High-resolution spatial modeling of household lawn fertilizer behavior: Implications for targeting nitrogen-reduction...

Residential lawns cover a substantial area within urbanized regions and household fertilizer applications represent a major nonpoint source of nitrogen pollution. This paper develops a novel methodology for an integrated model that produces parcel-level, large-scale predictions of residential lawn fertilizer use and corresponding nitrogen applications. An illustrative application of this methodology is provided for Connecticut coastal counties within the Long Island Sound watershed to allow the comparison of targeting strategies for nitrogen fertilizer-reduction programs. A two-part hurdle regression model of household lawn fertilizer behavior is estimated using household survey data. Hurdle model estimation results are then used to predict lawn-fertilizer behavior for all single-family households in the region (N = 431,349) based on observable property characteristics and household demographics. Predictions integrate model results with high-resolution land cover data to identify lawn footprints and a simulation procedure to incorporate demographic information for households that were not in the survey sample, thereby producing parcel-level estimates of nitrogen applied to lawns. The effectiveness of nitrogen-reduction scenarios is compared for the optimal, simple rule-based, and municipal-level targeting strategies. Large variations in fertilizer and nitrogen applications exist over household types and municipalities. The majority of predicted nitrogen application is due to a small subpopulation of households, primarily parcels with newer, larger homes on large exurban lots. Results demonstrate that targeting strategies play a key role in the potential effectiveness of nitrogen-reduction programs. Findings such as these illustrate the insights for policy guidance that can be provided through integrated approaches that predict household-level fertilizer use across the landscape.