Publication Date

8-2020

Date of Final Oral Examination (Defense)

4-29-2020

Type of Culminating Activity

Thesis

Degree Title

Master of Science in Hydrologic Sciences

Department

Geosciences

Supervisory Committee Chair

Alejandro Flores, Ph.D.

Supervisory Committee Member

David Wilkins, Ph.D.

Supervisory Committee Member

Nancy Glenn, Ph.D.

Abstract

Climate change poses serious threats to global agriculture, however some localities and crops may benefit from increasing temperatures. Grape production in southern Idaho may be a beneficial example as vineyard acreage has increased over 300% since the designation of the Snake River American Viticultural Area (SRVAVA) in 2007. We perform a statistical characterization of agroclimate within the SRVAVA that centers around four primary objectives: utilization of a novel, 30-year high resolution climate dataset to provide insight and agrometrics unavailable at coarser resolutions, climatic implications of the unique topography within the SRVAVA, identification of statistical trends, and correlation of SRVAVA climate to large-scale climate indicators such as the El Nino Southern Oscillation (ENSO). In Chapter 3 we build on the identified correlations to large scale climate and utilize a long short-term memory (LSTM) model in conjunction with empirical mode decomposition (EMD) to create a novel, data driven method to forecast regional temperature trends with lead times up to one year. Favorable results for local viticulture include an increase in growing degree days and season length, as well as reduced frequency of freezing events. Possible disadvantages include increased risk to shoulder season freezing events with warmer winters, increased magnitude of strong freezing events, mid-season heat stress, and higher susceptibility to powdery mildew outbreaks. Additionally, with strong correlations identified with large-scale climate indicators, we find EMD an effective method to increase modeling power by using multiple frequencies of the signals as input into a LSTM machine learning algorithm that can accurately predict temperature trends up to one year in advance. This climatic characterization and modeling framework could potentially inform many agricultural management decisions such as cultivar choice, vineyard site selection, fungicide spray timing, irrigation strategy, and canopy management.

DOI

10.18122/td/1703/boisestate

Included in

Hydrology Commons

Share

COinS