usefulness and uses of climate forecasts for agricultural extension in south carolina, usa

11
ORIGINAL ARTICLE Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA Scott R. Templeton M. Shane Perkins Heather Dinon Aldridge William C. Bridges Jr. Bridget Robinson Lassiter Received: 1 September 2012 / Accepted: 30 July 2013 Ó Springer-Verlag Berlin Heidelberg 2013 Abstract Farmers and extensionists can use forecasts about agro-climatic conditions to reduce risks of agricul- tural production. Eighteen extension agents, researchers, consultants, and farmers provided feedback about decision support tools that utilize such forecasts during focus groups that were conducted in Florence, South Carolina on Janu- ary 14, 2011. Climate Risk and County Yield Database were the tools most selected as potentially useful for agricultural extension in South Carolina. An irrigation scheduler was the most frequently mentioned tool to be developed. Also, a survey of Clemson University’s exten- sion personnel was conducted in January and February 2011 to assess interest of South Carolina’s growers and producers in using climate forecasts, eleven potential uses of climate forecasts by extension’s clientele, and potential usefulness to extensionists of twelve specific forecasts. Clemson’s extensionists represent approximately 97 % of the state’s agricultural extensionists. They are more likely than not to agree that growers and producers are interested in using climate forecasts. Most of the state’s extension personnel also think that farmers could use a climate forecast to improve irrigation management and planting schedules. A majority of the state’s extensionists thinks that a freeze alert could be useful to them and the pro- portion that thinks the forecast could be useful exceeds the proportion that thinks any other forecast could be useful. Most extensionists also think that a forecast of plant moisture stress could be useful to help farmers schedule irrigation. The key survey results are remarkably similar to those from surveys of extension personnel at North Caro- lina State University in early 2009 and University of Florida in late 2004. Keywords Agricultural extension AgroClimate Benefits of forecasts Cochran and McNemar statistics Decision support tools South Carolina agriculture Usefulness and uses of forecasts Introduction The El Nin ˜o-Southern Oscillation (ENSO) and its associ- ated El Nin ˜o, Neutral, and La Nin ˜a phases influence the seasonal climate in the southeastern USA (e.g., Fraisse et al. 2006). As a result, ENSO creates production and revenue risks for farmers in the region (e.g., Hansen et al. 1998; Nadolnyak et al. 2008). Farmers in the southeastern USA and elsewhere can use information about climate variability to change their crop management and, thereby reduce these risks, and improve profitability (e.g., Jones et al. 2000; Solow et al. 1998). Forecasts and historical data about climate become more valuable to farmers if they are S. R. Templeton (&) John E. Walker Department of Economics, Clemson University, Clemson, SC 29634-1309, USA e-mail: [email protected] M. Shane Perkins Tri-County Technical College, P.O. Box 587, Pendleton, SC 29670, USA H. D. Aldridge State Climate Office of North Carolina, North Carolina State University, Raleigh, NC 27695-7236, USA W. C. Bridges Jr. Department of Mathematical Sciences, Clemson University, Clemson, SC 29634, USA B. R. Lassiter Crop Science Department, North Carolina State University, Raleigh, NC 27695-7620, USA 123 Reg Environ Change DOI 10.1007/s10113-013-0522-7

Upload: bridget-robinson

Post on 23-Dec-2016

213 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

ORIGINAL ARTICLE

Usefulness and uses of climate forecasts for agricultural extensionin South Carolina, USA

Scott R. Templeton • M. Shane Perkins •

Heather Dinon Aldridge • William C. Bridges Jr. •

Bridget Robinson Lassiter

Received: 1 September 2012 / Accepted: 30 July 2013

� Springer-Verlag Berlin Heidelberg 2013

Abstract Farmers and extensionists can use forecasts

about agro-climatic conditions to reduce risks of agricul-

tural production. Eighteen extension agents, researchers,

consultants, and farmers provided feedback about decision

support tools that utilize such forecasts during focus groups

that were conducted in Florence, South Carolina on Janu-

ary 14, 2011. Climate Risk and County Yield Database

were the tools most selected as potentially useful for

agricultural extension in South Carolina. An irrigation

scheduler was the most frequently mentioned tool to be

developed. Also, a survey of Clemson University’s exten-

sion personnel was conducted in January and February

2011 to assess interest of South Carolina’s growers and

producers in using climate forecasts, eleven potential uses

of climate forecasts by extension’s clientele, and potential

usefulness to extensionists of twelve specific forecasts.

Clemson’s extensionists represent approximately 97 % of

the state’s agricultural extensionists. They are more likely

than not to agree that growers and producers are interested

in using climate forecasts. Most of the state’s extension

personnel also think that farmers could use a climate

forecast to improve irrigation management and planting

schedules. A majority of the state’s extensionists thinks

that a freeze alert could be useful to them and the pro-

portion that thinks the forecast could be useful exceeds the

proportion that thinks any other forecast could be useful.

Most extensionists also think that a forecast of plant

moisture stress could be useful to help farmers schedule

irrigation. The key survey results are remarkably similar to

those from surveys of extension personnel at North Caro-

lina State University in early 2009 and University of

Florida in late 2004.

Keywords Agricultural extension � AgroClimate �Benefits of forecasts � Cochran and McNemar

statistics � Decision support tools � South Carolina

agriculture � Usefulness and uses of forecasts

Introduction

The El Nino-Southern Oscillation (ENSO) and its associ-

ated El Nino, Neutral, and La Nina phases influence the

seasonal climate in the southeastern USA (e.g., Fraisse

et al. 2006). As a result, ENSO creates production and

revenue risks for farmers in the region (e.g., Hansen et al.

1998; Nadolnyak et al. 2008). Farmers in the southeastern

USA and elsewhere can use information about climate

variability to change their crop management and, thereby

reduce these risks, and improve profitability (e.g., Jones

et al. 2000; Solow et al. 1998). Forecasts and historical data

about climate become more valuable to farmers if they are

S. R. Templeton (&)

John E. Walker Department of Economics, Clemson University,

Clemson, SC 29634-1309, USA

e-mail: [email protected]

M. Shane Perkins

Tri-County Technical College, P.O. Box 587, Pendleton,

SC 29670, USA

H. D. Aldridge

State Climate Office of North Carolina, North Carolina State

University, Raleigh, NC 27695-7236, USA

W. C. Bridges Jr.

Department of Mathematical Sciences, Clemson University,

Clemson, SC 29634, USA

B. R. Lassiter

Crop Science Department, North Carolina State University,

Raleigh, NC 27695-7620, USA

123

Reg Environ Change

DOI 10.1007/s10113-013-0522-7

Page 2: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

presented as decision support tools in non-technical lan-

guage with an interactive computer interface (Breuer et al.

2008b; Fraisse et al. 2006; McCown et al. 2002).

Decision support tools on the website AgroClimate

(http://agroclimate.org/tools) initially became available for

South Carolina in 2010. In particular, County Yield Data-

base and Climate Risk for South Carolina became available

in May and September 2010. Eight additional tools were

operationalized between December 2010 and April 2011.

Peanut Leaf Spot Advisory was added in 2012. AgroCli-

mate’s decision support tools for the state cover drought,

other climate risks, crop yields, crop diseases, and degree

days and chill hours.

Most of AgroClimate’s decision support tools incorpo-

rate past and future information about climate. The tools

were developed with input from farmers and extension

agents in southeastern states of the USA other than South

Carolina, primarily Florida (e.g., Breuer et al. 2008b;

Fraisse et al. 2006). To assess the potential adoption of a

climatic information system for agricultural extension in

South Carolina and give feedback to developers of the

system, we have asked and provided initial answers to four

questions. First, are South Carolina’s growers and pro-

ducers interested in using climate forecasts? Second, what

are the potential uses of a forecast by the state’s farmers?

Third, which climate forecasts are potentially useful to

extension personnel in the state? Fourth, which decision

support tools in AgroClimate might be useful to farmers

and extension agents in the state?

Similar questions have been asked about decision sup-

port systems and climate forecasts for other states in the

USA or other countries. Answers to the questions have

been based on various methodologies, most of which have

not relied on structured surveys or extensively used sta-

tistical inference from survey data. For example, according

to three case studies (McCown et al. 2002), decision sup-

port systems enabled cotton producers in Australia to

substantially reduce their pesticide use and costs, wheat

farmers there to select a better cultivar, and livestock

producers in Texas to analyze the nutritional status of their

animals on pasture and, thereby, increase their profits

through changes in livestock feeding. The decision support

systems were all developed with input from farmers

(McCown et al. 2002). The case studies drew on researcher

recollection of their experiences with farmers in Australia

and Texas.

Anecdotal evidence from interviews, workshops, meet-

ings, and focus groups with farmers and extension agents in

Florida indicated that the potential for adoption of AgCli-

mate, the precursor of AgroClimate, to reduce risks related

to climate variability was encouraging (Fraisse et al. 2006).

Extension agents, row-crop farmers, and ranchers indicated

in 81 sondeos—informal, semi-structured discussions led

by teams of multi-disciplinary scholars—between March

1999 and March 2004 in north-central and other parts of

Florida that improved climate forecasts could help farmers

decide what, where, and when to plant (Breuer et al.

2008b). A majority of 26 extension agents from 13 counties

in southwest and west-central Florida agreed in sondeos

during March 2008 that they could benefit from knowing

which phase of ENSO was predicted and using AgClimate

to plan ahead and protect their crops from an anomalous

climate event (Breuer et al. 2008a). Selection of a crop or a

variety of a crop was the most commonly mentioned

potential use of a seasonal climate forecast by 38 farmers

from 21 southern Georgia counties during 31 semi-struc-

tured interviews between December 2006 and March 2007

(Crane et al. 2010). Changes in planting times and input

use were the second and third most frequently mentioned

potential uses of a forecast about seasonal climate (Crane

et al. 2010).

Two assessments of the usefulness of forecasts and uses

of climate forecasts have been based primarily on quanti-

tative summaries of answers to closed-ended questions

from two closely related, structured surveys. The original

36-question survey was created and conducted by

researchers with the Southeast Climate Consortium

(SECC) among extension agents with the University of

Florida (UF) during November and December 2004

(Cabrera et al. 2006). A subsequent 55-question survey,

which contained most of the original 36 questions, was

conducted by SECC researchers among extension agents

with North Carolina State University (NCSU) from March

6 through April 3, 2009 (Breuer et al. 2011). The extension

agents in both survey populations worked in agricultural

and natural resource management. Seventy-seven, or

86.5 %, of the 89 respondents from UF and 71, or 65.1 %,

of the 109 respondents from NCSU strongly agreed or

agreed that agricultural producers were interested in using

climate information (Breuer et al. 2011 and Cabrera et al.

2006). Moreover, planting schedules, irrigation manage-

ment, and nutrient management were selected by 68.5,

65.2, and 52.8 % of the respondents from UF as activities

that people with whom the respondents worked could use a

climate forecast to improve (Cabrera et al. 2006). Planting

schedules, harvest planning, and selection of a crop or

variety were selected by 85.3, 66.1, and 62.4 % of the

respondents with NCSU as activities that people with

whom the respondents worked could use a climate forecast

to improve (Breuer et al. 2011).

Statistical tests were conducted for differences in the

mean willingness of extension agents at NCSU to provide

advice about climate forecasts conditional on the agent’s

age and gender, work region, or clientele’s farm size

(Breuer et al. 2011). However, although unconditional

sample proportions differed and some exceeded 50 % in

S. R. Templeton et al.

123

Page 3: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

the UF and NCSU survey data, no statistical test for

majorities or differences in unconditional probabilities

among extension agents was conducted. If majority support

for or relative popularity of a forecast or a potential use of a

climate forecast is to guide development of decision sup-

port tools, statistical tests about populations of extension

agents are important. Viewpoints of South Carolina’s ex-

tensionists about climate forecasts are also important for

future tool development.

In this study, we use survey data and the probability

distributions of binomial random variables to test whether a

majority of extensionists at Clemson University and, thus,

in South Carolina share viewpoints about the uses of cli-

mate forecasts and usefulness of forecasts. We also use a

chi-squared statistic of Cochran (Conover 1999) to test for

differences in proportions of Clemson’s extension person-

nel who select which forecasts are useful or which mana-

gerial activities that their clientele could improve with a

climate forecast. If differences exist, we then use the square

root of McNemar’s statistic (Conover 1999) to test whether

one proportion exceeds another. Scholars in multiple dis-

ciplines have used Cochran’s statistic (e.g., Van Berckelaer

et al. 2011) and McNemar’s statistic (e.g., Faravelli 2007)

for related hypothesis tests. We also use focus-group data

to provide preliminary insights into the potential usefulness

of AgroClimate’s decision support tools, but the data can-

not be used for statistical inference.

Methodology

Focus-group data

Our assessment of the potential usefulness of decision

support tools in AgroClimate is primarily based on data

from focus-group participants. Participants were recrui-

ted from people who attended a SECC workshop about

AgroClimate on January 14, 2011 at Clemson Univer-

sity’s Pee Dee Research and Education Center in Flor-

ence, South Carolina. Twenty-two people attended the

workshop and received 50 min of instruction about South

Carolina’s climate from State Climatologist Hope Mizz-

ell. They then received 75 min of instruction from Clyde

Fraisse, Climate Extension Specialist with the University

of Florida, about the County Yield Database, Climate

Risk, and other decision support tools in AgroClimate for

South Carolina.

Two focus groups were conducted after lunch for

1 hour, led by Templeton and Jessica Whitehead, then

Regional Climate Extension Specialist for S.C. Sea Grant.

Attendees had been notified before the workshop about the

focus groups and at the end of the workshop were

encouraged, but not required, to participate. Perkins and

Lassiter helped to develop the focus-group questions, pre-

pare materials, and record responses. Mizzell and Fraisse

provided technical expertise for each group. Nine partici-

pants were preassigned to each focus group for diversity:

three extension agents, two experiment-station or Agri-

cultural-Research-Service researchers, one farmer, and

three others, such as a crop loan specialist, plant inspector,

crop consultant, other extension agent, or other researcher.

After introducing themselves and their backgrounds, par-

ticipants shared their impressions of AgroClimate. Each

participant was then asked to write and publicly state their

answers to three questions: (1) Which three decision sup-

port tools in AgroClimate would be most useful to exten-

sion agents? (2) Which three tools in AgroClimate would

be most useful to farmers? (3) What is missing in Agro-

Climate, or what new decision support tool should be

developed? In one of the groups, ‘decision support tools in

AgroClimate’ meant existing and yet-to-be-developed

tools. Each participant’s six votes—the three most useful

forecasts for extension agents and the three for farmers—

were publicly tallied.

Survey data

Our assessments of interest in and potential uses of climate

forecasts and usefulness of forecasts for South Carolina are

based on analysis of data from a structured survey of

Clemson University’s extension personnel conducted in

January and February 2011. The survey was adapted from

the one conducted in North Carolina (Breuer et al. 2011).

Our survey population consisted of 171 employees, 154

permanent and 17 temporary, who were extension associ-

ates, agents, or specialists for Clemson University. The

Small Farm Assistance and Outreach Program of South

Carolina State University, the state’s 1890 land-grant

institution, had six extension personnel in 2011, according

to Dr. Edoe Agbodjan, the program’s director. Thus, our

survey population represented almost 97 % of South Car-

olina’s agricultural extensionists.

Procedures of Dillman et al. (2009) were followed to

enhance the quality and quantity of survey responses. The

Experiment Station and Extension Directors endorsed the

survey and provided contact information about extension

personnel. An email preview of the upcoming survey was

sent on December 20, 2010 to permanent employees and a

request to participate was emailed on January 6, 2011. A

reminder to complete the survey was sent on January 13,

2011, 1 day before the AgroClimate workshop, to several

extension personnel who had registered for the workshop.

A follow-up request to complete the survey was sent by the

Experiment Station and Extension Directors on January 28,

2011 to all permanent employees who had not attended the

workshop. A final request to complete the survey was sent

Usefulness and uses of climate forecasts

123

Page 4: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

on February 3, 2011, roughly 4 weeks after the initial

invitation, to all permanent employees who had not atten-

ded the workshop.

Similar procedures were followed for extension per-

sonnel who were temporary employees and did not attend

the workshop. They were sent a preview of the survey on

February 2, 2011 and an invitation to participate in the

survey on February 10, 2011. A follow-up email from the

Experiment Station and Extension Directors was sent out to

the temporary extension personnel on February 18, 2011.

The final survey reminder was sent out to them on February

24, 2011.

Forty-nine, or 29 %, of Clemson’s extension personnel

responded. The response rate was within the range of

reported rates of response to various online surveys and

near the mean response rate, 33 % (Nulty 2008). Males

accounted for 81 % of the respondents but 63 % of the

survey population. Twenty-six, or 53.1 % of all, respon-

dents were extension agents. Eighteen, or 36.7 % of all,

respondents were extension specialists. Five respondents,

or 10.2 %, were extension associates. Extension agents,

specialists, and associates account for 63.2, 28.1, and 8.8 %

of Clemson’s extension personnel. In short, extension

associates were proportionally represented, males and

specialists slightly over proportionally represented, and

extension agents slightly under proportionally represented

in the sample.

Variables from survey data

Categorical responses to two statements and a question in

the survey are the sources of data for variables. The first

statement is, ‘‘In my opinion, growers and producers

(including forest owners, livestock producers, etc.) in my

region are interested in using climate forecasts. (1) strongly

agree, (2) agree, (3) neither agree nor disagree, (4) dis-

agree, or (5) strongly disagree.’’ Let Yr = one if respondent

r selects ‘strongly agree’ or ‘agree’ and zero if not and let

Y �P49

r¼1 Yr be the number of respondents who select

‘strongly agree’ or ‘agree.’

The second statement is, ‘‘People I work with can use

climate forecasts to improve … (Check all that apply.).’’

The managerial activities that might be improved with

climate forecasts are these: (1) planting schedules, (2)

allocation of land to crops or activities, (3) labor man-

agement, (4) harvest planning, (5) waste management, (6)

nutrient management, (7) irrigation management, (8)

marketing, (9) variety or crop selection, (10) spacing or

stand density, (11) integrated pest management, and (12)

other. Let Ira equal one if respondent r checks managerial

activity a and zero if not and Ia �PR

r¼1 Ira be the number

of respondents who check activity a.

The question is ‘‘Which of these forecasts could be

useful to you? (Check all that apply.).’’ The potentially

useful forecasts are these: (1) freeze alert, (2) wildfire risk,

(3) climate risk, (4) disease risk, (5) El Nino or La Nina

phase, (6) growing degree days, (7) cooling–heating degree

days, (8) lawn and garden moisture index, (9) yield risk,

(10) chill hours accumulation, (11) plant moisture stress,

(12) livestock heat stress index, and (13) other. Let Urf = 1

if respondent r checks forecast f and 0 if not and Uf �P49

r¼1 Urf be the number of respondents who check forecast

f.

Statistical methods for hypothesis tests

The random variables Y, Ia, and Uf have binomial distri-

butions in which PY , PIa, and PUf

are, by definition, the

probabilities that a extensionist thinks farmers are inter-

ested in using climate forecasts, the people with whom he

or she works could use a climate forecast to improve

activity a, and forecast f could be useful to him or her. Of

course, each of the probabilities is also a population pro-

portion and a hypothesis test of a probability in excess of

0.50 is equivalent to a test for a majority. Consider, for

example, the null hypothesis that a majority of extension-

ists think farmers are not interested in using climate fore-

casts. Let y be the actual number of respondents who

strongly agreed or agreed that farmers were interested. If

the probability that at least y respondents would do so is

small enough, then we reject the null and conclude that a

majority of extensionists think farmers are interested. The

binomial probability that at least y respondents would

strongly agree or agree is PrðY � yÞ ¼P49

j¼y

49

j

� �

0:549

and was calculated in a spreadsheet with a macrofunction

for combinations. Hypotheses about PIaand PUf

exceeding

0.50 were tested with probabilities similarly calculated in a

spreadsheet.

Do the probabilities that a climate forecast could

improve specific managerial activities differ? Does the

probability that an extensionist thinks a forecast could be

useful differ from the probability that he or she thinks all

other forecasts could be? If so, for which pairs of uses or

forecasts do the probabilities differ? Empirical evidence to

answer these questions cannot come from test statistics that

require independently drawn samples from which sample

proportions are calculated. The survey and the sampling

procedure—the same respondents were asked about all

potential uses of a forecast and all potentially useful fore-

casts—represent an experimental design called randomized

complete block (Conover 1999). Each respondent is a

randomized ‘block,’ each forecast or use of a forecast is a

‘treatment,’ and each forecast or use of a forecast that a

S. R. Templeton et al.

123

Page 5: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

respondent checks is a treatment ‘success.’ Instead, evi-

dence can come from Cochran’s statistic and the square

root of McNemar’s statistic (Conover 1999).

Cochran’s statistic to test whether the probabilities of an

extension client using a climate forecast to improve at least

two managerial activities differ is CSI ¼ 11�

10

P11

a¼1Ia� I

11ð Þ2P49

r¼1Ir 11�Irð Þ

, in which Ir �P11

a¼1 Ira is the number of

managerial activities that respondent r indicates could be

improved by a climate forecast and I �P49

r¼1

P11a¼1 Ira is

the number of times that all respondents check any activity

as capable of being improved with a climate forecast. The

subscript a runs to 11 because the twelfth activity, ‘other,’

is too vague. Given the null hypothesis of universal

equality and a ‘large’ sample, the distribution of CSI is

approximately chi-squared with 10 degrees of freedom. We

reject HIall0 : PI1

¼ PI2¼ � � � ¼ PI11

for possible improve-

ments in all eleven activities in favor of HIall1 : PIa

6¼ PIb;

a 6¼ b for at least two different activities, if Pr v210

�CSI� 0:05.

The hypothesis that PIa, the proportion of extensionists

who think people with whom they work will use a forecast

to improve managerial activity a, exceeds PIb, the pro-

portion of extensionists who think that their farm clients

will use a forecast to improve managerial activity b, is

tested with the square root of McNemar’s statistic,ffiffiffiffiffiffiffiffiffiMSIp

¼ IajðIb¼0Þ½ �� IbjðIa¼0Þ½ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiIajðIb¼0Þ½ �þ IbjðIa¼0Þ½ �

p . InffiffiffiffiffiffiffiffiffiMSIp

, IbjðIa ¼ 0Þ �P49

r¼1 IrbjðIra ¼ 0Þ½ � is the number of respondents who

check that a climate forecast could improve activity b but

not activity a and IajðIb ¼ 0Þ �P49

r¼1 IrajðIrb ¼ 0Þ½ � is the

number who indicate a climate forecast could improve

activity a but not activity b. Given a ‘large’ sample and the

null hypothesis that probabilities for the two different

activities are equal, the square root of McNemar’s statistic

is approximately distributed as standard normal. We reject

HItwo0 : PIa

�PIbfor the two activities in favor of HItwo

1 :

PIa[ PIb

if Pr Z�ffiffiffiffiffiffiffiffiffiMSIp� �

� 0:05.

Cochran’s statistic to test for inequality among the

twelve probabilities that an extensionist thinks forecasts

could be useful is CSU ¼ 12 � 11

P12

f¼1Uf�U

12ð Þ2P49

r¼1Ur 12�Urð Þ

, in which

Ur �P12

f¼1 Urf , U �P49

r¼1

P12f¼1 Urf , and f runs to 12

because the thirteenth forecast, ‘other,’ is too vague. The

square root of McNemar’s statistic to test whether PUf, the

probability that an extensionist thinks forecast f could be

useful, exceeds PUg, the probability that he or she thinks

forecast g could be useful, isffiffiffiffiffiffiffiffiffiffiffiMSUp

¼Uf jðUg¼0Þ½ �� UgjðUf¼0Þ½ �ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

Uf jðUg¼0Þ½ �þ UgjðUf¼0Þ½ �p , where Uf jðUg ¼ 0Þ �

P49r¼1 Urf j�

ðUrg ¼ 0Þ� and UgjðUf ¼ 0Þ �P49

r¼1 UrgjðUrf ¼ 0Þ� �

.

Given a ‘large’ sample and the null hypotheses that all

twelve probabilities and two specific probabilities are

equal, the respective distributions of CSU andffiffiffiffiffiffiffiffiffiffiffiMSUp

are

approximately chi-squared with 11 degrees of freedom and

standard normal. The statistics were calculated and alter-

native hypotheses tested with JMP�, Ver. 9 (1989–2012)

software.

Results

Focus group

County Yield Database and Climate Risk were most and

second most likely among decision support tools in Agro-

Climate to be useful to extension agents in South Carolina,

according to focus-group votes (Fig. 1). Agricultural Ref-

erence Index for Drought (ARID) and Strawberry Advisory

System received the third most and same number of votes

for usefulness to extension agents in the state (Fig. 1).

Climate Risk and Yield Risk were equally and most likely to

be useful to farmers. County Yield Database and Agricul-

tural Reference Index for Drought were the third and fourth

most popular tools for usefulness to farmers. An irrigation

scheduler was identified as the most important tool yet to

be developed. Smart-phone applications for disease

advisories were the next priority for development.

Survey

Seventy-one percent of the respondents agreed or strongly

agreed that growers and producers in their region were

interested in using climate forecasts. The probability that at

Fig. 1 Focus-group votes for potential usefulness of decision support

tools to farmers, extension agents, or both

Usefulness and uses of climate forecasts

123

Page 6: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

least 35 of 49 extensionists would agree is 0.0019, so the

null of no majority is rejected.

Six managerial activities that extension’s clientele could

improve with a climate forecast were checked by at least

half of the respondents (Table 1). However, the null

hypothesis of no majority is strongly and weakly rejected

for only two most checked activities. In particular, the

probabilities that irrigation management and planting

schedules are at least as popular as the respondents indi-

cated, given the null hypotheses, were 0.011 and 0.076.

Moreover, the Cochran statistic under the null hypothesis

of equal proportions is 75.65. The associated p value is less

than 0.001 and the null is rejected. Thus, the probabilities

that extension clientele can use a climate forecast to

improve eleven management activities in South Carolina

are not all equal. As a result, 55 pairwise comparisons were

made of potential uses of a climate forecast (Table 1).

The probability that an extensionist thinks that farmers

could use a climate forecast to improve irrigation manage-

ment is not necessarily greater than the probability that he or

she thinks that farmers could use a climate forecast to

improve planting schedules or harvest planning (Table 1).

Farmers are also more likely to use forecasts to improve

allocation of land to crops or activities, selection of crops or

crop varieties, and integrated pest management than to

improve the spacing or stand density of planted trees, mar-

keting, labor management, or waste management (Table 1).

The probabilities that a climate forecast could improve

nutrient management and the spacing or stand density of

planted trees do not necessarily differ but exceed the prob-

abilities that forecasts could improve marketing, labor

management, and waste management, which are activities

least likely to be improved by a climate forecast (Table 1).

Five forecasts were checked by a majority of respon-

dents as potentially useful to them (Table 2). However, the

null hypothesis of no majority in the population is rejected

for only the two most checked forecasts. The probabilities

that a freeze alert and a forecast of plant moisture stress are

at least as popular as the respondents indicated, given the

null hypotheses, were 2.86 9 10-8 and 0.043. Moreover,

the Cochran statistic under the null hypothesis of equal

proportions is 72.75. The associated p value is less than

0.001 and the null of universal equality is rejected. In short,

the proportions of Clemson extension personnel who think

various forecasts are useful are not all equal. As a result, 66

comparisons of the potential usefulness of two forecasts

were made.

The probability that a freeze alert could be useful is

significantly greater than the probability that any other

forecast could be useful (Table 2). Although less likely to

be useful than a freeze alert, forecasts about plant moisture

stress, the El Nino or La Nina phase, growing degree days,

and chill hours accumulation are more likely to be useful

than the five forecasts with the lowest sample frequencies

of being useful (Table 2). A minority of extensionists

thinks that forecasts of disease, wildfire, and yield risks and

a livestock heat stress index would be useful to them.

Discussion

A top vote getter in the focus groups Climate Risk displays

probabilistic forecasts of the upcoming ENSO phase and

information about historical monthly temperatures and

precipitation in each county for each ENSO phase and all

phases combined. Climate Risk might have tied for most

popular tool because the focus-group participants had just

learned about the relatively accurate forecasts of the ENSO

phase in the southeastern USA, particularly during fall and

winter months (Hansen et al. 1998), and had received

detailed hands-on instruction about how to use the tool. In

contrast, most extensionists do not think a forecast of cli-

mate risk could be useful for the possible reason that most

do not know about it. Respondents were not provided a

description of any forecast in the survey.

County Yield Database, the other top vote getter, was

also described and used in an instructional exercise for

almost as much time as Climate Risk was in the workshop.

A display of historical county-level crop yields, County

Yield Database is not a forecast, however, and, thus, was

not in the survey. The three other most popular tools—

Yield Risk, ARID, and the Strawberry Advisory System—

were also introduced during the workshop. None of the

three was operational for South Carolina at the time of the

workshop, however, and only yield risk was included in the

survey as a possibly useful forecast.

The popularity of decision support tools based on votes

of 18 focus-group participants cannot be generalized.

However, the popularity of forecasts based on survey

responses can be generalized with caution. For example,

most South Carolina extensionists think that farmers would

be interested in using climate forecasts and, in particular,

using one to improve irrigation management or planting

schedules. This generalization reflects the statistical evi-

dence already presented and additional survey evidence:

irrigation planning and planting dates were the first and

second most frequently checked managerial activities—by

36 and 35 of the 49 respondents—that crop growers could

improve if they had better climate forecast information.

Most extension agents with NCSU and the UF tend to

concur (Breuer et al. 2011; Cabrera et al. 2006). In par-

ticular, farmer use of a climate forecast to improve planting

schedules was most frequently selected by NCSU and UF

agents. Irrigation management was the second and fourth

most selected use of a climate forecast by UF and NCSU

agents.

S. R. Templeton et al.

123

Page 7: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

Ta

ble

1S

amp

lefr

equ

enci

esan

dp

val

ues

of

McN

emar

-bas

ed,

on

e-si

ded

Zte

sts

for

exce

edan

ces

inp

rob

abil

itie

sth

atp

eop

lew

ith

wh

om

exte

nsi

on

ists

wo

rkco

uld

use

acl

imat

efo

reca

stto

imp

rov

em

anag

eria

lac

tiv

itie

s(n

=4

9)

Man

ager

ial

acti

vit

yS

amp

le

freq

uen

cies

Pla

nti

ng

sch

edu

les

Har

ves

t

pla

nn

ing

All

oca

tio

no

fla

nd

to

cro

ps

or

acti

vit

ies

Var

iety

or

cro

p

sele

ctio

n

Inte

gra

ted

pes

t

man

agem

ent

Nu

trie

nt

man

agem

ent

Sp

acin

go

r

stan

dd

ensi

ty

Mar

ket

ing

Lab

or

man

agem

ent

Was

te

man

agem

ent

Irri

gat

ion

man

agem

ent

33

0.2

19

30

.15

87

0.0

63

30

.01

63

0.0

29

70

.00

38

0.0

00

3\

0.0

00

1\

0.0

00

1\

0.0

00

1

Pla

nti

ng

sch

edu

les

30

0.3

90

80

.19

69

0.0

82

80

.14

86

0.0

22

80

.00

37

0.0

00

1\

0.0

00

1\

0.0

00

1

Har

ves

tp

lan

nin

g2

90

.27

43

0.1

72

90

.17

29

0.0

63

30

.00

58

\0

.00

01

\0

.00

01

\0

.00

01

Lan

dal

loca

tio

nto

cro

ps

or

acti

vit

ies

26

0.4

09

30

.42

88

0.2

16

40

.03

68

0.0

00

90

.00

06

0.0

00

1

Var

iety

or

cro

p

sele

ctio

n

25

0.5

00

00

.21

93

0.0

26

10

.00

22

0.0

00

20

.00

02

Inte

gra

ted

pes

t

man

agem

ent

25

0.2

45

70

.04

48

0.0

01

40

.00

03

0.0

00

1

Nu

trie

nt

man

agem

ent

22

0.1

58

70

.01

09

0.0

01

40

.00

01

Sp

acin

go

rst

and

den

sity

18

0.0

44

80

.01

63

0.0

06

2

Mar

ket

ing

11

0.3

69

50

.20

27

Lab

or

man

agem

ent

10

0.2

39

8

Was

tem

anag

emen

t8

Usefulness and uses of climate forecasts

123

Page 8: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

Ta

ble

2S

amp

lefr

equ

enci

esan

dp

val

ues

of

McN

emar

-bas

ed,

on

e-si

ded

Zte

sts

for

exce

edan

ces

inp

rob

abil

itie

sth

attw

ofo

reca

sts

cou

ldb

eu

sefu

lto

anex

ten

sio

nis

t(n

=4

9)

Fo

reca

stS

amp

le

freq

uen

cy

Pla

nt

mo

istu

re

stre

ss

El

Nin

oo

r

La

Nin

a

ph

ase

Gro

win

g

deg

ree

day

s

Ch

ill

ho

urs

accu

mu

lati

on

Co

oli

ng

or

hea

tin

g

deg

ree

day

s

Cli

mat

e

risk

Law

nan

d

gar

den

mo

istu

re

ind

ex

Dis

ease

risk

Wil

dfi

re

risk

Yie

ld

risk

Liv

esto

ck

hea

tst

ress

ind

ex

Fre

eze

aler

t4

30

.00

14

0.0

00

50

.00

03

0.0

00

1\

0.0

00

1\

0.0

00

1\

0.0

00

1\

0.0

00

1\

0.0

00

1\

0.0

00

1\

0.0

00

1

Pla

nt

mo

istu

rest

ress

31

0.3

27

40

.23

35

0.1

37

60

.04

41

0.0

14

50

.00

14

0.0

00

30

.00

27

0.0

01

40

.00

04

El

Nin

oo

rL

aN

ina

ph

ase

29

0.4

13

70

.27

43

0.1

00

40

.05

42

0.0

25

00

.00

53

0.0

07

90

.00

47

0.0

01

3

Gro

win

gd

egre

ed

ays

28

0.2

96

50

.06

59

0.0

89

90

.01

95

0.0

08

20

.01

70

0.0

02

40

.00

06

Ch

ill

ho

urs

accu

mu

lati

on

26

0.2

19

30

.19

69

0.0

54

20

.01

95

0.0

34

00

.01

27

0.0

03

4

Co

oli

ng

–h

eati

ng

deg

ree

day

s

23

0.4

13

70

.14

25

0.1

10

40

.08

08

0.0

72

20

.00

92

Cli

mat

eri

sk2

20

.23

35

0.1

12

70

.10

04

0.0

89

90

.04

17

Law

nan

dg

ard

en

mo

istu

rein

dex

19

0.3

18

70

.25

64

0.2

33

50

.08

99

Dis

ease

risk

17

0.4

17

40

.38

15

0.2

07

1

Wil

dfi

reri

sk1

60

.50

00

0.2

45

7

Yie

ldri

sk1

60

.23

35

Liv

esto

ckh

eat

stre

ss

ind

ex

13

S. R. Templeton et al.

123

Page 9: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

Extensionists think that farmers are interested in using

forecasts to improve management of irrigation and planting

because, we hypothesize, the benefits to farmers of using

the forecast are likely to exceed the costs—time, money,

and mental effort of farmers—of learning about and

applying it. Our hypothesis is consistent with the concept

of a valuable forecast, i.e., one that generates incremental

benefits (Murphy 1993). A climate forecast is more likely

to be valuable, or beneficial, for a strategic, preseason

decision than for a tactical, during-season decision (Fraisse

et al. 2006). A farmer’s planning for seasonal irrigation

demand and scheduling of plantings are related strategic,

preseason decisions. The seasonal demand for irrigation

depends on seasonal rainfall and a crop’s growth stage,

which, in turn, depends on the planting date. In contrast,

only a minority of extension personnel thinks farmers

could use a climate forecast to improve marketing, labor

management, or waste management because, we hypothe-

size, they perceive, rightly or wrongly, that the activity

does not depend much on seasonal climate.

A freeze alert is also likely to be useful to a majority of

extension personnel in and beyond South Carolina. Freeze

alerts were the most frequently checked forecast, by size-

able margins, for usefulness to extension agents in Florida

(Cabrera et al. 2006) and North Carolina (Breuer et al.

2011). The high probabilities that freeze alerts could be

useful might reflect our survey having been conducted

during a relatively cold winter and the other two surveys

having been conducted in late fall and late winter. How-

ever, freezing of crops can cause substantial economic

damages. For example, a cold-air outbreak between April

6–10 and well-below-freezing temperatures on April 8,

2007 contributed to losses of approximately 79 % for

peach, 85 % for apple, and 39 % for winter-wheat harvests

and $39.3 million in farmgate revenues in South Carolina

(NOAA-USDA 2008).

Most South Carolina extension personnel also think that

a forecast of plant moisture stress could be useful to them,

although the proportion that thinks the forecast could be

useful is statistically less than the proportion that thinks a

freeze alert could be useful. Extension personnel could

teach farmers how to use a forecast of plant moisture stress

to schedule irrigation. Irrigation has become more impor-

tant for South Carolina’s agriculture. The area of irrigated

cropland in the state increased from 35,362 ha. in 1997 to

37,148 ha. in 2002 to 49,944 ha. in 2007 (NASS 2004,

2009). Irrigated area’s share of total harvested area

increased from 5.04 % in 1997 to 6.68 % in 2002 to

7.95 % in 2007. Participants in the AgroClimate workshop

in early 2011 reported that irrigation was still expanding in

the state.

A freeze alert and a forecast of plant moisture stress

could be useful to most extensionists because the forecasts

are short-term and, as such, are generally more accurate

than most climate forecasts. A short-term, or weather,

forecast can help an extensionist teach or advise farmers

when to take tactical actions—such as when to protect a

crop from imminent freeze damage or when next to irrigate

a crop for reduction in plant moisture stress—that create

immediately tangible benefits whereas some long-term, or

climate, forecasts are less likely to do so. In contrast,

Freeze Risk Maps ranked only sixth in potential usefulness

among AgroClimate’s decision support tools in the focus

groups (Fig. 1) for the possible reason that the county-level

maps could only help extension agents and farmers ascer-

tain, in light of the forecasted ENSO phase, the odds that

protection of crops from freeze damage during an

upcoming winter would be needed but not help them

decide when during the winter such action should be taken.

Conclusion

AgroClimate lacks decision support tools for irrigation

management and freeze protection in the Carolinas. A

freeze alert and a map of freeze probabilities, the displays

of which would vary with the ENSO phase, are also likely

to be useful to farmers of various crops, not just a majority

of extension personnel. The information to create such

tools already exists.

A tool that incorporates a forecast of plant moisture

stress for irrigation scheduling is also likely to be useful to

South Carolina’s farmers, not just extensionists. Most

irrigators rely on rules of thumb for scheduling or sched-

ulers that utilize data from neighboring states whose agro-

climatic conditions might not apply to the state (Farahani

et al. 2008). However, the lower and upper threshold

temperature functions for different crops in South Carolina

must first be determined through research if such sched-

ulers are to be developed (Farahani et al. 2008).

If probabilistic forecasts of the ENSO phase in an

upcoming spring and summer are to be useful for managing

irrigation and scheduling of planting, the forecasts need to

have sufficient accuracy, or quality (Murphy 1993), and the

impact of the ENSO phase on precipitation needs to be

sufficiently discernible. The scientific knowledge for such

accuracy and discernment in South Carolina might not yet

be adequate, however. The ENSO phase has been more

difficult to accurately predict for spring and summer than

other seasons in the southeastern USA (e.g., Barnston et al.

2012; Hansen et al. 1998). Also, although the ENSO signal

clearly affects the climate of the coastal plain of the Car-

olinas (Hansen et al. 1998), its effects on the climate of the

Sandhill and Piedmont regions are small or not yet well

understood. Moreover, recent research of Gail Wilkerson,

Professor of Crop Sciences at North Carolina State

Usefulness and uses of climate forecasts

123

Page 10: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

University, suggests that the impact of El Nino or La Nina

on rainfall in April and May might not be discernible in the

coastal plain of the Carolinas because factors other than

ENSO tend to dominate the inter-annual variability during

these months.

Our assessment might be limited by the scope and nature

of our data. Our focus groups and survey population

excluded South Carolina State University’s extension per-

sonnel, who represented about 3 % of the state’s exten-

sionists. Thus, our results and interpretations might not

apply to them and their clients. Most of their clients are

black or African-American farmers, who represented 7 %

of South Carolina’s farmers in 2007 (NASS 2009). More-

over, our survey data represent stated preferences about

forecasts, most of which have not yet been incorporated

into decision support tools for South Carolina. Stated

preferences might change as extension personnel learn

about climate forecasts that have already been incorporated

into decision support tools, such as Climate Risk.

Our conclusions are based primarily on the results of

statistical tests for majorities and differences in uncondi-

tional probabilities. However, respondents might have been

more enthusiastic about forecasts than non-respondents.

Moreover, the gender, age, and other characteristics of

extensionists might also affect their assessments of a

forecast (e.g., Breuer et al. 2011; Cabrera et al. 2006). For

example, if extensionists are younger or more computer

literate than others are, then the likelihood that a forecast is

potentially useful might be higher for these subpopulations

than for others. Analysis of conditional probabilities of

forecasts being useful or managerial activities being

improved by forecasts is important for future research.

Finally, a forecast that is likely to be useful or a managerial

activity that is likely to be improved by a climate forecast is

not necessarily a forecast or use of it that will be most bene-

ficial, on net, to farmers or society. For example, the expected

net benefits of an index of livestock heat stress, which is not be

regarded by many as potentially useful, might exceed the

expected net benefits of a forecast of climate risk. Estimation

of the discounted total benefits and costs of the forecasts to

farmers, consumers, and others would be required. None-

theless, if potentially widespread adoption matters to those

who develop decision support tools for extension, our focus-

group and survey results have provided a baseline of com-

plementary information about potential usefulness of fore-

casts and uses of climate forecasts.

Acknowledgments Our initial research was conducted under a

subcontract with Florida State University for the project ‘Decision

Support System for Reducing Agricultural Risks Caused by Climate

Variability,’ which was funded by the United States Department of

Agriculture’s Cooperative State Research, Educational, and Extension

Service. The paper was written and revised as part of a subcontract

with the University of Florida for the project ‘Climate Variability to

Climate Change: Extension Challenges and Opportunities in the

Southeast USA,’ which is funded by Competitive Grant No.

2011-67003-30346 from the USDA’s National Institute of Food and

Agriculture. We thank George Askew, Nelle Bridges, Teresa Kelley,

Eleanor Massey, and Steve Meadows for help with survey adminis-

tration; Kathryn Boys, Rebecca Davis, Patricia DeHond, Hamid Fa-

rahani, Bruce Fortnum, Mandy Stephan, and Jessica Whitehead for

help with workshop arrangements; and Ryan Boyles, Norman Breuer,

Wolfgang Cramer, Clyde Fraisse, Keith Ingram, Jim Jones, Vasu

Misra, Tom Mroz, Jim O’Brien, Gail Wilkerson, and two anonymous

reviewers for their comments. We are responsible for any remaining

errors. Aldridge and Bridges share third authorship.

References

Barnston AG, Tippett MK, L’Heureux ML, Li S, DeWitt DG (2012)

Skill of real-time seasonal ENSO model predictions during

2002–11: is our capability increasing? Bull Am Meteor Soc

93:631–651

Breuer NE, Adhikarim S, Brown-Salazar R, Clavijo JA, HansPetersen

HN, Kawa NC, Patarasuk R, Hildebrand PE (2008a) Extension

agent perspectives of climate, seasonal climate forecasts, and the

AgClimate decision support system. Southeast Climate Consortium

Technical Report Series: SECC-08-001, Gainesville FL, September

Breuer NE, Cabrera VE, Ingram KT, Broad K, Hildebrand PE

(2008b) AgClimate: a case study in participatory decision

support system development. Clim Change 87:385–403

Breuer NE, Dinon H, Boyles R, Wilkerson G (2011) Extension agent

awareness of climate and new directions for research in North

Carolina. J Service Climatol 5:1–20

Cabrera VE, Breuer NE, Bellow JG, Fraisse CW (2006) Extension

agent knowledge and perceptions of seasonal climate forecasts in

Florida. Southeast Climate Consortium Technical Report Series,

SECC Technical Report 06-001, Gainesville FL

Conover WJ (1999) Practical nonparametric statistics, 3rd edn. John

Wiley and Sons, New York

Crane TA, Roncoli C, Paz J, Breuer N, Broad K, Ingram KT,

Hoogenboom G (2010) Forecast skill and farmers’ skills:

seasonal climate forecasts and risk management in the south-

eastern United States. Weather Clim Soc 2:44–59

Dillman DA, Smyth JD, Christian LM (2009) Internet, mail, and

mixed-mode surveys: the tailored design method. John Wiley

and Sons, Hoboken NJ

Farahani H, Khalilian A, Smith WB (2008) Irrigation water manage-

ment in South Carolina – trends and needs. In: Proceedings of

the 2008 South Carolina Water Resources Conference, Clemson

University Restoration Institute and Clemson Center for

Watershed Excellence, North Charleston SC, October 14–15

Faravelli M (2007) How content matters: a survey based experiment

on distributive justice. J Public Econ 91:1399–1422

Fraisse CW, Breuer NE, Zierden D, Bellow JG, Paz J, Cabrera VE,

Garcia y Garcia A, Ingram KT, Hatch U, Hoogenboom G, Jones

JW, O’Brien JJ (2006) AgClimate: a climate forecast informa-

tion system for agricultural risk management in the southeastern

USA. Comput Electron in Agric 53:13–27

Hansen JW, Hodges AW, Jones JW (1998) ENSO influences on

agriculture in the southeastern United States. J Clim 11:404–411

JMP�, Ver. 9 (1989–2012) SAS Institute Inc., Cary, NC

Jones JW, Hansen JW, Royce FS, Messina CD (2000) Potential

benefits of climate forecasting to agriculture. Agric Ecosyst

Environ 82:169–184

McCown RL, Hochman Z, Carberry PS (2002) Probing the enigma of

the decision support system for farmers: learning from experi-

ence and from theory. Agric Syst 74:1–10

S. R. Templeton et al.

123

Page 11: Usefulness and uses of climate forecasts for agricultural extension in South Carolina, USA

Murphy AH (1993) What is a good forecast? An essay on the nature

of goodness in weather forecasting. Weather Forecast 8:281–293

Nadolnyak D, Vedenov D, Novak J (2008) Information value of

climate-based yield forecasts in selecting optimal crop insurance

coverage. Am J Agric Econ 90:1248–1255

NASS (2004) 2002 Census of Agriculture, South Carolina: State and

County Data, Volume 1, Geographic Area Series, Part 40, AC-

02-A-40. National Agricultural Statistics Service, U.S. Depart-

ment of Agriculture, issued June, Washington, DC

NASS (2009) 2007 Census of Agriculture, South Carolina: State and

County Data, Volume 1, Geographic Area Series, Part 40, AC-

07-A-40. National Agricultural Statistics Service, U.S. Depart-

ment of Agriculture, issued Feb. and updated Dec., Washington,

DC

NOAA-USDA (2008) The Easter Freeze of 2007: a climatological

perspective and assessment of impacts and services. National

Climatic Data Center’s Technical Report 2008-01, National

Oceanic and Atmospheric Administration and the US Depart-

ment of Agriculture

Nulty DD (2008) The adequacy of response rates to online and paper

surveys: what can be done? Assess Eval Higher Educ 43:301–314

Solow AR, Adams RM, Bryant KJ, Legler DM, O’Brien JJ, McCarl

BA, Nayda W, Weiher R (1998) The value of improved ENSO

prediction to US agriculture. Clim Change 39:47–60

Van Berckelaer AC, Mitra N, Pati S (2011) Predictors of well child

care adherence over time in a cohort of urban medicaid-eligible

infants. BMC Pediatrics 11:36

Usefulness and uses of climate forecasts

123