Assessing the Effects of Conservation Practices and Fertilizer Application Methods on Nitrogen and Phosphorus Losses from Farm Fields -- A Meta Analysis

摘要

Nitrogen and phosphorus runoff from agricultural lands and the subsequent impact on water quality has been of great concern in the United States, due to harmful algal blooms and anoxic zones in areas such as Lake Erie and the Gulf of Mexico. Conservation practices have been widely used to reduce the quantity of nutrients leaving a field, but there is a lack of research on the effectiveness of these practices using field scale data. The objective of this thesis is to quantify the effect of conservation practices on nitrogen and phosphorus runoff in farmlands. A meta-analysis was conducted using the Measured Annual Nutrient loads from AGricultural Environments (MANAGE) database created by the USDA-ARS. MANAGE is a compilation of 65 publications including data on nitrogen and phosphorus loads, runoff, land use, fertilizer application, and other field characteristics. The observational nature of the dataset makes direct comparisons from field to field impossible because of large variations in field characteristics. Thus, additional steps must be taken to estimate the effect of conservation practices on nutrient loss. To quantify this effect, I used propensity score matching and multilevel modeling, two statistical methods common for observational data. Propensity score matching shows that conservation practices have a significant reduction of 67.5% in total phosphorus, 83% in particulate phosphorus, and 67.3% in particulate nitrogen. Multilevel modeling results -- calculated using two different computational methods -- support these findings by showing a significant reduction of 57.7% in total phosphorus, 76.2% and 82.1% in particulate phosphorus (via the two methods), and 63.7% in particulate nitrogen. When examining different land uses and fertilizer application methods, the multilevel modeling showed that conservation practices had the most impact on row crops (e.g. corn and soybeans) and on farms fertilized via the injected or surfaced applied method. The results from this work represent the average effect of conservation practices on a national scale. At a regional scale, the effects of conservation practices may vary because of regional differences in agricultural practices and climate. To assist future research at regional and local scales, this thesis provides a Bayesian modeling framework for future quantification of these effects.