Enso and Pdo Climate Variability Impacts on Regional

ENSO AND PDO CLIMATE VARIABILITY IMPACTS ON REGIONAL

MISSOURI CROP YIELDS

A Thesis Presented to the Faculty of the Graduate School at the University of Missouri

In Partial Fulfillment of the Requirements for the Degree

Master of Science

CHASITY B. HENSON

Dr. Patrick Market, Thesis Advisor

MAY 2016

The undersigned, appointed by the dean of the Graduate School, have examined the thesis entitled

ENSO AND PDO CLIMATE VARIABILITY IMPACTS ON REGIONAL MISSOURI

CROP YIELDS presented by Chasity B. Henson, a candidate for the degree of master of science, and hereby certify that, in their opinion, it is worthy of acceptance.

______

Professor Patrick Market

______

Professor Anthony Lupo

______

Professor Mark Palmer

ACKNOWLEDGEMENTS

My deepest gratitude is expressed to Dr. Patrick Market for being my advisor and motivator. I would also like to thank Dr. Anthony Lupo for being my co-advisor and for his help with the methodologies used in this study. Dr. Patrick Guinan also deserves a thank you for his contributions to this research. I recognize my thesis committee members, especially Dr. Mark Palmer, for taking the time to assess my performance as a graduate student. Suggestions and explanations from all four of these professors have greatly improved my education and the quality of this thesis. Lastly, I acknowledge Ryan

Difani, my fellow graduate student, for his support and advice, specifically on the creation of Fig. 5.1.

This work would not have been possible without support from Missouri EPSCoR

(Experimental Program to Stimulate Competitive Research). Being a chapter of the

National Science Foundation, the official disclaimer is as follows:

This material is based upon work supported by the National Science

Foundation under Award IIA-1355406. Any opinions, findings, and

conclusions or recommendations expressed in this material are those of the

author(s) and do not necessarily reflect the views of the National Science

Foundation.

TABLE OF CONTENTS

ACKNOWLEDGEMENTS ...... ii

LIST OF FIGURES ...... v

LIST OF TABLES ...... viii

ABSTRACT ...... x

CHAPTER 1. INTRODUCTION ...... 1

1.1 Objectives ...... 3

1.2 Statement of Thesis ...... 4

CHAPTER 2. LITERATURE REVIEW ...... 5

2.1 Definitions...... 5

2.2 ENSO ...... 8

2.2.1 ENSO Characteristics ...... 8

2.2.2 ENSO Impacts ...... 13

2.3 PDO Characteristics and Impacts ...... 16

2.4 ENSO and PDO Relationships ...... 21

2.4.1 PDO Dependence on ENSO ...... 22

2.4.2 ENSO Modulated by PDO ...... 23

2.4.3 PDO Modulated ENSO in the Midwest ...... 27

2.5 Climate Impacts on Crops ...... 30

iii

CHAPTER 3. DATA AND METHODOLOGY ...... 34

3.1 Data Sources ...... 34

3.2 Methods...... 36

3.2.1 Power Spectrum Analysis ...... 39

3.2.2 PDO Influence on ENSO ...... 41

CHAPTER 4. RESULTS ...... 43

4.1 Climatological Analysis ...... 43

4.2 ENSO and PDO Phase Interactions ...... 47

CHAPTER 5. DISCUSSION ...... 52

CHAPTER 6. CONCLUSIONS ...... 58

APPENDIX A ...... 60

APPENDIX B ...... 84

REFERENCES ...... 88

LIST OF FIGURES

Figure 3.1. A map of Missouri outlining the six climate regions, determined by the National Oceanic and Atmospheric Administration (NOAA) using the Climate Divisional Dataset. NOAA defines these regions as Climate Divisions of Missouri...... 35

Figure 3.2. Annual corn yields for region 1 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. Data provided by the United States Department of Agriculture (USDA), graphed as yield (bushels per acre) versus time (year)...... 37

Figure 4.1. Power spectrum resulting from the Fourier transform of a) detrended corn yields for region 1 in Missouri, found in Fig. 3.2b, b) average temperature (°F) for region 1 over the corn growing season of April through September, c) average precipitation (in.) for region 1 from April through September, and d) the convolution of the corn yield spectrum (Fig. 4.1a) with the spectra of seasonal temperature (Fig. 4.1b) and seasonal precipitation data (Fig. 4.1c). Average temperature and precipitation data were downloaded from the Midwestern Regional Climate Center (MRCC) for the years of 1919 to 2013. The ordinate displays wave power, which is the magnitude of the Fourier coefficients, while the abscissa displays wave number. The dashed (dotted) line represents the 95% confidence level against the red (white) noise background continuum (Wilks 2006)...... 44

Figure 4.2. As in Fig. 4.1, except involving the convolution of a) the annual SOI spectrum with b) the spectrum of annual mean values for the PDO index. Annual SOI data were obtained from the Bureau of Meteorology (2005) for the time period of 1901 to 2004 (www.environment.gov.au/node/22307). Annual mean values for the PDO index were obtained from the Japan Meteorological Agency (JMA) for the same time period of 1901 to 2004 (ds.data.jma.go.jp/tcc/tcc/products/elnino/decadal/pdo.html)...... 48

Figure 4.3. A map of Missouri outlining the six climate regions, defined by NOAA, including the statistically significant results found for both crops. El Niño/La Niña refers to the statistically significant difference in yields between El Niño years and La Niña years, found in Table 4.2. PDO/El Niño refers to the significant difference in yields between El Niño/Positive PDO years and El Niño/Negative PDO years, found in Table 4.3...... 51

Figure 5.1. Timeline representing the crop growing season of 2012 (yellow), which runs from April 2012 to September 2012, and the JMA defined ENSO year of 2012 (red), which begins in October 2012 and ends in September 2013...... 54

Figure A.1. As in Fig. 4.1d, except involving average July temperature (T) and average July precipitation (P) data for region 1...... 61

Figure A.2. As in Fig. 3.2, except annual corn yields for region 2 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. ...62

Figure A.3. As in Fig. 4.1, except power spectrum resulting from the Fourier transform of a) detrended corn yields for region 2 in Missouri, found in Fig. A.2b, b) the convolution of the corn yield spectrum (Fig. A.3a) with the spectra of seasonal temperature (not shown) and seasonal precipitation data (not shown), and c) the convolution of the corn yield spectrum (Fig. A.3a) with the spectra of average July temperature (not shown) and average July precipitation data (not shown)...... 63

Figure A.4. As in Fig. 3.2, except annual corn yields for region 3 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. ...64

Figure A.5. As in Fig. A.3, except involving detrended corn yields for region 3 in Missouri, found in Fig. A.4b...... 65

Figure A.6. As in Fig. 3.2, except annual corn yields for region 4 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. ...66

Figure A.7. As in Fig. A.3, except involving detrended corn yields for region 4 in Missouri, found in Fig. A.6b...... 67

Figure A.8. As in Fig. 3.2, except annual corn yields for region 5 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. ...68

Figure A.9. As in Fig. A.3, except involving detrended corn yields for region 5 in Missouri, found in Fig. A.8b...... 69

Figure A.10. As in Fig. 3.2, except annual corn yields for region 6 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed. ...70

Figure A.11. As in Fig. A.3, except involving detrended corn yields for region 6 in Missouri, found in Fig. A.10b...... 71

Figure A.12. As in Fig. 3.2, except annual soybean yields for region 1 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 72

Figure A.13. As in Fig. 4.1, except power spectrum resulting from the Fourier transform of a) detrended soybean yields for region 1 in Missouri, found in Fig. A.12b, b) the convolution of the soybean yield spectrum (Fig. A.13a) with the spectra of seasonal temperature (not shown) and seasonal precipitation data (not shown), and c) the convolution of the soybean yield spectrum (Fig. A.13a) with the spectra of average August temperature (not shown) and average

August precipitation data (not shown). Average temperature and precipitation data were downloaded from the Midwestern Regional Climate Center (MRCC) for the years of 1944 to 2013...... 73

Figure A.14. As in Fig. 3.2, except annual soybean yields for region 2 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 74

Figure A.15. As in Fig. A.13, except involving detrended soybean yields for region 2 in Missouri, found in Fig. A.14b...... 75 Figure A.16. As in Fig. 3.2, except annual soybean yields for region 3 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 76

Figure A.17. As in Fig. A.13, except involving detrended soybean yields for region 3 in Missouri, found in Fig. A.15b...... 77

Figure A.18. As in Fig. 3.2, except annual soybean yields for region 4 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 78

Figure A.19. As in Fig. A.13, except involving detrended soybean yields for region 4 in Missouri, found in Fig. A.18b...... 79

Figure A.20. As in Fig. 3.2, except annual soybean yields for region 5 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 80

Figure A.21. As in Fig. A.13, except involving detrended soybean yields for region 5 in Missouri, found in Fig. A.20b...... 81

Figure A.22. As in Fig. 3.2, except annual soybean yields for region 6 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed...... 82

Figure A.23. As in Fig. A.13, except involving detrended soybean yields for region 6 in Missouri, found in Fig. A.22b...... 83

Figure B.1. Timeline representing the crop growing season of 2012 (yellow), which runs from April 2012 to September 2012, and the JMA defined ENSO year of 2011 (red), which begins in October 2011 and ends in September 2012...... 85

vii

LIST OF TABLES

Table 2.1. The years examined in this study, corresponding to each phase of the El Niño Southern Oscillation (ENSO), as defined by the Center for Ocean- Atmospheric Prediction Studies (COAPS, found online at coaps.fsu.edu/jma)...... 6

Table 2.2. The years examined in this study, corresponding to each phase of the Pacific Decadal Oscillation (PDO), adapted from Birk et al. (2010)...... 7

Table 3.1. Correlation coefficients resulting from a multiple correlation function, with detrended crop yields as the dependent variable and independent variables of temperature and precipitation data. A correlation coefficient of zero represents no correlation between the variables...... 38

Table 3.2. The years examined in this study, corresponding to each El Niño Southern Oscillation (ENSO) phase, as well as each Pacific Decadal Oscillation (PDO) phase...... 42

Table 4.1. Regional Missouri crop yield variability resulting from each analyzed power spectrum. Periodicities determined from each spectrum were compared to each climate variability and listed as ENSO, ENSO-PDO, or PDO. ENSO refers to an interannual variability of 2 to 7 years, corresponding to the El Niño Southern Oscillation. ENSO-PDO refers to an interdecadal variability of roughly 10 to 15 years, corresponding to PDO modulated ENSO-related variability. PDO refers to a multidecadal variability of roughly 20 years, corresponding to the Pacific Decadal Oscillation...... 46

Table 4.2. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation (ENSO) phase, listed in Table 2.1. Bold values represent statistically significant (90% confidence level) differences in yield between La Niña and El Niño years, with both data sets involved containing at least 5 samples (yield per ENSO phase year)...... 49

Table 4.3. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation-Pacific Decadal Oscillation (ENSO- PDO) phase combination, listed in Table 3.2. Bold values represent statistically significant (90% confidence level) differences in yield for El Niño years of different PDO phases, with both data sets involved containing at least 5 samples (yield per phase year combination)...... 50

viii

Table B.1. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation (ENSO) phase of the previous JMA ENSO year, listed in Table 2.1. Bold values represent statistically significant (90% confidence level) differences in yield between La Niña and El Niño years, with both data sets involved containing at least 5 samples (yield per ENSO phase year)...... 86

Table B.2. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation-Pacific Decadal Oscillation (ENSO- PDO) phase combination of the previous JMA ENSO year, listed in Table 3.1. Bold values represent statistically significant (90% confidence level) differences in yield for El Niño years of different PDO phases, with both data sets involved containing at least 5 samples (yield per phase year combination)...... 87

ABSTRACT

An analysis of crop yields for the state of Missouri was completed to determine if an interannual or multidecadal variability existed as a result of the El Niño Southern

Oscillation (ENSO) and the Pacific Decadal Oscillation (PDO). Corn and soybean yields were recorded in bushels per acre for each of the six climate regions of Missouri. An analysis using the Mokhov “method of cycles” demonstrated interannual, interdecadal, and multidecadal variations in crop yields. Cross-spectral analysis was then used to determine which region was impacted the greatest by ENSO and PDO influenced seasonal (April – September) and monthly temperature and precipitation. Interannual

(multidecadal) variations found in the spectral analysis represent a relationship to ENSO

(PDO) phase, while interdecadal variations represent a possible interaction harmonic between ENSO and PDO. A cross-spectral analysis was also completed using annual

Southern Oscillation Index data and annual mean values for the PDO index in order to verify that an interdecadal variation exists between ENSO and PDO. Average crop yields were then calculated for each combination of ENSO and PDO phase, displaying a pronounced increase in corn and soybean yields when ENSO is warm and PDO is positive. Climate regions 1, 2, 4, and 6 displayed statistically significant (90% confidence level) differences in yields between El Niño and La Niña years, representing 55-70% of

Missouri soybean and corn productivity, respectively. Final results give the opportunity to produce seasonal predictions of corn and soybean yields, specific to each climate region in Missouri, based on ENSO and PDO phase.

CHAPTER 1. INTRODUCTION

Variations in climate, especially those related to the El Niño Southern Oscillation

(ENSO), are commonly researched in the field of atmospheric science (Kung and Chern

1995; Lupo et al. 2005, 2012a). However, climate variability related to both ENSO and the Pacific Decadal Oscillation (PDO) is a growing topic after correlations were found with North American climate (Gershunov and Barnett 1998; Enfield and Mestas-Nuñez

1999; Ding and McCarl 2014), including Midwestern temperatures and precipitation

(Berger et al. 2002; Lupo et al. 2007; Birk et al. 2010; Newberry et al. 2016). ENSO leads to changes in upper air conditions, specifically mid-tropospheric circulation and jet stream positioning across the United States (Keables 1992; Kung and Chern 1995; Lee and Kung 2000), which impacts surface temperature and precipitation (Lupo et al. 2007,

2008).

A relationship has been found between temperature and precipitation and crop yields in Missouri (Hu and Buyanovsky 2003), but only a few of studies have applied

ENSO and PDO-related climate variability to crop yields in the United States (Ding and

McCarl 2014). Crop yield variations are important to understand in order to create seasonal forecasts for field crop production. Seasonal forecasts of crop yields, on a timescale of three to twelve months, may become more useful if a link can be made to climate variability, specifically ENSO and PDO.

ENSO events have been linked to interannual variations of two to seven years in local and regional climates (Kung and Chern 1995; Enfield and Mestas-Nuñez 1999;

Lupo et al. 2005), while PDO events have been linked to multidecadal variations of twenty to thirty years in North American climate (Lupo et al. 2007; Ding and McCarl

2014). A third signal occurs on an interdecadal timescale of roughly ten to fifteen years and is referred to as a PDO modulated ENSO-related variability of Midwestern climates

(Gershunov and Barnett 1998; Berger et al. 2002; Birk et al. 2010). Lupo et al. (2007) extended the work of Kung and Chern (1995) showing different clusters of monthly mean

Pacific Region sea surface temperature (SST) anomalies corresponding to different phases of ENSO, as well as different phases of PDO. Results from Lupo et al. (2007) suggest that ENSO events are modulated by long-term SST variability associated with

PDO; with strong El Niño and weak La Niña SST distributions occurring mostly during a positive PDO phase, weaker El Niño and stronger La Niña SST distributions common during the negative PDO phase, and El Niño events during the negative PDO phase featuring warm SST anomalies in the east-central Pacific.

Further research involving sea level pressure (SLP), temperatures, and precipitation supported the findings of Lupo et al. (2007), showing strong El Niño signals occurring during years of positive PDO, while strong La Niña events occur during years of negative PDO (Gershunov and Barnett 1998; Birk et al. 2010). More evidence of an interdecadal modulation is found in studies involving Australia and Asia rainfall, and

Arizona winter precipitation (Power et al. 1999; Goodrich 2004; Wang et al. 2008).

Correlations between ENSO and PDO phases in the aforementioned studies strongly suggest that PDO phase can contribute to the strength and frequency of the ENSO phase, therefore creating the interdecadal variability in climate. It is acknowledged that the

ENSO and PDO relationship is controversial, in that research studies supporting the PDO

dependence on ENSO, rather than the ENSO dependence on PDO, exist in atmospheric science literature (Newman et al. 2003; Zhang and Delworth 2015). This study will focus on how ENSO phase impacts on Missouri crop yields differ depending on PDO phase.

1.1 Objectives

With crop yields in Missouri, specifically corn, showing a correlation to temperature and precipitation (Hu and Buyanovsky 2003), and Midwest temperatures and precipitation correlating to ENSO and PDO (Berger et al. 2002; Birk et al. 2010), the hypothesis stands that Missouri crop yields will also correlate to ENSO and PDO.

Another question remains as to whether or not Missouri crop yield data will support the

Gershunov and Barnett (1998) and Birk et al. (2010) findings of PDO modulated ENSO events. Thus, the objectives of this research are as follows:

1. Determine if variability in Missouri crop yields exists in relationship to ENSO

and PDO variability.

2. Determine if interdecadal variability is present in crop yields, consistent with

the theory of PDO and ENSO-related variability (Gershunov and Barnett

1998; Birk et al. 2010).

1.2 Statement of Thesis

This research is intended for the advancement of knowledge relating climate variability to crop yields in Missouri. Specifically, this study will describe:

1. ENSO and PDO impacts differ for each climate region in Missouri.

2. PDO can enhance or diminish the ENSO impact on Missouri corn and

soybean yields.

3. The quantitative relationship between crop yields and ENSO and PDO phases

can be calculated for future use in seasonal yield predictions.

CHAPTER 2. LITERATURE REVIEW

2.1 Definitions

The ENSO definition is provided by the Japan Meteorological Agency (JMA) and is used in several studies (Berger et al. 2002; Lupo et al. 2007; Birk et al. 2010; Lupo et al. 2012a). The JMA ENSO Index was used in this study in order to stay consistent with previously referenced studies and due to the fact that Hanley et al. (2003) found the JMA

ENSO Index to be the most sensitive to La Niña events compared to five other ENSO indices. The JMA ENSO Index is described by the Center for Ocean-Atmospheric

Prediction Studies (COAPS) and is a 5-month running mean of spatially averaged sea surface temperature (SST) anomalies over an area in the tropical Pacific (4°S-4°N,

150°W-90°W). There are three phases of ENSO: La Niña (cool phase) is based on SST anomalies at or below -0.5 °C, neutral is between -0.5 °C and 0.5 °C, and El Niño (warm phase) is based on SST anomalies at or above 0.5 °C. The JMA ENSO Index requires the index values for each phase to be observed for 6 consecutive months, including October,

November, and December, in order to be classified.

The ENSO year for JMA starts on 1 October and continues through the following

September. Corn and soybean mean growing season in Missouri runs from April to

September, meaning the ENSO year begins at the end of the crop growing season, after harvest. For example, crop yields for 2012 are the product of the April 2012 to September

2012 growing season and are compared to the ENSO year for 2012, which JMA defines

Table 2.1. The years examined in this study, corresponding to each phase of the El Niño Southern Oscillation (ENSO), as defined by the Center for Ocean-Atmospheric Prediction Studies (COAPS, found online at coaps.fsu.edu/jma).

La Niña Neutral El Niño 1922 1919-1921 1918 1924 1923 1925 1938 1926-1928 1929 1942 1931-1937 1930 1944 1939 1940 1949 1941 1951 1954-1956 1943 1957 1964 1945-1948 1963 1967 1950 1965 1970 1952 1969 1971 1953 1972 1973-1975 1958-1962 1976 1988 1966 1982 1998 1968 1986 1999 1977-1981 1987 2007 1983-1985 1991 2010 1989 1997 1990 2002 1992-1996 2006 2000 2009 2001 2003-2005 2008 2011-2013

as beginning in October 2012 and ending in September 2013. A list of years by ENSO phase is found in Table 2.1, which was adapted from COAPS (coaps.fsu.edu/jma).

PDO can be described as a long-lived ENSO-like pattern of SST anomalies oscillating over a period of about twenty years (Mantua et al. 1997; Minobe 1997; Zhang et al. 1997; Minobe 2000; Mantua and Hare 2002). There are two phases of PDO: the

Table 2.2. The years examined in this study, corresponding to each phase of the Pacific Decadal Oscillation (PDO), adapted from Birk et al. (2010).

Positive (1) Negative (2) 1925-1946 1919-1924 1977-1998 1947-1976 1999-2013

positive phase (phase 1) is characterized by cold western and north central Pacific waters with warm eastern and tropical Pacific waters, as well as an anomalously deep Aleutian low; the negative phase (phase 2) occurs during the opposite conditions (Gershunov and

Barnett 1998; Mantua and Hare 2002; Zhang and Delworth 2015). A list of years by PDO phase is found in Table 2.2.

ENSO and PDO phases are determined and quantified by indices. ENSO events can be measured by the Southern Oscillation Index (SOI), which is a standardized index based on observed sea level pressure (SLP) differences between the western and eastern tropical Pacific, according to NOAA. Positive SOI values are associated with La Niña, while negative SOI values are associated with El Niño. PDO is measured by the PDO index, which is defined by JMA as the projections of monthly mean SST anomalies onto their first Empirical Orthogonal Function (EOF) vectors (derived for the period of 1901 to 2000) in the North Pacific, poleward of 20°N. Positive PDO index values represent phase 1, while negative PDO index values represent phase 2. Annual SOI data were downloaded from the Bureau of Meteorology (2005) for the time period of 1901 to 2004

(www.environment.gov.au/node/22307). Annual mean values for the PDO index were

downloaded from JMA for the same time period of 1901 to 2004

(ds.data.jma.go.jp/tcc/tcc/products/elnino/decadal/pdo.html).

2.2 ENSO

The El Niño Southern Oscillation (ENSO) is a popular topic when it comes to studying climate. For decades, research has been completed to determine why and how

ENSO impacts different parts of the United States. Studies focused on different aspects of

ENSO, from specific characteristics of the phenomenon on a global scale, to regional and local variables impacted by ENSO variability.

2.2.1 ENSO Characteristics

Enfield and Mestas-Nuñez (1999) described how ENSO influences interannual variations in global climate characteristics. Specifically, Enfield and Mestas-Nuñez

(1999) used an EOF analysis to represent ENSO modes in sea surface temperature (SST) anomalies from 1870 to 1991. From this analysis, the spatial amplitude and phase, as well as the temporal behavior of the mode can be determined. The spatial functions resulting from this analysis showed what Enfield and Mestas-Nuñez (1999) refer to as “classic features” that are normally associated with ENSO in the Pacific Ocean. These features include a region of increased amplitude within ten degrees of the equator and east of the date line, with smaller amplitudes over a wedge-shaped region stretching poleward along

the eastern Pacific; phase propagation to the north along the coast of North America, with the Gulf of Alaska and the Bering Sea having maximum lags of one to two seasons; and regions of intermediate amplitude and opposite phase in the central North Pacific region between 30° and 45°N, as well as the central South Pacific region between 20° and 40°S.

Less apparent features of ENSO include one-season lags in the central Pacific region between 20°N and 20°S and a one-season lead precursor off of central Chile, which lies along a well sampled ship route just to the north of Cape Horn. Enfield and Mestas-

Nuñez (1999) speculate a possible cause of the precursor to be weakening trade winds and associated surface heat fluxes off of Chile, prior to the main weakening of trade winds at lower latitudes.

On the global scale, spatial functions associated with ENSO can be found in the tropical Atlantic and Indian Oceans (Enfield and Mestas-Nuñez 1999). Both oceans were found to have coherence and lag structures relative to the equatorial Pacific Ocean.

Specifically, the tropical Atlantic Ocean has a one to three season lag with phase propagation toward the equator, along with the greatest amplitude being between 15°-

20°N and 15°-20°S and having a two-season lag. Enfield and Mestas-Nuñez (1999) found little coherent variability in the North Atlantic, north of 20°-30°N, as well as the region off of Angola between 0°-20°S and the western equatorial Atlantic. The Indian Ocean is similarly lagged as the Atlantic Ocean, with phase propagation to the east.

Enfield and Mestas-Nuñez (1999) continued their analysis by comparing non-

ENSO modes to tropospheric conditions, using distributions of December to February geopotential height for the 500-hPa level. Anomalous tropical SST distributions can change heat and mass distributions of the troposphere and can alter prevailing wind

patterns across the globe (Keables 1992; Enfield and Mestas-Nuñez 1999). Keables

(1992) focused on how these SST distributions, especially those associated with warm

ENSO (El Niño) events, can alter the mid-tropospheric circulation pattern over North

America in the winter.

Keables (1992) found there to be three primary responses of the winter mid- tropospheric circulation during ENSO events. These responses are referred to as types of the Pacific/North American (PNA) pattern. PNA Type 1 features the most characteristic synoptic circulation pattern, with a pronounced ridge over western North America and a trough positioned to the east. This results in a strong meridional circulation regime across the continent with an anomalous northerly component of flow over the central portion of

North America. Therefore, the region extending from the central Great Plains to the

Northwest experiences above average temperatures, due to the warm core associated with the ridge aloft (Keables 1992). The southeastern United States experiences below average temperatures, due to the northerly flow associated with the position of the trough. As for precipitation, the ridge and increased convergence leads to below average precipitation for the western and central portions of the country, with average conditions over the central Midwest and the East Coast.

Keables (1992) describes PNA Type 2 as having an eastward displacement of the

Pacific negative and North American positive anomalies, with a westward displacement of the North American negative anomaly. PNA Type 3 is the same as PNA Type 1, except all anomalies are displaced eastward. A reverse PNA pattern was also found in

Keables (1992) study, associated with zonal flow across North America. The positioning of troughs and ridges determines the spatial distribution of temperature and precipitation

for the United States during PNA winters associated with specific ENSO events (Keables

1992).

Synoptic circulation patterns were also analyzed by Kung and Chern (1995) and

Lupo et al. (2007, 2008). Kung and Chern (1995) used a principal component analysis to represent fundamental large-scale modes of variation in global sea surface temperatures

(SSTs) and Northern Hemisphere tropospheric circulation over the period of 1955 to

1993. Lupo et al. (2007) extended this analysis period from 1993 to 2005, while Lupo et al. (2008) extended it even further to 2007.

Kung and Chern (1995) generated seven classifications of global, monthly SST anomaly fields, labeled type A through G. Each type of anomaly field reflected different principal components of SSTs. For example, Kung and Chern (1995) found the F type of anomaly distribution to be observed during ENSO events, reflecting the 1st and 2nd principal components of SSTs. However, the C type was also observed during ENSO events, but only reflects the 2nd principal component. Each type of anomaly distribution

(A-G) has specific characteristics. ENSO events can be observed with types C, D, and F.

From these classifications, Kung and Chern (1995) determined the prevailing climate pattern for the 1980s and 1990s to be the result of ENSO mode dominance since the late

1970s, causing recurring major ENSO events. The ENSO-dominated climate is associated with exceptionally strong anomalies in the tropical ocean and subtropical to mid-latitude SST patterns, which are shown in the 2nd and 3rd principal components.

Kung and Chern (1995) used the monthly SST anomaly classifications to describe tropospheric circulation. The anomaly types associated with ENSO events can describe the circulation patterns associated with the same ENSO events. Type C displayed

negative SST anomalies in the North Pacific and Atlantic Ocean, north of the equator, with high pressure areas in the Northwestern Pacific Ocean, eastern Asian continent,

Northwestern Atlantic Ocean, and Canada. Type C is comparable to PNA Type 1 found in Keables (1992). Type F represents the strongest ENSO events with strong high pressure areas over the North American continent, North Atlantic Ocean, and the Asian continent, along with a low pressure area over the Pacific Ocean. Therefore, type F is analogous to the previously described PNA Type 2. Type D is similar to type F, except for an eastward shift in the general circulation pattern, reflecting cooling in the North

Pacific and Atlantic Oceans. The eastward shift in the circulation pattern was also found in PNA Type 3 from Keables (1992).

Kung and Chern (1995) developed an El Niño Index (ENI) from the weighted means of the coefficients of the 1st and 2nd principal components of SSTs, which can be cross-correlated with the following tropospheric circulation to indicate the seasonal range predictability of the Northern Hemisphere circulation. Lupo et al. (2007, 2008) used the expanded Kung and Chern (1995) SST anomaly classifications to create seasonal forecasts, four to five months in advance, for temperature and precipitation in the mid-

Mississippi valley region. Results determined SST classifications of types B and G to represent La Niña events, types C, D, and F to represent El Niño events, and types A and

E representative of neutral conditions in the Pacific Ocean. Type B was found to be associated with warmer temperatures, type C produced cooler temperatures and greater than normal precipitation, while type D was cool and dry. Type G could represent zonal flow over North America causing warm (cool) conditions during the cold (warm) season for the mid-Mississippi region, which is consistent with the characteristics of the reverse

PNA pattern described in Keables (1992). Type F, which was associated with strong El

Niño events, caused very warm winters for the region.

Lee and Kung (2000) also created seasonal-range forecasts for temperature and precipitation in the Ozark Highlands area, using the predictors of global SSTs and hemispheric upper air fields. Lee and Kung (2000) used a similar analysis as Lupo et al.

(2007, 2008) involving principal components, and found this method to be superior to a multiple linear regression scheme for forecasting. Lee and Kung (2000) found significant correlations between January temperature and precipitation in the Ozark region and preceding SSTs, upper air temperatures, geopotential heights, and geostrophic winds.

Therefore, these four variables can be used as predictors for local temperature and precipitation in the Ozark Highlands area. It is now commonly recognized that SSTs can alter the heating distributions of the tropical troposphere, which influences the general circulation of the atmosphere, making SSTs viable predictors for long-range forecasting of atmospheric conditions (Kung and Chern 1995; Enfield and Mestas-Nuñez 1999; Lupo et al. 2007, 2008; Birk et al. 2010).

2.2.2 ENSO Impacts

ENSO occurs on an interannual timescale, with El Niño events occurring when central tropical Pacific SSTs are warmer than normal and La Niña events occurring when these SSTs are cooler than normal. In general for North America, El Niño events lead to wetter/cooler than average conditions in the southeast and warmer/drier conditions in the northwest (Keables 1992). However, each ENSO phase has a specific impact on certain

regions of the United States. For example, the Southwest experiences above average winter precipitation during El Niño events and La Niña is associated with below average winter precipitation and prolonged drought (Goodrich 2004). One of the most common regions in the United States to be studied in regard to ENSO impacts, is the Midwest region, including Missouri.

ENSO has an impact on Midwestern temperature and precipitation through the relationship between SST anomalies and tropospheric circulation (Lee and Kung 2000;

Lupo et al. 2007, 2008). Several studies have been completed to investigate possible associations between Midwest weather variables and ENSO. These variables range from dew point temperatures to snowfall, covering regional or local areas.

Snowfall in Missouri has been examined for possible links to ENSO variability by two different studies (Berger et al. 2002; Lupo et al. 2005). Berger et al. (2002) researched the interannual variability of snowfall occurrences in Northwest Missouri.

Results displayed more snowfall events occurring during La Niña events than during El

Niño events. The reasoning for fewer snowfall events in El Niño winters is because of fewer mid-latitude cyclones over the eastern two-thirds of the United States during El

Niño winters (Berger et al. 2002). Synoptic-scale flow regime was also considered in the

Lupo et al. (2005) study involving Southwest Missouri snowfall. Large-scale flow over

North America tends to be a more zonal pattern during El Niño years (Keables 1992;

Lupo et al. 2005), resulting in fewer cyclones and more southwest low events. Lupo et al.

(2005) found that El Niño years may be associated with warmer, more humid air in the

Southwest Missouri region, resulting in larger snowfall totals with fewer snowfall events than in La Niña years.

Lupo et al. (2012a) found consistent results after examining dew point temperatures across the state of Missouri in search for interannual variability related to

ENSO. El Niño years had higher dew point temperatures than La Niña years, meaning more moisture in the air, due to the location and strength of the low-level jet. La Niña years were associated with drier air, which led to more wildfires during those years for

Missouri (Lupo et al. 2012a). Long-range forecasting was the goal for Lupo et al.

(2012a), but ENSO behavior, referring to the strength and transitioning of ENSO phases, over the summer season became an issue. Conclusions were that the summer transition of

ENSO could be an important aspect when creating forecasts (Lupo et al. 2012a).

Recently, this issue of ENSO transitioning has been specifically examined for the

Midwestern United States (Ratley et al. 2002; Newberry et al. 2016), as well as for a region in Russia (Lupo et al. 2014). A severe drought in Western Russia during the summer of 2010 inspired the research of Lupo et al. (2014). It was found that ENSO phase didn’t correlate exactly to drier summers, but the transition between ENSO phases did. Drier summers were more common in Western Russia during transitions toward La

Niña. Conflicting results did occur, with a smaller region in Western Russia represented by drier summers during transitions toward El Niño. Nonetheless, the transition of ENSO phases was found to be an important indicator of summer conditions in Russia.

Ratley et al. (2002) performed a study on the transition from a spring to summer flow regime, which involved the transition of ENSO as well, for the Missouri Ozarks region. Results found there to be an abrupt transition between flow regimes, along with an abrupt change in the average period between heavy precipitation events. Lower precipitation frequencies were found to be associated with La Niña summers, as

compared to El Niño summers. Further examination determined late arriving summers for the Missouri Ozarks to be associated with transitioning ENSO phases. Newberry et al.

(2016) contains similar results for the West Central Plains region, with ENSO transitions toward La Niña events associated with warmer and drier summers. Therefore, it is important to consider not only the ENSO phase, but the transition into the upcoming

ENSO phase, when making long-range forecasts in the Midwestern United States.

2.3 PDO Characteristics and Impacts

A long-lived ENSO-like variability of Pacific SSTs is referred to as the Pacific

Decadal Oscillation (Mantua et al. 1997; Zhang et al. 1997; Mantua and Hare 2002). The

Pacific Decadal Oscillation (PDO) fluctuates in two different periodicities, one about 20 to 25 years and another from 50 to 70 years (Minobe 2000; Mantua and Hare 2002;

Zhang and Delworth 2015). Zhang et al. (1997) were the first to compare the long-lived

PDO variability to ENSO variability. Using a complex method involving empirical orthogonal function/principal component analysis of SST anomaly fields, Zhang et al.

(1997) were able to document the spatial signature of PDO variability after removing the

ENSO signature through linear regression. Results show ENSO and PDO having similar spatial signatures in global SST, SLP, and wind stress fields. However, ENSO has an equatorially confined SST signature in the eastern Pacific, while PDO has a more prominent SST and SLP signature over the extratropical North Pacific. Further, PDO is associated with a 500-mb height pattern similar to the Pacific/North American (PNA)

pattern found in Keables (1992), with ENSO associated with positive regression coefficients that extend farther east across Canada. The differences in 500-mb height patterns are consistent with the differences in surface air temperature patterns (Zhang et al. 1997).

Finding a difference between ENSO-related interannual variability and PDO variability led to further investigation of PDO. Minobe (1997) specifically studied the 50 to 70 year oscillation pattern, but acknowledges shifts in PDO occurring about 20 years apart. Using empirical orthogonal analysis with reconstructed temperature data from instrumental measurements and tree-ring widths, Minobe (1997) found strong evidence to support the existence of PDO variability in North America. Shifts in PDO phase were found to be associated with the deepening of the Aleutian low, with a strong (weak)

Aleutian low corresponding to warmer (cooler) winter and spring temperatures in western

North America (Minobe 1997). SLP and North American temperature is related due to the strengthening (weakening) of the Aleutian low causing an enhancement (reduction) of warm air advection onto the west coast of North America. Minobe (1997) found similar results as Zhang et al. (1997), with spatial distributions of SLP differences exhibiting patterns similar to the PNA circulation pattern (Keables 1992). Results from Minobe

(1997) suggest the PDO variability to essentially be an internal oscillation in the coupled atmosphere-ocean system.

Shifts in PDO-related climate regime were found to occur in the 1890s, 1920s,

1940s, 1970s, and 1990s (Mantua et al. 1997; Minobe 1997; Zhang et al. 1997). Mantua et al. (1997) studied the influence of these regime shifts on salmon production in the

North Pacific Ocean, concluding that positive PDO phases contributed to increased

production. Mantua et al. (1997) also described how the winter PNA pattern associated with PDO (Minobe 1997; Zhang et al. 1997) suggests that a PDO in positive phase will lead to higher than average precipitation for Alaska, due to an enhanced cyclonic flow of warm, moist air. British Columbia and Washington state will also experience an increase in warm, moist air, but the dynamics associated with an enhanced anticyclonic circulation keep the precipitation about average during positive PDO events. Similarly, positive

(negative) PDO years contribute to reduced (enhanced) Pacific Northwest snowpack, due to warm (cool) winter temperatures and low (high) precipitation. Further analysis suggests that PDO and the associated PNA pattern impacts Idaho snowpack, Alaska temperature, SST near the North American west coast, and annual water discharge from

Alaskan rivers (Mantua et al. 1997). PDO was found to be positively correlated with winter precipitation over Alaska, northern Mexico, and south Florida, while negatively correlated with that over central North America and Hawaii (Mantua et al. 1997).

Minobe (2000) also performed an analysis on PDO regime shifts using wavelets.

It was concluded that the shifts in the 1920s, 1940s, and 1970s involved phase reversals of the bidecadal (referring to the 20 year fluctuation in PDO) and the pentadecadal (50 to

70 year) variations, verifying the two periodicities related to PDO variability. Minobe

(2000) suggests that the two variabilities interact with each other, with evidence found in the North Pacific Index and Alaska temperatures.

With several studies hinting at the existence and characteristics of PDO, it was time for a complete description of this climate variability. Mantua and Hare (2002) describe PDO in terms of temporal scales, paleoclimate, spatial patterns, dynamics, and impacts on surface climate and marine ecosystems. Mantua and Hare (2002) reiterate the

fact that PDO fluctuates at two different periodicities, consistent with the aforementioned studies. The periodicity of approximately 20 years was found in a reconstruction of PDO using tree-ring chronologies from the North American west coast, going back to the year

1650 (Mantua and Hare 2002). Evidence of observed, rather than reconstructed, phase changes display only two full PDO cycles in the past century. Beginning in 1890, PDO was in a “cool” or negative phase until 1924, and again from 1947 to 1976. PDO was in a

“warm” or positive phase from 1925 to 1946 and 1977 to 1998.

A positive PDO phase is associated with anomalously cool SSTs in the central

North Pacific and anomalously warm SSTs along the west coast of the Americas (Mantua and Hare 2002; Zhang and Delworth 2015). Average SLP anomalies, from November to

March, display low pressure over the North Pacific (a strengthened Aleutian low shifted southward) for positive PDO phases, causing enhanced counterclockwise winds. High pressure is associated with positive PDO over the northern subtropical Pacific, causing enhanced clockwise winds (Mantua et al. 1997; Mantua and Hare 2002; Zhang and

Delworth 2015). These characteristics suggest that PDO circulation anomalies extend through the depth of the troposphere and are consistent with the PNA pattern (Minobe

1997; Zhang et al. 1997; Mantua and Hare 2002).

Zhang and Delworth (2015) agree with Mantua et al. (1997) that the cyclonic wind associated with a positive PDO phase drives an anomalous northwestward current along the North American coast. Consistent with Mantua et al. (1997), Zhang and

Delworth (2015) explain that this current creates warm air advection from low latitudes to high latitudes, generating warmer temperatures for the North Pacific and higher than average precipitation for Alaska. Conversely, anticyclonic wind drives a southeastward

current along the coast, creating cold air advection from high to low latitudes, generating cooler temperatures for the northern subtropical Pacific. These patterns of PDO were derived from linear analyses, making the anomalies associated with a negative PDO phase simply the opposite of those associated with a positive PDO phase.

Mantua and Hare (2002) speculated on the physical mechanisms of PDO, linking

SLP to wind and temperature fields which creates a positive feedback system involving

Rossby waves and thermocline anomalies. Zhang and Delworth (2015) describe possible

PDO mechanisms in great detail, involving Rossby wave propagation, interactions with the tropics, and extratropical air-sea interactions. The specific cause of PDO is still unclear, but using PDO as a predictor is still possible due to the consistent periodicity of roughly 20 years and the already discovered impacts on surface climate (Mantua et al.

1997; Minobe 1997; Zhang et al. 1997; Mantua and Hare 2002).

PDO impacts on surface climate are not limited to North America. A positive

PDO phase is associated with dry periods in eastern Australia, Korea, Japan, the Russian

Far East, interior Alaska, a zonally elongated belt in the United States from the Pacific

Northwest to the Great Lakes, the Ohio Valley, and in much of Central America and northern South America (Mantua and Hare 2002). In contrast, a positive PDO phase is associated with wet periods in the coastal Gulf of Alaska, the southwest United States and Mexico, southeast Brazil, south central South America, and western Australia. As for temperature anomalies, a positive PDO phase coincides with anomalously warm temperatures in northwestern North America, northern South America, and northwestern

Australia, with cool temperatures in eastern Asia, the southeast United States, and

Mexico (Mantua and Hare 2002). PDO temperature signals are most prominent in the

boreal spring, rather than winter season, in North America (Minobe 2000; Mantua and

Hare 2002).

Even though PDO may be related to surface climate around the globe, there are no strong PDO signatures found in the Atlantic or Indian Ocean SST and SLP fields

(Mantua et al. 1997). PDO impacts may be more common in countries along the Pacific

Ocean, like North America and Australia. Power et al. (1999) found PDO-related variability in eastern Australian climate, with positive PDO contributing to warm and dry conditions and negative PDO associated with cool and wet conditions. Marine ecosystem variability has been related to PDO variability and the associated Aleutian low, but only for ecosystems in the Pacific Ocean, like off the coasts of Japan, California, Alaska, and

Peru (Mantua et al. 1997; Minobe 1997; Mantua and Hare 2002). Due to PDO persistence and the varied associations, PDO-related forecasts have the opportunity to be skillful.

However, PDO-related forecasts will become more confident when the physical processes of PDO are better understood (Mantua and Hare 2002).

2.4 ENSO and PDO Relationships

Despite having slight differences in characteristics and impacts, ENSO and PDO are both related to SST variability in the Pacific Ocean. Mantua et al. (1997) believes

ENSO and PDO climate patterns to be clearly related, both spatially and temporally, hence the use of the term “ENSO-like” to describe PDO variability. However, the question of how ENSO and PDO are related is still uncertain. It would be easy to

describe PDO-related climate shifts as responses to ENSO events, but it is equally conceivable that the long-term PDO variability constrains the envelope of the shorter

ENSO variability (Mantua et al. 1997).

2.4.1 PDO Dependence on ENSO

There are a few studies in atmospheric science literature that agree on the idea of

PDO variability being forced by ENSO (Newman et al. 2003; Zhang and Delworth

2015). Using Pacific Ocean SSTs and atmospheric variability for the years 1900 to 2001,

Newman et al. (2003) modeled PDO variability as the sum of direct forcing by ENSO, the reemergence of North Pacific SST anomalies in subsequent winters, and white noise atmospheric forcing. As a result, Newman et al. (2003) created a possible null hypothesis for PDO, stating that variability in North Pacific SSTs on decadal timescales results from

“reddening” of the ENSO signal. Specifically, Newman et al. (2003) concludes that PDO is the reddened response to both atmospheric noise and ENSO, while being dependent upon ENSO at all timescales.

In agreement with Minobe (2000) and Mantua and Hare (2002), Newman et al.

(2003) explains that decadal variability of North Pacific SSTs is largely a phenomenon of winter and spring seasons. The cycle of ENSO and PDO is suggested as ENSO forcing

PDO in summer, the subsequent ENSO phase forces PDO the following winter and spring, and the resultant North Pacific SST anomaly persists into the early part of the following summer (Newman et al. 2003). A similar analysis of ENSO and PDO relations has been completed, with Zhang and Delworth (2015) also suggesting that ENSO

variability plays a role in PDO dynamics because of PDO-related SST anomalies in the tropical Pacific. As for predicting the phase of PDO, Newman et al. (2003) and Zhang and Delworth (2015) agree that it is directly related to the skill of forecasting ENSO.

Since ENSO can only be predicted a couple years into the future, PDO shifts may not be apparent until a couple years after they occur (Newman et al. 2003).

2.4.2 ENSO Modulated by PDO

While the work of Newman et al. (2003) and Zhang and Delworth (2015) is notable, several studies have also supplied evidence of the opposite ENSO and PDO relationship. Mokhov et al. (2004) believes there to be two approaches to examining

ENSO tendencies. Usually, ENSO events are diagnosed when indices deviate from the annually varying basic state, remaining unchanged for a certain period of time. The second approach allows the basic state of ENSO to evolve over time, prohibiting permanent states of El Niño or La Niña. For the second approach, ENSO can be modulated by longer term variability, which could be caused by anthropogenic forcing

(human induced, involving greenhouse gases) or by natural long-term variations in climate (PDO variability), according to Mokhov et al. (2004).

The first study to address this possible PDO modulation of ENSO came from

Gershunov and Barnett (1998). By examining SLP anomalies in the Pacific Ocean, it was found that SLP patterns associated with ENSO tend to be stronger and more stable during constructive ENSO and PDO phase pairings. The El Niño pattern in Pacific SLP consists of an Aleutian low with low pressure anomalies stretching and diminishing toward the

southeast along the coast of North America. However, during years of a positive PDO, El

Niño SLP signals are much more distinct. The Aleutian low and the low pressures along the American coast become almost twice as strong as those associated with El Niño alone. For years of a negative PDO, El Niño signals become insignificant with no strong

SLP anomalies in the eastern North Pacific. La Niña is associated with a SLP pattern near opposite of El Niño, with high pressure in the North Pacific. Gershunov and Barnett

(1998) determined the La Niña signal to strengthen during negative PDO years, along with the emergence of a significant low SLP anomaly over the central United States.

Contrary to El Niño, La Niña SLP patterns become insignificant and distorted during positive PDO years.

With a constructive ENSO and PDO phase relationship determined from SLP anomalies, it is likely that this association between phenomena has an impact on North

American climate. Gershunov and Barnett (1998) expected this PDO modulation of

ENSO to create a stronger climate response to El Niño during positive PDO years and La

Niña during negative PDO years, with the opposite phase pairings leading to weaker climate responses. This was found to be true with the examination of United States precipitation during these phase combination years. For example, the La Niña pattern in precipitation was found to correlate to the negative PDO pattern, with a correlation coefficient of 0.86 and sharing 74 percent of spatial variability (Gershunov and Barnett

1998). However, these statistics weaken when La Niña patterns are compared to positive

PDO patterns, with only 14 percent of spatial variability shared and a pattern correlation coefficient of only 0.38. Therefore, PDO has an influence on United States ENSO impacts and should be considered when creating ENSO-based statistical forecasts in

North America, especially in western and central United States (Gershunov and Barnett

1998).

One year after the Gershunov and Barnett (1998) study came the Power et al.

(1999) study. Using PDO index values derived from Zhang et al. (1997), as well as an

EOF analysis of global SST data, Power et al. (1999) determined how PDO phase influences the ENSO impact on Australian rainfall. Specifically, the examination of the frequency distribution of rainfall prediction skill scores determined the negative phase of

PDO to be most important. The negative PDO phase allowed for significantly enhanced seasonal predictability of rainfall anomalies, which are based on a linear, lagged relationship with the Southern Oscillation Index (SOI). These results suggest that the

ENSO impact on Australian rainfall varies in association with PDO.

From average temperature and precipitation over the entire continent of Australia,

Power et al. (1999) provided evidence of how PDO can modulate ENSO impacts through correlation coefficients. Results found a significant correlation of 0.7 when comparing variations in precipitation to both the SOI and a negative PDO index. That correlation coefficient fell to 0.1 when precipitation was compared to SOI and a positive PDO index.

Power et al. (1999) explains that when PDO is positive, the tropical Pacific waters are very warm, but flanked by cold water to the north and south. When SSTs increase in the tropical Pacific Ocean, there is no significant relationship between Australian climate variations and ENSO, and prediction skill scores decline. However, when tropical Pacific

SSTs decrease during a negative PDO phase, the association between ENSO and

Australian climate are strong and predictions are much more accurate (Power et al. 1999).

Thus, the ENSO phase impacts on Australian climate are influenced by the phase of

PDO.

With two studies (Gershunov and Barnett 1998; Power et al. 1999) discovering the PDO modulation of ENSO, more research has been focused on this phenomenon and how it can improve climate predictability. Goodrich (2004) investigated the influence of

PDO on the ENSO impacts involving Arizona winter precipitation, from January to

March. ENSO has been correlated to winter precipitation in the southwestern United

States via SSTs in the central tropical Pacific impacting the path of mid-latitude cyclones

(Goodrich 2004). Specifically, ENSO is associated with a southward shift in the mid- latitude cyclone track during El Niño events, with a northward shift in the track during La

Niña events. Therefore, as previously mentioned (section 2.2.2), El Niño is associated with above average winter precipitation over the Southwest, while La Niña is associated with prolonged drought and below average precipitation.

Goodrich (2004) hypothesized, from the results found in Gershunov and Barnett

(1998), that the Southwest should experience much wetter winters when El Niño events occur during a positive PDO. Since La Niña events were found to be stronger during negative PDO phases (Gershunov and Barnett 1998), Goodrich (2004) expected the

Southwest to be much drier during the years of this phase combination. These predictions were found to be true when Goodrich (2004) used a difference of means test on Arizona winter precipitation, as binned by each ENSO and PDO phase combination. Positive

(negative) PDO phases enhanced the El Niño (La Niña) impact, while negative (positive)

PDO phases kept Arizona winter precipitation near normal during El Niño (La Niña) events. Goodrich (2004) speculates the reasoning to be because PDO phase can also alter

mid-latitude cyclone tracks, with the strong (weak) Aleutian low associated with a positive (negative) PDO phase causing the track to shift south (north).

North American climate is most impacted by the positive PDO phase combined with El Niño events (Gershunov and Barnett 1998; Goodrich 2004), while Australia is most impacted by ENSO during the negative PDO phase (Power et al. 1999). Wang et al.

(2008) wanted to determine which PDO phase would have the greatest influence on the relationship between ENSO and the east Asian winter monsoon. Using similar methodology as Gershunov and Barnett (1998), winter precipitation was categorized and examined by ENSO and PDO phase combinations. Results found the relationship between ENSO and the east Asian winter monsoon to be weak and insignificant during positive PDO years (Wang et al. 2008). The opposite was found to be true for the negative PDO years, with ENSO having a strong impact on the winter monsoon and significant changes in low-level temperature occurring over east Asia. Therefore, the

ENSO relationship with the winter monsoon depends on the PDO phase, with the negative PDO phase having the greatest influence. Wang et al. (2008) concludes by stating that the PDO phase should be taken into account when ENSO is used as a predictor for the east Asian winter monsoon.

2.4.3 PDO Modulated ENSO in the Midwest

As for the Midwestern United States, a few studies have come across the possible existence of PDO modulated ENSO impacts on climate variables (Berger et al. 2002;

Lupo et al. 2007, 2008; Birk et al. 2010). Berger et al. (2002) were the first to discover

the influence of PDO on ENSO impacts in Missouri, after examining the climatology of snowfall events. The focus was initially on the ENSO impact, but after further review,

Berger et al. (2002) found the ENSO-related variability in snowfall occurrences to be superimposed on the long-term PDO variability. Results confirm the research by

Gershunov and Barnett (1998) and Goodrich (2004), with significantly fewer snowfall events occurring during years of El Niño and positive PDO, compared to that of only El

Niño years. Overall, Berger et al. (2002) found there to be a significant difference in

ENSO-related variability for each phase of PDO. Lupo et al. (2005) found the same results when examining Southwest Missouri snowfall. Like the aforementioned studies,

Berger et al. (2002) and Lupo et al. (2005) tribute ENSO and PDO impacted mid-latitude cyclone tracks for the examined snowfall occurrences.

Lupo et al. (2007, 2008) and Birk et al. (2010) found similar results when associating ENSO and PDO to Midwestern temperature and precipitation. Lupo et al.

(2007, 2008) used the Kung and Chern (1995) SST anomaly classifications to describe

ENSO, as well as PDO phases. The ENSO-specific synoptic classifications (A-G) were described in section 2.2.1. PDO phase can be described with classifications of type A through D for negative phase and types C, E, and F for positive phase. Comparing these synoptic classifications to temperature and precipitation displayed cooler than normal summer months and warmer than normal winter months for type F, which is associated with strong El Niño events and positive PDO years. Refer to Lupo et al. (2007, 2008) for each type’s specific associations to Midwestern temperature and precipitation. Further analysis displayed a dominance of El Niño events, with the majority of those being strong, during positive PDO years (Lupo et al. 2007, 2008).

Birk et al. (2010) focused solely on the PDO modulation of ENSO-related interannual variability of Midwestern climates. Using power spectrum analysis, it was determined that an interdecadal variability on the order of twelve to fifteen years exists in

Midwestern temperature and precipitation. Birk et al. (2010) considers this interdecadal variability to represent the time periods in which PDO acts to modulate the ENSO-related variability. Birk et al. (2010) performed a statistical analysis on monthly average temperature and precipitation values for the Midwest, in order to quantify interannual variability and to identify the relationships between ENSO and PDO variability.

Consistent with Berger et al. (2002), ENSO-related departures and variability were found to be stronger during positive PDO periods and lower during negative PDO periods. The most significant ENSO-related temperature variability during a positive PDO period occurred in the winter season months through the northern Midwest (Birk et al. 2010).

Monthly temperature frequencies during El Niño and negative PDO events differ from that of El Niño and positive PDO events, with the former associated with a dramatic increase in the percentage of cool and normal months, and the latter associated with an increase in warm months (Birk et al. 2010). El Niño events during positive PDO years were also found to produce above average spring and fall precipitation for the southern and western regions of the Midwest. However, the same phase combination produced drier conditions for the central, northern, and eastern regions during spring and summer.

La Niña years were found to be warmer than average during both phases of PDO. La

Niña events during positive PDO years experienced an absence of extreme events, with mostly average conditions overall (Birk et al. 2010). La Niña events during negative PDO years produced below average precipitation in the Midwest, with the exception of the

southeastern region. Similar to Lupo et al. (2007, 2008), Birk et al. (2010) compared the synoptic patterns of each SST distribution for each ENSO and PDO phase, in hopes of determining the physical reasoning of the final results.

In conclusion, strong evidence suggests that the behavior of ENSO events is determined by PDO phase (Birk et al. 2010). El Niño (La Niña) impacts on Midwestern climate are enhanced by a positive (negative) PDO phase, making the expected climatic conditions associated with a particular ENSO phase differ for each event. Thus, PDO modulates the ENSO impacts on climate variables and should be taken into consideration when creating long-range forecasts (Gershunov and Barnett 1998; Berger et al. 2002;

Goodrich 2004; Lupo et al. 2007, 2008; Birk et al. 2010). Specifically, the interdecadal variability associated with the modulation should be considered as being superimposed upon the interannual variations, in order to obtain more skillful seasonal to long-range forecasts of Midwestern variables (Berger et al. 2002; Lupo et al. 2007, 2008; Birk et al.

2010). To stay consistent with the Midwestern climate studies aforementioned, this thesis accepts the theory of PDO modulated ENSO-related variability on an interdecadal timescale.

2.5 Climate Impacts on Crops

It is known that ENSO and PDO-related climate variability can impact temperature and precipitation in the United States and Australia. Temperature and precipitation greatly impact agricultural performance, meaning it is possible for ENSO

and PDO to have an impact on crop production. If so, then ENSO and PDO can be used as a predictor for not only temperature and precipitation, but also crop yields. These possible connections led Power et al. (1999) to include wheat crop yield in the list of

Australian climate variables used to determine the impacts of ENSO and PDO. Results displayed the same impacts on wheat yields as those described for rainfall in section

2.4.2. PDO modulates the ENSO impact, with a stronger influence during the negative

PDO phase, on the domestic wheat crop of Australia (Power et al. 1999). From these results, Power et al. (1999) enhanced seasonal predictability of wheat crop yields in

Australia, with the highest forecast skill scores occurring when PDO was negative.

More recently, Ding and McCarl (2014) studied the impact of PDO and other decadal climate variabilities on several different crops in the Texas region of Edwards

Aquifer. The crops studied include corn, cotton, oats, sorghum, and winter wheat. From the results, it was determined that a certain phase combination between PDO and a few other decadal climate variabilities contributed to a decrease of 5-15% in all of the crop yields except irrigated sorghum (Ding and McCarl 2014). A different phase combination was associated with an increase in all of the crop yields (Ding and McCarl 2014).

Further, Ding and McCarl (2014) explain how climate variability can impact agricultural economics by increasing or decreasing the crop value.

There is a noticeable lack of knowledge on how climate variability can impact crop yields in the United States. For Missouri, specifically, no study has been completed on the ENSO and PDO impacts on crops. One study comes close by examining the climate effects on Missouri corn yields, but only temperature and precipitation data were used (Hu and Buyanovsky 2003).

Hu and Buyanovsky (2003) began research by classifying corn yield production by low, average, and high yield years. Average monthly temperature and precipitation for the corn growing season were compared to each group of low, average, and high yields. It was found that high yield years are characterized by lower rainfall in April and

September and higher rainfall from May through August, with the most significant above-average precipitation occurring in July. The temperature during high yield years was found to be above average from April through June and below average in July and

August (Hu and Buyanovsky 2003). Overall, high corn yields in Missouri are most likely produced when the spring months of April and May, which covers the planting period of corn, are warmer and drier, with the summer months of July and August being wetter and cooler. It was also noted that the summer months of July and August had stable or similar daytime temperatures during high yield years (Hu and Buyanovsky 2003). The opposite conditions, including large daily fluctuations of temperature in the summer months, were associated with low corn yields.

Hu and Buyanovsky (2003) successfully determined the climate impacts on

Missouri corn production by examining temperature and precipitation over the corn growing season. Therefore, it is conceivable that climate variability related to ENSO and

PDO can impact Missouri crop production, and possibly be used as a predictor of crop yields, because ENSO and PDO are related to and used to predict Missouri temperature and precipitation (Lupo et al. 2007, 2008; Birk et al. 2010). From the information learned in all of the aforementioned studies, it seems highly possible to relate ENSO and PDO- related climate variabilities to crop production. Thus, ENSO and PDO phases can

possibly be used to predict regional Missouri corn and soybean yields. This thesis is the first to research the specific impacts of ENSO and PDO on Missouri crop yields.

CHAPTER 3. DATA AND METHODOLOGY

3.1 Data Sources

Crop data were recorded from the United States Department of Agriculture’s

(USDA) National Agricultural Statistics Service (NASS) using Quick Stats 2.0

(quickstats.nass.usda.gov). Crop yields were collected for the two most common field crops grown statewide in Missouri: corn and soybean. Corn was recorded as grain yield in bushels per acre, while soybean was recorded as a general yield in bushels per acre.

Annual yields for each crop were downloaded by county in Missouri based on survey.

From the collected data, an average yield representing each crop was calculated for each of the six climate regions in Missouri based on the counties located in each region.

The six climate regions of Missouri (Fig. 3.1) are determined by the Climate

Divisional Dataset, which is a long-term temporally and spatially complete dataset used by the National Oceanic and Atmospheric Administration (NOAA) to generate historical climate analyses for the continental United States. Missouri experiences regional differences in climate (NCDC 2006), so analyzing data by the climate regions will give the opportunity for more accurate results. A total of 95 years of data, from 1919 to 2013, were used for corn in each region, while only 70 years of soybean data were collected from 1944 to 2013.

Temperature and precipitation data were obtained from NOAA, via the

Midwestern Regional Climate Center (MRCC) cli-MATE application

(mrcc.isws.illinois.edu/CLIMATE). Temperature was recorded in degrees Fahrenheit and precipitation in inches. Monthly averages for temperature and precipitation were downloaded directly for each climate region in Missouri, using the MRCC Climate

Division Data. The number of years downloaded for temperature and precipitation depended on the number of years of data collected for each crop in each region. Thus, monthly average temperatures for each region were collected for the years 1919 to 2013 to match with corn, while the same was collected for the years 1944 to 2013 to match with soybean yields.

Available crop data ranging from 70 to 95 years, for soybean and corn yields respectively, represent sufficiently long time periods for identifying interannual, interdecadal, and multidecadal variabilities. The climate variability this study is concerned with is related to the El Niño Southern Oscillation (ENSO) and the Pacific

Decadal Oscillation (PDO). ENSO varies every two to seven years (Kung and Chern

1995; Enfield and Mestas-Nuñez 1999; Lupo et al. 2005), while PDO fluctuates roughly every 20 years (Lupo et al. 2007; Ding and McCarl 2014).

Please note, corn yields in region 1 represent the longest timescale and largest yield values of all data collected. Therefore, only graphs resulting from corn yields for region 1 are displayed in this publication to help describe methods and analyses. The same methodology was used for both crops in all six regions, resulting in similar graphs as those of corn for region 1. The graphs for the remaining crops and regions are displayed in Appendix A for reference. The analysis results described in section 4.1 include both crops in all six regions.

3.2 Methods

Agricultural techniques and equipment have advanced over the years causing an increasing trend in crop yields (Hu and Buyanovsky 2003). To account for the influence of these advancements, including seed genetics and the use of fertilizers, the trend was eliminated from the yield data for both crops. Using minimized squared error techniques, a simple linear trend in the data was defined and then removed. Removing the trend in

a) Missouri Corn Yields - Region 1 200

150

100

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 1, Detrended 60

 30

Yield(Bu/Acre)  60 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

the data allows for the opportunity to evaluate the climate effect on crop yields most accurately. The crop yield data were plotted as a time series for each crop in each region; first as annual yields (Fig. 3.2a), and then as annual detrended yields (Fig. 3.2b). Figures

3.2a and 3.2b represent corn yields in bushels per acre for region 1. The same graphs resulting from each crop in each region contain similar results and linear trends as corn for region 1 (Appendix A). As stated in section 3.1, only figures from corn for region 1 are displayed to explain the methodology used in this study.

Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Corn-Temp-Precip 0.4965 0.5480 0.6069 0.5753 0.4992 0.3005 (July)

Corn-Temp-Precip 0.2783 0.3707 0.3886 0.3800 0.3318 0.1564 (Season)

Soybean-Temp-Precip 0.5345 0.5848 0.6016 0.5488 0.4785 0.5026 (August)

Soybean-Temp-Precip 0.1803 0.2957 0.4597 0.4436 0.4124 0.3704 (Season)

Hu and Buyanovsky (2003) determined there to be a correlation between Missouri corn yields and temperature and precipitation. To verify this correlation, a multiple correlation function was used to calculate the correlation coefficient between detrended crop yields and temperature and precipitation data. Detrended crop yields were the dependent variable, while temperature and precipitation were independent variables, in determining the Pearson correlation coefficient from a linear regression model. The results using seasonal temperature and precipitation, as well as the highest correlating monthly temperature and precipitation, for each crop, are displayed in Table 3.1, with a value of zero representing no correlation between the variables. Seasonal temperature and precipitation values were calculated by averaging the values from the monthly data, for the crop growing season of April through September, downloaded from the Midwestern

Regional Climate Center (MRCC).

3.2.1 Power Spectrum Analysis

In order to identify interannual and multidecadal periodicities within each time series of detrended crop yields, the method of power spectra analysis following the

Mokhov et al. (2004) “method of cycles” was performed (e.g., Birk et al. 2010; Lupo et al. 2012a). This method of amplitude and period calculation results in phase portraits

Ẋ(푋) for a given time series 푋(푡). The process can be fitted by a harmonic oscillator:

Ẍ + 휔2푋 = 0 (3.1) or

푋(푡) = 퐴(푡) sin 휔(푡)푡 + 휙(푡) (3.2) if there is a statistically significant linear regression of Ẍ(푡) on 푋(푡) with a negative regression coefficient −휔2(푡). The amplitude of the process 퐴(푡), frequency 휔(푡), period 푃(푡) = 2휋/휔(푡), and initial phase 휙(푡), are assumed to change sufficiently slowly over time (Mokhov et al. 2004). The variables Ẋ and Ẍ are determined by taking the second-order finite difference of the original time series 푋. In this study, temperature, precipitation, and crop yield data are represented by 푋. The amplitude 퐴(푡), frequency

휔(푡), and period 푃(푡) are calculated using a least-squares method (e.g., Mokhov et al.

2004). Equations (3.1) and (3.2) can be solved using empirical orthogonal functions (e.g.,

Fourier series or wavelet transforms). Fourier analysis was chosen for this study.

Power spectra were constructed from the detrended crop data using Fourier coefficients. Fourier transforms convert data from Cartesian space to wave space and power is put into discrete wave numbers; resulting in a plot of wave power versus wave number that can then be analyzed for dominant periods in a time series. Spectral peaks

represent periodicities in the time series and can be tested for statistical significance against a red or white noise continuum (e.g., Wilks 2006), due to the expectation that low frequency would be dominant (red) or that all frequencies would be equally likely

(white). Resulting power spectra for both crops in all six regions were analyzed for statistically significant periodicities to determine if the resulting oscillations relate to

ENSO and PDO variability.

The same method was used to create power spectra for seasonal (April through

September), as well as monthly (July or August), temperature and precipitation data for each climate region and time range, corresponding to the crop data. Resulting power spectra were not analyzed for periodicities, but rather used to perform a cross-spectral analysis (e.g., Wilks 2006; Lupo et al. 2012b). This analysis involves the convolution of power spectra to create a final power spectrum used for the examination of periodicities.

Power spectra resulting from detrended crop data were cross-analyzed with power spectra resulting from seasonal or monthly temperature and precipitation, for the respective time range and climate region. The resulting covariance spectra were analyzed for dominant periods, which represent the periodicities shared by crop yields, temperature, and precipitation. The peaks found in the resulting spectra were tested for statistical significance against a white noise continuum, assuming no particular frequency to be dominant (e.g., Wilks 2006).

Each crop in each region is therefore represented by three different power spectra: detrended annual crop yields, covariance of detrended yields with seasonal temperature and precipitation, and covariance of detrended yields with July or August temperature and precipitation. Comparing periodicities from all three resulting spectra for each crop

in each region will help provide a more accurate representation of relationships to ENSO and PDO. Since ENSO and PDO correlate to Midwestern temperatures and precipitation

(Berger et al. 2002; Birk et al. 2010), variability resulting from a covariance of Missouri crop yields with temperature and precipitation data is more likely to be related to ENSO or PDO, as compared to variability found in crop yields alone.

3.2.2. PDO Influence on ENSO

In an effort to support the idea of an interaction harmonic between ENSO and

PDO, power spectra were created using the Mokhov et al. (2004) “method of cycles” for annual SOI data and annual mean values of the PDO index. A cross-spectral analysis, following the covariance technique found in Lupo et al. (2012b), was then performed and the resultant spectrum was analyzed for periodicities, in the search for an interdecadal variability. A second method was used to verify the existence of PDO modulated ENSO events, involving detrended crop data. Annual detrended crop yields for each crop in each region were binned according to ENSO phase, as well as to each ENSO-PDO phase combination (Table 3.2). Averages were then calculated, resulting in an average crop yield for each crop in each region for each ENSO phase and each ENSO-PDO phase combination.

Detrended crop yield averages are displayed as a positive or negative value, making it easy to see the impacts of each ENSO phase or ENSO-PDO phase combination on each crop and region in terms of productivity. Comparing crop yield averages from each ENSO phase to crop yield averages from each ENSO-PDO phase combination gave

Table 3.2. The years examined in this study, corresponding to each El Niño Southern Oscillation (ENSO) phase, as well as each Pacific Decadal Oscillation (PDO) phase.

Positive PDO (1) Negative PDO (2) La Niña Neutral El Niño La Niña Neutral El Niño 1938 1926-1928 1925 1922 1919-1921 1951 1942 1931-1937 1929 1924 1923 1957 1944 1939 1930 1949 1947 1963 1988 1941 1940 1954-1956 1948 1965 1998 1943 1982 1964 1950 1969 1945 1986 1967 1952 1972 1946 1987 1970 1953 1976 1977-1981 1991 1971 1958-1962 2002 1983-1985 1997 1973-1975 1966 2006 1989 1999 1968 2009 1990 2007 2000 1992-1996 2010 2001 2003-2005 2008 2011-2013

the opportunity to assess how PDO phase influences ENSO impacts. A Mann-Whitney test was completed on the phase-specific detrended crop yields to determine the statistical significance of the calculated averages, using one-tail probability. Only data samples containing at least 5 values and resulting in a p-value of 0.10 or less (90% confidence level or greater) were considered to be statistically significant.

CHAPTER 4. RESULTS

4.1 Climatological Analysis

Figure 3.2a represents one of twelve graphs created from the crop yield data collected from USDA NASS for each of the crops (corn and soybean) for all six climate regions in Missouri (Fig. 3.1). As stated in section 3.2, Fig. 3.2a specifically represents annual corn yields in bushels per acre from 1919 to 2013 for region 1. The linear trend was then removed from the data in Fig. 3.2a, to create the data shown in Fig. 3.2b.

Detrended crop yields for both crops in all six regions were then used for the remaining methods in this study.

Figure 4.1a represents the power spectrum resulting from the Fourier transform of detrended corn yields for region 1 (Fig. 3.2b). This spectrum was analyzed for periodicities, represented by peaks, to determine a correlation to climate variability in the form of ENSO or PDO. To determine the variability, or periodicities, in the detrended corn yields, the time interval of the data used is divided by the wave number of a statistically significant spectral peak, against both the red and white noise continuum. For example, Fig. 4.1a displays a large peak on the right half of the graph at a wave number of roughly 20. The time interval of the data used in Fig. 4.1a is 1919 to 2013 for corn in region 1, which is 95 years of data. Therefore, the periodicity represented by the large peak is roughly 5, meaning corn yields in region 1 vary significantly every 5 years. With

5 falling into the interannual variability range of two to seven years, this periodicity represents a possible relationship to ENSO phase. The same calculation was made for the

a) b) Corn Yields - Region 1 Seasonal Temperature - 1 3 1.610 20

3 1.210 15

800 10

Magnitude

Magnitude 400 5

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

c) d) Seasonal Precipitation - 1 Temp/Precip/Corn 1 3 5 410

3 3.75 310

3 2.5 210

Magnitude

Magnitude 3 1.25 110

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

remaining significant peaks shown in Fig. 4.1a, resulting in corn yields for region 1 having interannual, interdecadal, and multidecadal variability. Thus, corn yields in region

1 have a possible relationship to ENSO, PDO modulated ENSO, and PDO phase.

Figure 4.1b represents the same methodology used to create Fig. 4.1a, except seasonal temperature data were used for climate region 1. The temperatures were averaged for the corn growing season, which spans from April to September, and the years analyzed corresponded to the available corn data of 1919 through 2013. Figure 4.1c is the same as Fig. 4.1b, except seasonal precipitation data were used. Figure 4.1d displays the power spectrum representing the convolution of detrended corn yields (Fig.

4.1a) with seasonal temperatures (Fig. 4.1b) and precipitation (Fig. 4.1c) for region 1 over the period of 1919 to 2013 (corresponding to the time range of corn yields used for region 1). After dividing the time interval (again 95 years) by the wave number of each statistically significant peak, the covariance of corn, temperature, and precipitation displayed interannual and interdecadal variability. In other words, corn, temperature, and precipitation data from region 1 all vary at the same intervals as ENSO and PDO modulated ENSO.

The same method was completed using detrended corn yields and average July temperature and precipitation data for all six regions, based on the correlation results found in Table 3.1. Detrended corn yields for all six regions displayed the highest correlation coefficient to temperature and precipitation in July (Table 3.1), compared to that of any other growing season (April through September) month and to all growing season months combined, referred to as the seasonal temperature and precipitation.

Soybean yields displayed the highest correlation coefficient to August temperature and precipitation (Table 3.1).

The methodology used for all spectra in Fig. 4.1 were used for all power spectra involving both crops in all six regions with their corresponding time ranges of seasonal,

Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 ENSO ENSO ENSO ENSO ENSO ENSO Corn ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO PDO PDO Corn-Temp-Precip ENSO ENSO ENSO ENSO ENSO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO Covariance (July) PDO PDO PDO Corn-Temp-Precip ENSO ENSO ENSO ENSO ENSO ENSO ENSO-PDO ENSO-PDO Covariance (Season) PDO ENSO ENSO ENSO ENSO ENSO ENSO Soybean ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO PDO PDO Soybean-Temp-Precip ENSO ENSO ENSO ENSO ENSO ENSO ENSO-PDO ENSO-PDO ENSO-PDO ENSO-PDO Covariance (August) Soybean-Temp-Precip ENSO ENSO ENSO ENSO ENSO ENSO ENSO-PDO ENSO-PDO ENSO-PDO Covariance (Season) PDO PDO PDO PDO PDO

as well as highest correlating monthly, temperature and precipitation data. Three spectra were analyzed for each crop in each region, resulting in 36 analyzed spectra. Each periodicity resulting from the analyzed spectra was associated with ENSO, ENSO-PDO

(PDO modulated ENSO), and PDO, depending on the variability range of interannual (2-

7 years), interdecadal (10-15 years), or multidecadal (20 years), respectively. The final results are shown in Table 4.1 according to possible climate variability associations.

Interannual variations were found in 35 out of 36 (97%) analyzed spectra, while only 27 of the 36 (75%) contained interdecadal variabilities and 13 of the 36 (36%) contained multidecadal variabilities.

4.2 ENSO and PDO Phase Interactions

With interdecadal variability found in 75% of the analyzed crop spectra, it is important to understand the cause before an attempt at forecasting crop yields can be made. Further analyses were completed in an effort to support the theory of PDO modulated ENSO-related variability (Gershunov and Barnett 1998; Berger et al. 2002;

Birk et al. 2010). The first step was to examine ENSO and PDO for an interdecadal variability. In order to do this, the “method of cycles” was used to create power spectra representing annual SOI data (Fig. 4.2a), to determine variations in ENSO, and annual mean values of the PDO index (Fig. 4.2b). Figures 4.2a and 4.2b verify ENSO and PDO variability to be interannual (2-7 years) and multidecadal (roughly 20 years, as well as roughly 50 years), respectively. The convolution of annual SOI data and annual mean values of the PDO index is represented by the power spectrum in Fig. 4.2c. The cross- spectral analysis revealed a significant peak in Fig. 4.2c at a wave number of about 10.

After division of the time interval of 104 years (1901-2004) by the wave number of 10, the periodicity is about 10 years. Thus, the covariance of ENSO and PDO displays an interdecadal variability.

The second step was to determine if an interaction harmonic between ENSO and

PDO phases can be found in crop yields. Table 4.2 displays the averages calculated for each set of detrended yields, in bushels per acre, for each crop in each region based on

ENSO phase year (Table 2.1). Corn and soybean yields display a positive value for El

Niño years, meaning yields will most likely depart positively from the yield average in future El Niño years. La Niña years display mostly negative values for corn and soybean yields based on region, meaning yields will likely depart negatively from average when

a) b) Annual SOI Annual PDO 300 5

225 3.75

150 2.5

Magnitude Magnitude 75 1.25

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

c) SOI/PDO 180

135

Magnitude 45

0 0 6 12 18 24 30

Wave Number

La Niña events occur. Statistical significance testing of yields in El Niño years versus yields in La Niña years determined regions 1, 2, and 6 to be statistically significant at the

90% confidence level for both crops, with region 4 for corn also included (Table 4.2).

Regions 1, 2, 4, and 6 contribute to 70% of Missouri corn production, while regions 1, 2,

La Niña Neutral El Niño Corn 1 -5.267 0.476 4.445 Corn 2 -4.199 0.209 4.035 Corn 3 -3.063 0.732 1.268 Corn 4 -3.025 -0.172 3.843 Corn 5 -3.748 0.808 1.805 Corn 6 -5.732 1.531 1.905

Soybean 1 -0.883 -0.237 1.600 Soybean 2 -0.766 -0.113 1.154 Soybean 3 -0.477 0.057 0.396 Soybean 4 0.314 -0.429 0.731 Soybean 5 -0.763 0.175 0.421 Soybean 6 -1.751 0.678 0.268

and 6 contribute to roughly 55% of Missouri soybean production, for the range of years analyzed in this study.

Table 4.3 is similar to Table 4.2, only displaying average detrended crop yields based on ENSO-PDO phase combination years (Table 3.2). El Niño and La Niña show the same impacts on the crops, but the impacts are enhanced when combined with a positive PDO phase and diminished when combined with a negative PDO phase. For example, detrended corn yields in climate region 1 have an increase in the average yield by 4.4 bushels per acre during El Niño years (Table 4.2). The average yield increases by

7.1 bushels per acre for El Niño and positive PDO years, and increases by only 2.1 bushels per acre for El Niño and negative PDO years (Table 4.3). Corn in region 6, while

Table 4.3. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation-Pacific Decadal Oscillation (ENSO-PDO) phase combination, listed in Table 3.2. Bold values represent statistically significant (90% confidence level) differences in yield for El Niño years of different PDO phases, with both data sets involved containing at least 5 samples (yield per phase year combination).

Positive PDO (1) Negative PDO (2) La Niña Neutral El Niño La Niña Neutral El Niño Corn 1 -5.737 -2.183 7.056 -5.120 3.665 2.095 Corn 2 -7.891 -1.479 7.250 -3.045 2.235 1.141 Corn 3 -0.799 0.356 3.398 -3.771 1.183 -0.650 Corn 4 -2.800 -0.559 6.206 -3.096 0.292 1.716 Corn 5 -4.935 -0.019 2.193 -3.378 1.800 1.456 Corn 6 -9.131 2.059 6.834 -4.669 0.897 -2.532

Soybean 1 -2.011 -0.499 1.375 -0.641 -0.024 1.713 Soybean 2 -1.591 -0.213 1.500 -0.589 -0.032 0.982 Soybean 3 1.173 0.592 2.592 -0.831 -0.376 -0.702 Soybean 4 0.211 -1.194 1.346 0.336 0.191 0.423 Soybean 5 -1.239 -0.151 1.361 -0.661 0.439 -0.048 Soybean 6 -2.167 -0.230 -0.842 -1.662 1.412 0.823

soybean in regions 3 and 6, resulted in statistically significant differences in yields at the

90% confidence level when comparing yields from El Niño/Positive PDO years to yields from El Niño/Negative PDO years (Table 4.3).

A regional analysis of the results displayed in Tables 4.2 and 4.3 revealed which regions have the potential for the most accurate crop yield forecasts based on ENSO and/or PDO phase. The statistical significance tests showed promise for five out of six of the climate regions in Missouri. Regions 1, 2, 4, and 6 displayed statistically significant results for corn yields, while regions 1, 2, 3, and 6 displayed the same for soybean yields.

Figure 4.3 displays these findings on a map of the Missouri climate regions with the corresponding ENSO and/or PDO phase relationships. Region 6 shows the most

Corn: El Niño/ La Niña

Soybean: El Niño/ Corn: El Niño/ La Niña La Niña

Soybean: El Niño/ La Niña Corn: El Niño/La Niña PDO/El Niño Soybean: PDO/El Niño Soybean: El Niño/La Niña PDO/El Niño

Corn: El Niño/La Niña

Figure 4.3. A map of Missouri outlining the six climate regions, defined by NOAA (Fig. 3.1), including the statistically significant results found for both crops. El Niño/La Niña refers to the statistically significant difference in yields between El Niño years and La Niña years, found in Table 4.2. PDO/El Niño refers to the significant difference in yields between El Niño/Positive PDO years and El Niño/Negative PDO years, found in Table 4.3.

statistically significant results, with both corn and soybean yields having statistically significant differences in El Niño years compared to La Niña years, as well as in El

Niño/Positive PDO years compared to El Niño/Negative PDO years. Regions 1, 2, and 6 display significant results for both corn and soybean. Region 3 only shows significance with soybean yields, while region 4 only displays significant results for corn. Out of all six climate regions, only region 5 lacks statistically significant results in both analyses

(Tables 4.2 and 4.3).

CHAPTER 5. DISCUSSION

Annual crop yields in Missouri were examined for periodicities, using the techniques of Mokhov et al. (2004) and Lupo et al. (2012b), to determine possible associations with the climate variabilities related to ENSO and PDO. The highly produced field crops of corn and soybean were studied by climate region in Missouri, in order to account for climate differences across the state (NCDC 2006). Average seasonal temperature and precipitation ranging from April to September, to cover Missouri’s corn and soybean growing season, as well as the highest correlated monthly temperature and precipitation data were used for the analysis of each crop. July for corn and August for soybean proved to have the highest correlation between temperature, precipitation, and crop yield when compared to other months and the growing season average. Therefore,

July (August) temperature and precipitation will have the greatest impact on corn

(soybean) production in Missouri. Spectral analysis of each crop in each region displayed different variabilities of interannual, interdecadal, and/or multidecadal timescales in each set of data, suggesting slightly varied ENSO and PDO impacts across Missouri. Regional differences in ENSO and PDO impacts are not extreme, possibly due to the fact that

Missouri’s regional climates grade inconspicuously into each other (NCDC 2006).

With results varying for each crop in each region, including each power spectrum analyzed, the best way to choose the basis for seasonal forecasts is specifically with each crop and region. Table 4.1 displays all possible associations to ENSO and PDO phase, as well as to PDO modulated ENSO variability, for every power spectrum analyzed.

Therefore, results from Table 4.1 allow for an analysis of each region, specific to each crop. Corn results suggest the most climate variability associations to be found in regions

1, 2, and 6, while soybean yields display the most variability related to climate in regions

3 and 6. Table 4.1 can be used as a guideline in creating seasonal forecasts, by displaying the information needed to determine if each crop in each region is impacted by ENSO,

PDO modulated ENSO, and/or PDO phase. After determining if each crop and region is impacted by certain climate variabilities, the question of how they are impacted becomes a factor.

The results displayed in Tables 4.2 and 4.3 provide the answer to how each crop in each region is impacted by ENSO and PDO phases, by giving averages of each set of detrended yields based on each ENSO phase and ENSO-PDO phase combination. Table

4.2 displays a positive departure from average of corn and soybean yields for El Niño years and a negative departure from average for La Niña years. Table 4.3 displays similar results and is described in section 4.2. Crops require adequate precipitation with no extreme temperatures during the growing season, especially during July and August, in order to produce average or higher than average yields. Results suggest that these conditions are more likely to occur in growing seasons prior to El Niño events. Since a negative departure from average yields is found for La Niña years, it is possible that extreme temperatures and drought are common during the growing seasons prior to La

Niña events. ENSO impacts for each climate region in Missouri only differ slightly in magnitude, which remains consistent with the crop yield results in Tables 4.2 and 4.3.

Specific ENSO and/or PDO impacts on Missouri regional temperature and precipitation

Crop Yields 2012 ENSO 2012

can be found using several different applications online, including Useful to Usable

(U2U; mygeohub.org/groups/u2u/cpv), NOAA’s Climate Prediction Center (CPC; http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/states/MO.html), and NOAA’s Earth System Research Laboratory (ESRL; http://www.esrl.noaa.gov/psd/data/usclimdivs/correlation/).

ENSO years in this study begin after the harvest of the crops used in this study

(refer to section 2.1). Figure 5.1 displays a timeline graphic to represent the crop growing season, as compared to the JMA defined ENSO year. It is realized that the growing season of the crop falls in a time period that could display either similar index values to the JMA ENSO year or values that represent a transition into the phase defined by the

JMA ENSO year. For either situation, the growing season should not experience drastically different impacts or resulting yields than what the JMA ENSO year phase would suggest. July and August weather has the greatest impact on crop yields and those months fall slightly before the JMA ENSO year begins. Similar strategies have been used involving the time period before an ENSO phase occurs, which can be referred to as the

summer transition period (Lupo et al. 2008, 2014; Mokhov et al. 2014; Newberry et al.

2016).

The JMA ENSO year was chosen for this study based on the statistical significance of the results. Tables 4.2 and 4.3 were recreated using the ENSO year beginning in the winter before the crop growing season, but no statistically significant results were found (refer to Appendix B). Due to the results from this study displaying the relationship of crop yields to the ENSO phase beginning after the crop harvest (Fig.

5.1), the predictability of crop yields will depend on the advanced predictability of

ENSO. Therefore, crop yield forecasts will be seasonal, on the order of three to twelve months.

Seasonal forecasts may be more accurate for the regions displaying statistically significant differences in yield between ENSO phases or ENSO-PDO phase combinations, which are displayed in Tables 4.2 and 4.3, respectively. 70% of Missouri corn production from 1919 to 2013 and 55% percent of Missouri soybean production from 1944 to 2013 exhibit a statistically significant difference in yields between El Niño and La Niña years. It is fair to say ENSO has an impact on Missouri crops and has the possibility to be used as an aid in creating crop yield predictions. Table 4.3 displays less impressive results than Table 4.2, with only one region for corn and two for soybean exhibiting a statistically significant difference in yields between El Niño/Positive PDO years and El Niño/Negative PDO years. However, the averages in Table 4.3 are still compelling evidence of a possible interaction between ENSO and PDO, with PDO contributing to ENSO impacts on Missouri crop yields.

Out of all six of the Missouri climate regions, region 6 is most promising for accurate crop yield predictions. Corn and soybean yields differ significantly between El

Niño and La Niña years, as well as between El Niño/Positive PDO and El Niño/Negative

PDO years, showing high potential for skillful forecasts of yields based on ENSO and

PDO phases. These results are very impressive, especially because region 6 produces roughly 20% of both corn and soybean yields for the state of Missouri. Regions 1, 2, and

6 display statistically significant results for both crops, while region 3 is only significant for soybean and region 4 for corn. Therefore, regions 1 and 2 show the most promise after region 6. However, regions 1 and 2 are only significant when dealing with ENSO phase differences. When considering PDO as a predictor, only regions 3 and 6 show promise. The likelihood of skillful yield predictions based on ENSO and PDO phase depends on the specific crop and climate region in question. Although, it is safe to assume all predictions for region 5 will have the lowest skill compared to other climate regions, because no statistically significant differences in yields were found. Region 5 mostly consists of Ozark Mountains and Mark Twain National Forest, so it was no surprise to find the large region to only produce roughly 15% of Missouri corn and soybean yields, with no encouraging statistical results from the ENSO and PDO phase analyses.

The statistical analysis displayed more impressive results based on ENSO phase only, rather than both ENSO and PDO phase. Four climate regions showed significance when dealing with ENSO phase, while only two regions showed significance after including PDO phase. The power spectra analysis also reveals a less apparent PDO influence on Missouri crop yields, compared to the ENSO influence. Only 36% of the

analyzed spectra represented a multidecadal variability, while 97% displayed an interannual variability. However, PDO still has an impact on Missouri crops through an interdecadal variability found in 75% of the analyzed spectra, associated with a possible

ENSO-PDO interaction harmonic.

Results from the convolution of annual SOI and annual mean values of the PDO index (Fig. 4.2c), suggest that the interaction harmonic between ENSO and PDO occurs at an interdecadal timescale of roughly 10 years. Comparing averages from each ENSO phase alone (Table 4.2) to averages from each ENSO-PDO phase combination (Table

4.3), highly suggests that PDO influences ENSO impacts. An example was provided in section 4.2, supporting the conclusions from Lupo et al. (2007, 2008) that strong El Niño signals occur during years of positive PDO, while strong La Niña events occur during years of negative PDO (Gershunov and Barnett 1998; Birk et al. 2010). Therefore, the findings in this study support the theory of PDO modulated ENSO-related interdecadal variability found in Midwestern climates (Gershunov and Barnett 1998; Berger et al.

2002; Birk et al. 2010).

CHAPTER 6. CONCLUSIONS

The results in this study provide a solid foundation for future seasonal forecasts of corn and soybean yields, specific to each climate region in Missouri, based on both

ENSO and PDO phase. Correlation coefficients were used to confirm the fact that temperature and precipitation have a relationship with crop yields. Power spectra analysis revealed the variability of temperature, precipitation, and crop yield data for each climate region. The resulting variability was associated with interannual, interdecadal, and multidecadal variability to represent possible relationships to ENSO and PDO, as well as

PDO modulated ENSO. Statistical testing quantified the ENSO, PDO, and crop yield relationship for each crop in each region, verifying the possible existence of PDO influenced ENSO events.

The results displayed a slight difference in ENSO and PDO impacts across the climate regions of Missouri, depending on the crop analyzed. However, every crop in every region displayed a possible relationship to PDO influenced ENSO. Therefore, the future crop yield forecasts based on these results must be specific to each climate region in Missouri and must be based on both PDO and the future phase of ENSO, in order to have a chance for high skill scores. Yield forecasts created for the entire state of Missouri or based on only one climate variability, ENSO or PDO, have a higher risk of inaccuracy and low skill scores.

ENSO and PDO effect Missouri temperature and precipitation, while temperature and precipitation impact Missouri corn and soybean yields. Thus, ENSO and PDO impact

Missouri crops. Knowledge of crop yield changes in relationship to ENSO and PDO phase can greatly benefit the agricultural community and economics through the use of skillful forecasts on the seasonal time range of three to twelve months. This thesis has created the possibility for these skillful forecasts, while contributing knowledge of the

ENSO and PDO impacts on regional Missouri corn and soybean yields.

APPENDIX A

The following figures represent the time series and power spectra results from the data used in this study for the remaining crops and regions. The methodology used to create the following figures is consistent with the methodology described in section 3.2.

The list begins with the power spectrum resulting from the covariance of average July temperature and precipitation with detrended corn yields for region 1 (Fig. A.1). Figure

A.1 was not used to describe the methodologies in this paper for simplicity. The following pages list the remaining graphs for corn, region 2 through 6, and soybean, region 1 through 6. These figures represent annual, as well as detrended annual crop yields, along with power spectra representing detrended crop yields and convolutions of temperature, precipitation, and crop yields for the crop growing season and highest correlated monthly data. All three spectra results for each crop and region were analyzed and included in Table 4.1.

T/P/Corn 1 (July) 4 910

4 6.7510

4 4.510

Magnitude 4 2.2510

0 0 6 12 18 24 30

Wave Number

Figure A.1. As in Fig. 4.1d, except involving average July temperature (T) and average July precipitation (P) data for region 1.

a) Missouri Corn Yields - Region 2 200 150 100 50

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 2, Detrended

 50

Yield(Bu/Acre)  100 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.2. As in Fig. 3.2, except annual corn yields for region 2 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Corn Yields - Region 2 3 1.510

3 1.12510

750

Magnitude 375

0 0 6 12 18 24 30

Wave Number T/P/Corn 2 (Season) T/P/Corn 2 (July) b) 3 c) 4 410 910

3 4 310 6.7510

3 4 210 4.510

Magnitude Magnitude 3 4 110 2.2510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

a) Missouri Corn Yields - Region 3 150

100

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 3, Detrended 100

 50

Yield(Bu/Acre)  100 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.4. As in Fig. 3.2, except annual corn yields for region 3 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Corn Yields - Region 3 3 1.510

3 1.12510

750

Magnitude 375

0 0 6 12 18 24 30

Wave Number T/P/Corn 3 (Season) T/P/Corn 3 (July) b) 3 c) 4 510 810

3 4 3.7510 610

3 4 2.510 410

Magnitude Magnitude 3 4 1.2510 210

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.5. As in Fig. A.3, except involving detrended corn yields for region 3 in Missouri, found in Fig. A.4b.

a) Missouri Corn Yields - Region 4 150

100

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 4, Detrended 60

 30

Yield(Bu/Acre)  60 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.6. As in Fig. 3.2, except annual corn yields for region 4 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Corn Yields - Region 4 3 210

3 1.510

3 110

Magnitude 500

0 0 6 12 18 24 30

Wave Number b) T/P/Corn 4 (Season) T/P/Corn 4 (July) 3 c) 4 410 410

3 4 310 310

3 4 210 210

Magnitude Magnitude 3 4 110 110

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.7. As in Fig. A.3, except involving detrended corn yields for region 4 in Missouri, found in Fig. A.6b.

a) Missouri Corn Yields - Region 5 150

100

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 5, Detrended 40

 10

 35

Yield(Bu/Acre)  60 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.8. As in Fig. 3.2, except annual corn yields for region 5 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Corn Yields - Region 5 3 1.510

3 1.12510

750

Magnitude 375

0 0 6 12 18 24 30

Wave Number T/P/Corn 5 (Season) T/P/Corn 5 (July) b) 3 c) 4 410 210

3 4 310 1.510

3 4 210 110

Magnitude Magnitude 3 3 110 510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.9. As in Fig. A.3, except involving detrended corn yields for region 5 in Missouri, found in Fig. A.8b.

a) Missouri Corn Yields - Region 6 200

150

100

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Corn Yields - Region 6, Detrended 50

 25

Yield(Bu/Acre)  50 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.10. As in Fig. 3.2, except annual corn yields for region 6 in Missouri from 1919 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Corn Yields - Region 6 3 1.610

3 1.210

800

Magnitude 400

0 0 6 12 18 24 30

Wave Number T/P/Corn 6 (Season) T/P/Corn 6 (July) b) 3 c) 4 410 210

3 4 310 1.510

3 4 210 110

Magnitude Magnitude 3 3 110 510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.11. As in Fig. A.3, except involving detrended corn yields for region 6 in Missouri, found in Fig. A.10b.

a) Missouri Soybean Yields - Region 1 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 1, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.12. As in Fig. 3.2, except annual soybean yields for region 1 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 1 100

Magnitude

0 0 6 12 18 24 30

Wave Number b) T/P/Soybean 1 (Season) T/P/Soybean 1 (August) 3 c) 4 410 210

3 4 310 1.510

3 4 210 110

Magnitude Magnitude 3 3 110 510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

a) Missouri Soybean Yields - Region 2 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 2, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.14. As in Fig. 3.2, except annual soybean yields for region 2 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 2 100

Magnitude

0 0 6 12 18 24 30

Wave Number T/P/Soybean 2 (Season) T/P/Soybean 2 (August) b) 3 c) 4 410 210

3 4 310 1.510

3 4 210 110

Magnitude Magnitude 3 3 110 510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.15. As in Fig. A.13, except involving detrended soybean yields for region 2 in Missouri, found in Fig. A.14b.

a) Missouri Soybean Yields - Region 3 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 3, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.16. As in Fig. 3.2, except annual soybean yields for region 3 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 3 100

Magnitude

0 0 6 12 18 24 30

Wave Number b) T/P/Soybean 3 (Season) T/P/Soybean 3 (August) 3 c) 4 810 410

3 4 610 310

3 4 410 210

Magnitude Magnitude 3 4 210 110

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.17. As in Fig. A.13, except involving detrended soybean yields for region 3 in Missouri, found in Fig. A.15b.

a) Missouri Soybean Yields - Region 4 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 4, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.18. As in Fig. 3.2, except annual soybean yields for region 4 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 4 120

Magnitude

0 0 6 12 18 24 30

Wave Number T/P/Soybean 4 (Season) T/P/Soybean 4 (August) b) 3 c) 4 610 310

3 4 4.510 2.2510

3 4 310 1.510

Magnitude Magnitude 3 3 1.510 7.510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.19. As in Fig. A.13, except involving detrended soybean yields for region 4 in Missouri, found in Fig. A.18b.

a) Missouri Soybean Yields - Region 5 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 5, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.20. As in Fig. 3.2, except annual soybean yields for region 5 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 5 100

Magnitude

0 0 6 12 18 24 30

Wave Number T/P/Soybean 5 (Season) T/P/Soybean 5 (August) b) 3 c) 3 1.510 610

3 3 1.12510 4.510

3 750 310

Magnitude Magnitude 3 375 1.510

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.21. As in Fig. A.13, except involving detrended soybean yields for region 5 in Missouri, found in Fig. A.20b.

a) Missouri Soybean Yields - Region 6 60

Yield(Bu/Acre) 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

b) Missouri Soybean Yields - Region 6, Detrended 30

 15

Yield(Bu/Acre)  30 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020

Year

Figure A.22. As in Fig. 3.2, except annual soybean yields for region 6 in Missouri from 1944 to 2013, with a) the linear trend in yields shown and b) the trend removed.

a) Soybean Yields - Region 6 100

Magnitude

0 0 6 12 18 24 30

Wave Number T/P/Soybean 6 (Season) T/P/Soybean 6 (August) b) 3 c) 3 210 810

3 3 1.510 610

3 3 110 410

Magnitude Magnitude 3 500 210

0 0 0 6 12 18 24 30 0 6 12 18 24 30

Wave Number Wave Number

Figure A.23. As in Fig. A.13, except involving detrended soybean yields for region 6 in Missouri, found in Fig. A.22b.

APPENDIX B

Figure B.1 displays a timeline graphic of the crop growing season of 2012 and is comparable to Fig. 5.1, except it involves the JMA defined ENSO year of 2011. Tables

B.1 and B.2 represent the same methodology used to create Tables 4.2 and 4.3, respectively. However, Tables B.1 and B.2 compare detrended crop yields to the JMA

ENSO year prior to the year of crop yields. For example, crop yields in 2012 are compared to the JMA ENSO year of 2011, which is the ENSO phase that begins on 1

October 2011 and continues through the growing season of the crops in 2012, before ending in September 2012 (Fig. B.1). No statistically significant differences between

ENSO phases or ENSO and PDO phase combinations were found.

ENSO 2011 Crop Yields 2012

Soybean 1 -0.642 0.129 0.401 Soybean 2 -0.668 0.127 0.434 Soybean 3 -0.741 0.014 0.804 Soybean 4 -0.181 -0.555 1.611 Soybean 5 -0.127 0.093 -0.092 Soybean 6 -1.114 0.713 -0.545

Table B.2. Departure from average calculated from detrended crop yields in bushels per acre, for each crop in each climate region of Missouri, for the years associated with each El Niño Southern Oscillation-Pacific Decadal Oscillation (ENSO-PDO) phase combination of the previous JMA ENSO year, listed in Table 3.1. Bold values represent statistically significant (90% confidence level) differences in yield for El Niño years of different PDO phases, with both data sets involved containing at least 5 samples (yield per phase year combination).

Positive PDO (1) Negative PDO (2) La Niña Neutral El Niño La Niña Neutral El Niño Corn 1 -2.592 0.173 -2.271 0.787 0.793 -0.177 Corn 2 -1.229 0.791 -3.536 1.296 -0.778 1.727 Corn 3 0.368 -0.353 4.566 -1.255 -0.074 -1.534 Corn 4 1.105 -0.957 4.731 -3.226 1.855 -1.984 Corn 5 0.542 0.133 -1.207 -0.664 1.373 -1.819 Corn 6 0.953 1.732 2.261 -0.693 0.053 -6.785

Soybean 1 -6.252 0.203 0.236 0.106 0.070 0.511 Soybean 2 -3.725 0.546 -0.455 -0.260 -0.211 1.027 Soybean 3 -0.116 0.997 1.639 -0.825 -0.781 0.247 Soybean 4 0.158 -1.267 1.380 -0.226 0.021 1.766 Soybean 5 -0.855 0.448 -0.897 -0.030 -0.194 0.445 Soybean 6 -2.978 0.323 -2.360 -0.866 1.029 0.665

REFERENCES

Berger, C. L, A. R. Lupo, P. Browning, M. Bodner, M. D. Chambers, and C. C. Rayburn, 2002: A climatology of Northwest Missouri snowfall events: Long-term trends and interannual variability. Phys. Geo., 23, 427-448.

Birk, K., A. R. Lupo, P. Guinan, and C. E. Barbieri, 2010: The interannual variability of Midwestern temperatures and precipitation as related to the ENSO and PDO. Atmósfera, 23, 95-128.

Ding, J., and B. A. McCarl, 2014: Inter-decadal climate variability in the Edwards Aquifer: Regional impacts of DCV on crop yields and water use. AAEA 2014 Annual Meeting, Minneapolis, MN, Agricultural and Applied Economics Association. [Available online at http://purl.umn.edu/170216.]

Enfield, D. B., and A. M. Mestas-Nuñez, 1999: Multiscale Variabilities in Global Sea Surface Temperatures and Their Relationships with Tropospheric Climate Patterns. J. Climate, 12, 2719-2733.

Gershunov, A., and T. P. Barnett, 1998: Interdecadal modulation of ENSO teleconnections. Bull. Amer. Meteor. Soc., 79, 2715-2725.

Goodrich, G. B., 2004: Modulation of the winter ENSO Arizona climate signal by the Pacific Decadal Oscillation. J. Arizona-Nevada Acad. Sci., 36, 88-94.

Hanley, D. E., M. A. Bourassa, J. J. O’Brien, S. R. Smith, and E. R. Spade, 2003: A quantitative evaluation of ENSO indices. J. Climate, 16, 1249-1258.

Hu, Q., and G. Buyanovsky, 2003: Climate effects on corn yield in Missouri. J. Appl. Meteor., 42, 1626-1635.

Keables, M. J., 1992: Spatial Variability of Mid-tropospheric circulation patterns and associated surface climate in the United States during ENSO Winters. Phys. Geo., 13, 331-348.

Kung, E. C., and J.-G. Chern, 1995: Prevailing anomaly patterns of the Global Sea Surface temperatures and tropospheric responses. Atmósfera, 8, 99-114.

Lee, J. -W., and E. C. Kung, 2000: Seasonal-range forecasting of the Ozark climate by a principal component regression scheme with antecedent sea surface temperatures and upper air conditions. Atmósfera, 13, 223-244.

Lupo, A. R., P. S. Market, F. A. Akyuz, C. L. Allmeyer, D. Albert, and R. Hearst, 2005: Interannual variability of snowfall events of Southwest Missouri and snowfall-to- liquid water equivalents at the Springfield WFO. Nat. Wea. Digest, 29, 13-24.

Lupo, A. R., E. P. Kelsey, D. K. Weitlich, J. E. Woolard, I. I. Mokhov, P. E. Guinan, and F. A. Akyuz, 2007: Interannual and interdecadal variability in the predominant Pacific region SST anomaly patterns and their impact on climate in the mid- Mississippi valley region. Atmósfera, 20, 171-196.

Lupo, A. R., E. P. Kelsey, D. K. Weitlich, N. A. Davis, and P. S. Market, 2008: Using the monthly classification of global SSTs and 500 hPa height anomalies to predict temperature and precipitation regimes one to two seasons in advance for the mid- Mississippi region. Nat. Wea. Digest, 32, 11-33.

Lupo, A. R., N. B. Smith, P. E. Guinan, and M. D. Chesser, 2012a: The climatology of Missouri region dew points and the relationship to ENSO. Nat. Wea. Digest, 36, 81-91.

Lupo, A. R., R. S. Hayward, and G. W. Whitledge, 2012b: Synchronization of fishes’ temporal feeding patterns with weather in mid-Missouri. J. Freshwater Ecol., 27, 419-428.

Lupo, A. R., I. I. Mokhov, Y. G. Chendev, M. G. Lebedeva, M. Akperov, and J. A. Hubbart, 2014: Studying Summer Season Drought in Western Russia. Adv. Meteor., 2014, 1-9, doi:10.1155/2014/942027.

Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific interdecadal climate oscillation with impacts on salmon production. Bull. Amer. Meteor. Soc., 78, 1069-1079.

Mantua, N. J., and S. R. Hare, 2002: The Pacific Decadal Oscillation. J. Oceanogr., 58, 35-44.

Minobe, S., 1997: A 50-70 year climatic oscillation over the North Pacific and North America. Geophys. Res. Lett., 24, 683-686, doi:10.2019/97GL00504.

Minobe, S., 2000: Spatio-temporal structure of the pentadecadal variability over the North Pacific. Prog. Oceanogr., 47, 381-408.

Mokhov, I. I., D. V. Khvorostyanov, and A. V. Eliseev, 2004: Decadal and longer term changes in El Niño-Southern Oscillation characteristics. Int. J. Climatol., 24, 401- 414.

Mokhov, I. I., V. C. Khon, A. V. Timazhev, A. V. Chernokulsky, and V. A. Semenov, 2014: Hydrological anomalies and tendencies of change in the Amur River basin in relation to climate changes. Extreme Floods in the Amur River Basin: Causes, Forecasts, and Recommendations, Moscow, Roshydromet, 81-120.

NCDC, 2006: Climate of Missouri. NOAA, accessed 30 July 2015. [Available online at http://www.crh.noaa.gov/Image/dvn/downloads/Clim_MO_01.pdf.] Newberry, R. G., A. R. Lupo, A. D. Jensen, and R. A. Rodriges-Zalipynis, 2016: An Analysis of the Spring-to-Summer Transition in the West Central Plains for Application to Long Range Forecasting. Atmos. Climate Sci., in press.

Newman, M., G. P. Compo, and M. A. Alexander, 2003: ENSO-Forced Variability of the Pacific Decadal Oscillation. J. Climate, 16, 3853-3857.

Power, S., T. Casey, C. Folland, A. Colman, and V. Mehta, 1999: Inter-decadal modulation of the impact of ENSO on Australia. Climate Dynamics, 15, 319-324.

Ratley, C. W., A. R. Lupo, and M. A. Baxter, 2002: Determining the Spring to Summer Transition in the Missouri Ozarks using Synoptic Scale Data. Trans. MO Acad. Sci., 36, 55-62.

Wang, L., W. Chen, and R. Huang, 2008: Interdecadal modulation of PDO on the impact of ENSO on the east Asian winter monsoon. Geophys. Res. Lett., 35, L20702.

Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. Academic Press, 627 pp.

Zhang, L., and T. L. Delworth, 2015: Analysis of the Characteristics and Mechanisms of the Pacific Decadal Oscillation in a Suite of Coupled Models from the Geophysical Fluid Dynamics Laboratory. J. Climate, 28, 7678-7701.

Zhang, Y., J. M. Wallace, and D. S. Battisti, 1997: ENSO-like interdecadal variability: 1900-93. J. Climate, 10, 1004-1020.