Agro-Ecological Class Stability Decreases in Response to Climate Change Projections for the Pacific Northwest, USA

Harsimran Kaur,Richard A Rupp,John T Abatzoglou,John P Reganold,David R Huggins,Claudio O Stöckle

doi:10.3389/fevo.2017.00074

Abstract

Climate change will impact bioclimatic drivers that regulate the geospatial distribution of dryland agro-ecological classes (AECs). Characterizing the geospatial relationship between present AECs and their bioclimatic controls will provide insights into potential future shifts in AECs as climate changes. The major objectives of this study are to quantify empirical relationships between bioclimatic variables and the current geospatial distribution of six dryland AECs of the inland Pacific Northwest of the United States; and apply bioclimatic projections from downscaled climate models to assess geospatial shifts of AECs under current production practices. Two Random Forest variable selection algorithms, VarSelRF and Boruta, were used to identify relevant bioclimatic variables. Three bioclimatic variables were identified by VarSelRF as useful for predictive Random Forest modeling of six AECs: (1) Holdridge evapotranspiration index; (2) spring precipitation (March, April and May); and (3) precipitation of the warmest four-month season (June, July, August and September). Super-imposing future climate scenarios onto current agricultural production systems resulted in significant geospatial shifts in AECs. The Random Forest model projected a 58% and 63% increase in area under dynamic annual crop-fallow-transition (AC-T) and dynamic grain-fallow (GF) AECs, respectively. By contrast, a 46% decrease in area was projected for stable AC-T and dynamic annual crop (AC) AECs across all future time periods for Representative Concentration Pathway 8.5. For the same scenarios, the stable AC and GF AECs showed the least declines in area (8% and 13%, respectively), compared to other AECs. Future spatial shifts from stable to dynamic AECs, particularly to dynamic AC-T and dynamic GF AECs would result in more use of fallow, a greater hazard for soil erosion, greater cropping system uncertainty, and potentially less cropping system flexibility. These projections are counter to cropping system goals of increasing intensification, diversification, and productivity.

Highlights

Changing climatic conditions have resulted in substantial shifts in the geographic range of plant and animal species in natural ecosystems (Gonzalez, 2001; Wilson et al, 2005; Pauli et al, 2007; Chen et al, 2011) and are predicted to continue in the future (Schrag et al, 2008; Lawler et al, 2009; Monadjem et al, 2013)
Our objectives are the following: (1) identify bioclimatic predictors which can discriminate among current agroecological classes (AECs) using two different Random Forest variable selection methods; (2) assess the predictive capacity of the geospatial models for current AECs using bioclimatic variables; (3) use future climate scenarios to predict changes in identified bioclimatic variables; (4) model regional shifts in AECs that would result if future climate scenarios were imposed on current agricultural systems; and (5) interpret the relevance of any AEC shifts in terms of sustainable agricultural intensification, vulnerability to resource degradation, and priorities for agricultural research
The VarSelRF method indicated that three bioclimatic variables, (Holdridge evapotranspiration index (HBIO), spring (Mar–May) precipitation (Psp), and precipitation of the warmest 4-months, Jun–Sep (Pcm1), modeled all six current AECs with an accuracy of 67% compared to 75% when using all 44 variables (Data not shown)

Summary

Introduction

Changing climatic conditions have resulted in substantial shifts in the geographic range of plant and animal species in natural ecosystems (Gonzalez, 2001; Wilson et al, 2005; Pauli et al, 2007; Chen et al, 2011) and are predicted to continue in the future (Schrag et al, 2008; Lawler et al, 2009; Monadjem et al, 2013). Single or multiple crop agro-ecosystems within a biophysical and socio-economic context have been affected by changing climatic variables (Kumar et al, 2013; Zhang et al, 2013). It follows that shifts in the geographic suitability of crop species/systems would occur in response to a changing climate (Evangelista et al, 2013; Ovalle-Rivera et al, 2015). The AECs are based on the actual annual land use/cover derived from the Cropland data layer (USDA-NASS, 2008–2015) This classification approach for defining AECs provides the opportunity to quantify and test hypotheses regarding the drivers of spatio-temporal changes in cropping systems

Objectives

Methods

Results

Discussion

Conclusion