeding (RVR0). This partitioning variable is related to unit age because a young unit will usually have correspondingly light RVR0 values whereas an old unit is more likely to have a heavy unit RVR0 at the time of seeding. In most cases the percentage seeding effect is greatest in units having light precipitation at the time of initial seeding, somewhat less for units with medium RVR0 values and least for heavy RVR0 values. The obvious exceptions are the AB and PC projects that show little to negative effects of seeding.

Although the results of these and other analyses to be described in the presentation make a strong case for enhanced rainfall by most of the Texas operational seeding programs, such programs must not be viewed as substitutes for randomized seeding efforts that are conducted in conjunction with realistic cloud modeling that are followed by replication, preferably by independent groups for maximum credibility.