Date of Award

January 2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Atmospheric Sciences

First Advisor

David Delene

Abstract

North Dakota farmers’ interest in using weather modification to increase precipitation and reduce hail damage resulted in a managed cost-sharing program, the North Dakota Cloud Modification Project (NDCMP), being started in 1976. The goal of this research is to determine the effectiveness of NDCMP at increasing precipitation by using the rain gauge observations of National Weather Service (NWS) Cooperative Observer Program (COOP) and North Dakota Atmospheric Resource Board Cooperative Observer Network (NDARBCON). The rain gauge analysis uses target and control regions. Target regions are selected from counties in District I and II that have participated in all 41 years of the NDCMP. Control regions are counties adjacent to the target counties. Precipitation is evaluated on a monthly and seasonal (June, July, and August) basis over the 41-year program.

Multiple analysis methods are used to examine the available rain gauge data. In one method, rain gauge data is analyzed by overlaying a circle with a radius of 40 km over each target and control region. Rain gauges are weighted to a central point within the circle. Monthly and seasonal rainfall totals are calculated within these circles. In another method, rain gauge data from the entire county is used to calculate county-wide average monthly and seasonal rainfall totals. In a third approach, all rain gauges from the target and control counties are combined to generate one large target and one large control dataset. Single and double ratios are calculated for each target and control region in the circle-based and county-based analyses. Bootstrapping is applied to the single and double ratios to determine the natural variation over the 41 years. An exploratory analysis in which single and multiple linear regression is created using 1950-1975 seasonal rainfall to predict what the rainfall in the target area would be without the seeding effect. These various methods provide an exploratory statistical analysis of the program, not a physical process evaluation.

The circular method shows that four out of the nine target/control double ratios have the target region receiving 2 to 8% more precipitation during the NDCMP years; however, only one out of the four increases are considered statistically significant based off the one tailed significance test. The 95% confidence intervals for these target/control pairs range from a small decrease to a small increase. The county-based method shows that six out of the nine double ratios have a range of 2 to 12% more precipitation in the target than the control region during the NDCMP years. Of those six, three cases are determined to be statistically significant by one tailed significance test. The single linear regression method shows an increase of 1 to 12% in target areas during NDCMP years in all but two of the target/control pairings when the standard error of the estimate is less than 1.50 (± 6%). Multiple linear regression shows an increase of 3 to 7% in target areas during NDCMP years when the standard error is less than 1.50 (± 6%) in 7 out of 12 analyzed cases.

Increases in rainfall due to cloud seeding in most instances are small and hard to detect statistically because it is difficult to predict how much rain would have fallen in the target areas during NDCMP years if seeding had not occurred. Uncertainties are caused by limited rain gauge data in the years before the NDCMP, from which natural target/control relationships can be determined. Results show that McKenzie county has an increase in rainfall when compared to control counties to its south and southwest, likely due to these cloud seeding efforts.

Download

Share

COinS