You are on page 1of 12

A Farm-Level Analysis of Economic

and Agronomic Impacts of Gradual


Climate Warming
Harry M. Kaiser, Susan J. Riha, Daniel S. Wilks,
David G. Rossiter, and Radha Sampath

The potential economic and agronomic impacts of gradual climate warming are
examined at the farm level. Three models of the relevant climatic, agronomic, and
economic processes are developed and linked to address climate change impacts and

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


agricultural adaptability. Several climate warming scenarios are analyzed, which vary in
severity. The results indicate that grain farmers in southern Minnesota can effectively
adapt to a gradually changing climate (warmer and either wetter or drier) by adopting
later maturing cultivars, changing crop mix, and altering the timing of field operations
to take advantage of a longer growing season resulting from climate warming.

Key words: discrete stochastic sequential prograrnming, gradual climate warming,


Minnesota.

This paper examines potential economic and ag- climate change is simulated a s a gradual and dy-
ronomic impacts of gradual climate warming at namic process rather than the comparative static
the farm level. Three models of the relevant cli- approach of comparing a "doubled COz" in-
matic, agronomic, and economic processes are duced change in climate with our present cli-
developed and linked to address climate change mate. Given the focus on farm-level adaptation
impacts and agricultural adaptability. While the issues, it is important that the climate, crop, and
model could be adapted and applied to any re- economic models be dynamic. Second, com-
gion of the United States, a grain farm in south- pared with most previous research, there is a
ern Minnesota is used here a s a case study. This greater emphasis on simulating the effects on
region is part of the northern fringe of the U.S. "tactical" farm-level decisions. For example, the
corn belt which could be affected by climate model allows for adaptive management strate-
warming. Several climate warming scenarios are gies such as changing plant cultivar (variety) se-
analyzed, which vary in severity to simulate how lection and changing planting and harvesting dates
sensitive crop yields, crop mix, and farm rev- in response to a gradually changing climate. Fi-
enue are to climate change. nally, changes in the variability, as well as in
This research differs from most previous the averages of climatic variables, are modeled.
analyses of climate change in three respects. First, Climate change may result in more than changes
in mean values for climatic variables, with
potentially important consequences (Katz and
Kaiser and Sampath ate associate professor and research support Brown).
specialist, respectively, in the Department of Agricultural Econom-
ics, Cornell University; Riha, Wilks, and Rossiter are associate
professor in agronomy, associate professor in atmospheric science,
and former research associate in agronomy, respectively, in the De- Previous Economic Findings
partment of Soil, Crop, and Atmospheric Sciences at Cornell Uni-
versity.
Funding for this research was provided by a cooperative agree- Although most research on climate change has
ment with the Resources and Technology Division of the Economic
Research Service, U.S. Department of Agriculture. The support
been done in the physical sciences, there is
and useful suggestions by John Reilly are appreciated. The authors growing interest in exploring the economic as-
also thank Richayd Boisvert, Deborah Streeter, and three anony- pects of climate change. Economists have fo-
mous reviewers for their constructive comments, and Jonell Blake-
ley for preparation of the manuscript. cused mainly on the benefits and costs of reducing
Review coordinated by Richard Adams. or slowing the rate of growth of atmospheric
Amer. J. Agr. Econ. 75 (May 1993): 387-398
Copyright 1993 American Agricultural Economics Association
388 May 1993 Amer. J. Agr. Econ.

concentrations of CO2 (e.g., Edmonds and Reilly; growth in total capacity. Kokoski and Smith, in
Nordhaus and Yohe). In a related vein, there primarily a methodological study, show that
has been analysis of economic issues corre- partial equilibrium welfare measures of climate
sponding to potential treaties among countries to change may provide reasonable approximations
reduce CO2 and other greenhouse gas emissions of true welfare changes particularly when the
(e.g., Chapman and Drennen). Less economic magnitude of change is large. Lewandrowski and
research has been devoted to exploring impacts Brazee discuss how fama programs could be-
of a changing climate on vulnerable economic come costlier under climate change and may play
sectors such as agriculture.1 an important role in the farm adaptation pro-
There are, however, some notable excep- cess.
tions. Several studies empirically measure po- While previous studies have shown that cli-
tential impacts of climate change on agricultural mate change has the potential to substantially
supply, demand, prices, social welfare, and other alter prices, production, agricultural compara-
market variables at various levels of geographic tive advantage, trade, and welfare distributional
effects, questions still remain. Additional re-

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


aggregation. At the world level, Kane, Reilly,
and Bucklin examine potential climate change search is needed on many issues including the
impact on world agriculture due to a doubling linkage of climatic, agronomic, and economic
of COz atmospheric concentrations, concluding models, and on farm adaptability issues, which
that world prices for com, soybeans, and other are the focus of this article.
grains would increase significantly due to de-
creases in world production. At the national level,
Adams et al. (1990) estimate that prices, re- A Model of Impacts of Climate Change on
gional production, and welfare impacts vary the Agricultural Sector
widely depending upon which climate scenario
The model used here consists of three compo-
is used and whether or not a CO2 "fertilizer ef-
nents: atmospheric, agronomic, and economic.
fect" is assumed. In a study of the prairie region
of Canada, Arthur and Abizadeh conclude that The atmospheric component simulates daily val-
ues for minimum and maximum temperature,
revenue gains and losses to farmers are sensitive
precipitation, and solar radiation. Based on the
to the assumed climate change scenario and re-
values of the climatic variables, the agronomic
gion being examined. In the midwest, Easterly
et al. look at potential climate change impacts component estimates crop yields, grain moisture
content, and field time availability (i.e., the span
on Missouri, Iowa, Nebraska, and Kansas ag-
of time during which weather-related soil mois-
riculture, generally finding reductions in grain
yields (excluding wheat) and faster depletion of ture conditions allow farmers to perform field
operations). In turn, crop yields, grain moisture
ground water sources in the scena¡ examined.
Onal and Fang conclude that climate change will content, and field time availability become in-
puts in the farrn level economic component. Crop
cause corn production to decrease in Illinois ac-
prices, which also are a function of climate sce-
companied by a northern shift in wheat and soy-
nario, are generated by price-reduced-form
bean production. In a study of the western U.S.,
equations based on supply and demand vari-
Adams et a1.(1988) argue that climate change
ables. Finally, the output of the economic model
will not result in a major loss in agricultural pro-
includes optimal crop mix, scheduling of field
duction, but there will be adjustments in re-
operations, and expected net farm income.
gional production and resource use.
There have also been several conceptual stud-
ies of climate change and agriculture. Sonka and Atmospheric Component
Lamb present a conceptual model for analyzing
the impacts of climate change at the farm and The atmospheric component is a stochastic
sector levels, recommending the need to study weather generator based on Richardson's model.
this process within a dynamic, or evolving, con- It requires that daily historical weather data be
text instead of a comparative static framework. characterized in terms of the parameters of a
Crosson argues that climate change would alter particular stochastic process which describes daily
world and regional agricultural comparative ad- weather variations. Stochastic weather out-
vantage, but would not likely seriously limit comes consistent with present-day climates are
produced using stochastic process parameters fit
For a more detailed discussion of some of these studies, see
to observed weather data. Forecasts of changed
Kaiser (1991). climates are achieved through adjustments to the
Kaiser, Riha, Wilks, Rossiter, and Sampath Farm-Level Effects of Climate Warming 389

parameters of the daily weather process consis- a changing climate and to predict other aspects
tent with specified changes in mean values and of the cropping system including grain moisture
aspects of the variability (Wilks). Climate content and field time availability. While rela-
changes are specified in terms of monthly (rather tively complex, such models are generally pre-
than daily) means and variances, since this is ferred over simpler alternatives for climate impact
the level of information available from general assessments (Sonka and Lamb) since problems
circulation models (GCM) integrations. The link of extrapolation beyond developmental data are
between monthly statistics and parameters of the less severe, and mechanisms of interaction with
daily stochastic process is made using the sam- the environment can be examined explicitly.
pling distributions of sums of the daily values The crop simulation software used (General
(Katz; Wilks). purpose Atmosphere Plant Soil or GAPS) was
Four climate change scenarios are generated developed by Buttler and Riha. This model pro-
in order to simulate potential crop and economic vides for flexible simulation along the lines of
responses. Scenario 1 (baseline) reflects no cli- similar efforts by crop physiologists in order to
mate change and employs the unmodified pa- model crop growth in the field environment (e.g.,

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


rameters of the stochastic weather model. The van Keulen and Wolf). GAPS allows different
changing climate scenarios are developed from representations of the physical environment to
plausible, but arbitrary, estimates of the evolu- be interfaced with different crop models. Con-
tion of the climate system in the medium term sequently, the degree to which particular results
(50 to 100 years). Scenarios 2 and 3 are based depend upon particular crop or environmental
on the globally averaged temperature changes component models can be analyzed. 2
portrayed in the transient "scenario B" of Han- Increasing atmospheric concentrations of COz
sen et al. Here the temperature changes are rel- should have an enhancing effect on crop yields.
atively mild, amounting to 2.5~ at the time of However, there is currently debate among plant
equivalent CO2 doubling (year 2060), with half scientists over the magnitude of this enhance-
the warming occurring during 2030-2060. In ment, or how this effect might interact with other
scenario 2, the average monthly precipitation is environmental or plant physiological condi-
increased linearly at a rate sufficient to increase tions. In addition, not all of the enhanced ra-
this quantity by 10% in 2060. In scenario 3, av- diative forcing will be due to CO2. Other green-
erage precipitation is linearly decreased through house gases r e p r e s e n t a significant part of
10% in 2060. Scenario 4 is based on the Hansen anticipated enhanced radiative forcing, and these
et al. "Scenario A," exhibiting a 4.2~ increase gases will have no direct effect on crop physi-
by 2060, again with half of this increase occur- ology. Given these factors, we assume in this
ring from 2030 to 2060. Average precipitation study that there is no enhancing effect on crop
in scenario 4 decreases linearly by 20% of pres- yields due to increasing greenhouse gas emis-
ent levels by 2060. sions in the atmosphere. Tbis is a simplifying
Imposed variability changes are based on the assumption that will need to be improved upon
time-dependent mean changes. Standard devia- once more knowledge is gained about the po-
tions of monthly average temperature are lin- tential positive effect of increased CO2 emis-
early decreased with increases in average monthly sions.
temperature, and these changes are reflected in
the variances of the daily temperatures. This is
consistent with the limited investigations of Economic Component
changes in temperature variability in GCM sim-
ulations (e.g., Rind et al.; Wilson and Mitch- As the climate changes, farmers will be forced
ell). The variance of monthly precipitation is as- to re-evaluate their production decisions, and in
sumed to follow a power-law relationship with
average monthly precipitation (Waggoner); it 2 Presently included in the GAPS modeling framework are maize
increases in scenario 2 and decreases in scenar- (com) and sorghum. The com model in GAPS was initially de-
veloped by Stockle and Campbell, and modified to include grain
ios 3 and 4. yield. The sorghum crop model in GAPS is SORKAM, developed
by Rosenthal et al. Sorghum is included because it exhibits superior
tolerance to heat and drought, and thus has the potential to displace
other crops in a warming environment. SORKAM was modified
Agronomic Component and incorporated into GAPS so the model maize and sorghum crops
would share the same simulated soil, water, and evapotranspiration
modules. While soybeans are not currently part of the GAPS en-
Process-oriented models of crop-environmental vironment, a soybean crop model (SOYGROW) developed by
interactions are used to predict crop response to Wilkerson et al. is used here.
390 May 1993 Amer. J. Agr. Econ.

Table 1. C a l e n d a r of Production Operations

Plant Harvest

Stage/Period Plow Com Soybeans Sorghum Corn Soybeans Sorghum

1 Apr 7 - A p r 22 X
2 Apr 2 3 - M a y 11 X X X X
3 May 1 2 - M a y 31 X X X X
4 Jun l - J u n 8 X X
5 Sep 15-Sep 30 X X
6 Oct 1-Oct 16 X X X X
7 Oct 17-Oct 31 X X X
8 Nov 1 - N o v 30 X

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


particular, the mix of crops. This study simu- knowledge of which stage 1 state has occurred,
lates future decisions for each climate scenario but only probabilistic knowledge of which stage
using discrete stochastic sequential program- 2 state will occur. The stage 2 states (ten states,
ming (Cocks; Rae). Discrete stochastic sequen- conditional on each of the three stage 1 states)
tial programming (DSSP) is a mathematical consist of discrete random parameters for field
programming technique that treats the decision time availability, crop yields, grain drying costs,
process as a multi-stage, sequential process. and crop prices. Each of these states consists of
Consequently, decisions made at any particular field hours available in each of the four stage 2
point in time depend upon decisions made in periods, crop yields and drying costs associated
previous stages and outcomes of previous ran- with each of six planting and harvest dates, and
dom events, or states of nature (Kaiser and output prices for the three crops. Again, it is
Apland). assumed that the farmer expects each of the stage
The economic model divides the decision- 2 states to be equally likely. There are 30 joint
making process into two stages: stage 1 (pre- net revenue events possible, each corresponding
harvest) and stage 2 (harvest). Stage 1 decisions to a unique sequence of a stage 1 state followed
include spring plowing and planting operations, by a stage 2 state.
which can take place in four periods (see table Values for the field time availability, yield,
1).3 Stage 2 decisions include fall plowing and and crop moisture states of nature are generated
harvesting, which can take place in four periods by GAPS. Grain drying cost states of nature are
(table 1). The constraining resources for both determined by grain moisture content at harvest,
stages include full and part-time labor by pro- yield level, and costs per bushel per percentage
duction period and crop acreage. Risk is cap- point of moisture removed using the following
tured by a Freund-type objective function which formula:
maximizes expected net revenue minus a risk
term adjusted by a risk aversion coefficient. DCi =- 0.024 Yi (Mi - 17) I17i
Four important sources of risk are included in
the model: field time availability, crop yields, where DCi is drying cost per acre, harvest state
grain drying costs, and crop prices. At the be- i; Yi is yield per acre, harvest state i; M; is grain
ginning of stage 1, the farmer makes spfing moisture content, harvest state i, and I17i is equal
plowing and planting decisions while facing three to one if the moisture content in harvest state i
states of nature on field time availability. Each is greater than 17%, or equal to zero otherwise.
of these states differs by field time availability It is assumed that it costs $0.024 per bushel to
in each of the four stage 1 periods. It is assumed remove one percentage point of moisture, and
that the farmer expects each of the three stage that grain must be dried only if it has a moisture
1 states to be equally likely. At the beginning content of 17% or higher. Values for each crop's
of stage 2, the farmer makes harvest and fall price states of nature are generated by a price
plowing decisions. Here the farmer has perfect reduced-form equation of supply and demand
which includes crop yield as an explanatory
3 Post-plant cultivation, fertilization, and pesticide operations ate variable (procedures are discussed in the next
not included as activities in the model because they are not time- section).
limiting. Rather, the variable costs and technical coefficients as-
sociated with these operations are incorporated into the planning The mathematical formulation of the eco-
activities. nomic model is the following:
Kaiser, Riha, Wilks, Rossiter, and Sampath Farm-Level Effects of Climate Warming 391

3 lO C l l , C12 , C22 = v a r i a b l e cost vectors for


(1) maximize: E - r@ E E d~j spring plowing, planting, and fall plowing
i= 1 j=l Xll, XI2 = spring plowing and planting
vectors, preharvest stage
subject to the following constraints: C21ij = variable cost vector for harvest,
preharvest state i, harvest state j
X2~i, X22i = harvest and fall plowing vec-
Accounting Constraints tors, preharvest state i
P = output price vector
(2) Rgj -~- C l l X l l -~- Cl2X12 -F- C21ª MI = marketing decision vector, prehar-
+ C22X22i - PMi = 0
vest state i
( i = 1 . . 3 ; j - - 1 . . . 10)
o~i = probability of preharvest state i oc-
3 lO
curring
(3) E E afljR,j - E = 0 /3j -- probability of harvest state j occur-
i=l j=l

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


ring, given preharvest state i
(4) R~j - E - d~j -< 0
all, al2 a 2 1 , a 2 2 = matrices of resource
requirements for all field operations in stages 1
(i= 1...3;j= 1 . . . 10)
and 2
bl; = vector of stage 1 resource endow-
ments, preharvest state i
Resource Constraints
LI = vector of ones
b 2 = total crop land endowment
(5) a~~X]r -+- a12X12 "< bli
b3ij = vector of stage 2 resource endow-
(6) LlX12 -< b 2 ments, preharvest state i, harvest state j
I'Iij = vector of crop yields, preharvest
(7) a21X21 i qt_ a22X22 i ~ b3ij state i, harvest state j
(i= 1...3;j= 1 . . . 10)
B~, . . . B7 = sequence preserving matri-
(8) -HijX21i + Mi <- 0 ces for field operations
(i= 1...3;j= 1... 10) I = identity matrix

The objective function in (1) is to maximize


Sequencing Constraints expected net revenue (gross revenue minus vari-
able costs) minus a risk adjustment term, where
(9) --B1Xll + B2X12 -< 0 risk is measured as the standard deviation of net
revenue calculated by the product of total ab-
(10) --/Xi2 + B 3 X21 i ~ 0 solute negative deviations from the mean times
(i = 1 . . . 3) the parameter q) (Hazell). The risk measurement
(11) is discounted by the risk aversion coefficient, r.
- B 4 X 2 1 i q- BsX22 i ~ 0
(i = 1 . . . 3) Constraints (2) through (4) are accounting con-
3
straints which define the 30 joint net revenue
events (2), define expected net revenue (3), and
(12) BTXll -- E aiB6X22i "< 0 define negative deviation from expected net
i=1
revenue (4).
(i = 1 . . . 3) Constraint (5) restricts the use of farm labor
(13) Rij, dij, Xi1, Xi2, X21i, X22i, M i :> 0 (full- and part-time) by field operations in the
pre-harvest stage to endowed levels. Note that
where E = expected net revenue the right-hand-side parameter in this constraint
r = risk aversion coefficient is stochastic, corresponding to available field
= constant that converts total negative hours by production period for the three stage 1
deviations into proxy for standard deviation states of nature. Constraint (6) is the land con-
(2/s)H,/2(s - 1)05, where H = 2 2 / 7 and s is straint that limits acres planted to endowed ley-
sample size els. Constraint (7) restricts the use of farm labor
dij = negative deviation from expected by field operations in the harvest stage to en-
net revenue, preharvest state i, harvest state j dowed levels. The right-hand-side parameters
Rij = total net revenue preharvest state i, for this constraint are also stochastic, corre-
harvest state j sponding to the available field hours by pro-
392 May 1993 Amer. J. Agr. Econ.

duction period for the 30 stage 2 states of na- vesting combination, three cultivars are simu-
ture. Crop output constraints are represented by lated: early, mid, and late maturing varieties.
(8), which limits the amount of crop that can be However, only one cultivar is used in the eco-
sold to the amount that is harvested for each pre- nomic model for each planting-harvesting com-
harvest state of nature. Finally, constraints (9) bination for each decade based on the following
through (12) are sequencing restrictions which decision rule. It is assumed that farmers make
preserve the proper sequence of field operations cultivar decisions on the basis of yield perfor-
in the model. These constraints guarantee that mance in the previous decade. Specifically, the
sp¡ plowing occurs prior to planting (9), that cultivar having the highest average yield for a
planting activities are matched with harvest ac- particular plant-harvest period in the previous
tivities (10), that harvesting occurs before fall decade is selected by the farmer for the current
plowing for each stage 1 state (11), and that any decade.
acreage not plowed in the fall is plowed in the To generate the 30 joint events (crop yields,
spring (12). grain moisture content, and field hours) for the

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


economic component, the weather component
produces 30 realizations (years) of daily data
Simulation Procedures representative of each decade for the agronomic
component. Based on these 30 weather realiza-
The simulation procedures begin with a Monte tions, the agronomic component then produces
Carlo simulation of the stochastic weather model 30 yield, field time availability, and grain mois-
to generate daily weather values for climates ture states of nature per decade for the economic
(scenarios) changing over the 100 year period model. Since these 30 joint states of nature are
1980-2079. These results are used by the ag- assumed to be equally representative of a de-
ronomic model to generate annual values for crop cade, the solution to the economic model can be
yields, grain moisture, and field time availabil- thought of a s a representative year within the
ity. 4 These values are tabulated decade-by- decade. This process is repeated ten times for
decade, yielding ten sets of agronomic results each climate change scenario to generate solu-
for each scenario. Finally, the resulting crop tions for each decade for 1980-2070.
yields, grain moisture, and field time availabil- Because this is a microlevel model, crop prices
ity parameters are used by the farm level eco- cannot be endogenously determined. Yet it is
nomic model, which generates optimal manage- unlikely that climate change would not affect crop
ment strategies and expected net revenue. Costs, prices over time. To construct price trajectories
technical parameters, and resource endowments by decade for each crop based on each climate
are held constant at their 1980 values. However, scenario, price reduced-form equations are es-
cultivar selection, crop yields, grain moisture timated using time series data from 1960 through
content, and grain drying costs are different for 1988. In addition to several exogenous demand
each decade of each climate change scenario, and supply shifters, county average crop yield
according to results of the agronomic simula- is included as an explanatory variable in the price
tions. reduced-form equation. In theory, crop yield at
For each crop and each possible planting-har- the micro-level should not explain output price,
but Minnesota yields are highly correlated with
4 The crop model generates these values disaggregated by time
period. For example, for each crop six yields are generated--one
national yields which do influence price (cor-
for each combination of planting and harvest periods. Likewise, six relation coefficients for corn and soybeans be-
grain-moisture-content values are generated for each crop--one for tween county and national yields are 0.84 and
each combination of planting and harvesting periods. Field time
availability output from the crop model is generated in the form of 0.75, respectively). The three estimated crop
available field hours for each of the eight production periods. price reduced-form equations are

lnPC,=- 46.33 - 0.44 l n Y C , + 9.11 lnPOP,+ 0.34 lnPC,_~ + 0.59 lnSPC,+ 0.52
(26.5) (0.21) (5.05) (0.15) (0.29) (0.09)
DUM73-75 - 0.08 Tt
(0.05)
R 2 = 0.87 D.W. -- 1.81
lnPSB,= 0.90 - 0.33 ln YSB,+ 0.16 lnPOP,+ 0.53 lnPSBt_~ + 0.10 lnSPSBt
(7.10) (0.33) (1.34) (0.17) (0.39)
+ 0.29 DUM73_75
(0.17)
Kaiser, Riha, Wilks, Rossiter, and Sampath Farm-Level Effects of Climate Warming 393

R 2 = 0.68 D.W. = 2.33


ln PSG, = 3.32 -0.99 In YSG, + 0.07 In lNC + 0.20 ln PSG,_I + 0.56 ln SPSG,
(3.19) (0.34) (0.34) (0.16) (0.23)
+ 0.47 DUM73-75
(0.10)
R 2 = 0.88 D.W. = 1.61

where PC, = real com price per bushel (nomi- to illustrate the type of analyses that can be con-
nal price divided by Consumer Price Index 1988 ducted using this model. This region is the
= 1.0), year t, northem limit of com production in southern
YC, = county average corn yield, year t, Minnesota (Murray). The predominant soil se-
POP, = U.S. civilian population, year t, ries in this region is Ves, which is deep and has
SPC, = real com support p¡ per bushel, excellent water holding capacity.
year t, Most of the data for the economic component
DUM73_75 = dummy variable equal to 1

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


for the hypothetical Minnesota farm not gener-
for 1973-75, equal to zero otherwise, ated by GAPS are presented in Kaiser (1985),
7", = time trend, 1960 = 1, 1961 = 2, including resource requirements and variable
costs. It is assumed that the farm is endowed
PSB, = real soybean price per bushel, with 600 acres of tillable land on which com,
year t, soybeans, and/or sorghum can be grown. There
YSB, = county average soybean yield, are two full-time workers and one additional part-
year t, time worker that can be hired at a cost of $6.00
SPSB, = real soybean support price per per hour. The farm uses a conventional tillage
bushel, year t, system. Data for sorghum, which currently is
PSG, = real sorghum support price per not commonly grown in this region, are based
bushel, year t, on national average statistics.
YSG, = national average sorghum yield, The yield data for the three crops are gener-
year t, ated by GAPS and SOYGROW. Because GAPS
INC = U.S. per capita real income, and SOYGROW simulate potential rather than
year t, actual farm yields, all yields are normalized to
SPSG, = real sorghum support price per 1980-88 county averages by multiplying each
bushel, year t, GAPS (or SOYGROW) predicted yield by the
R 2 = coefficient of variation, ratio of the average county yield for the 1980s
D.W. = Durbin-Watson statistic, and to the average GAPS (or SOYGROW) potential
( ) = standard error. yield simulated for the 1980s. While GAPS pre-
dicted com yields are higher than actual yields,
To simulate crop prices from 1990-2079, fu- the predicted soybean and sorghum yields are
ture values for the exogenous variables are nec- quite close to actual yields.
essary. It is assumed that U.S. civilian popu- Field hours by production period and state of
lation increases by 1% per year throughout the nature are also generated by GAPS. Thirty sets
simulation period. Real support prices for the of values for field hours are generated to rep-
three crops are assumed to decrease by 1% per resent the 30 joint states of nature for each de-
year, reflecting a trend towards a market-ori- cade. All 30 values are used for the four stage
ented farm policy. Real income is forecast based 2 periods. However, because stage 1 requires
on a regression equation with income in the two only three states of nature for field hours, the
previous years a n d a time trend as explanatory following values are used to represent the three
variables. For each decade, the price states of stage 1 states: average field hours minus one
nature are generated by substituting the annual standard deviation, average field hours, and av-
average yield states of nature, generated by erage field hours plus one standard deviation.
GAPS, and the values for the other exogenous Several adjustments were made after compar-
variables into the price reduced-form equations. ing actual observations for the 1980s with pre-
dicted values. First, GAPS over-predicted field
hours in the first period, consequently, subse-
Results for a Southern Minnesota Farm quent period 1 field hours were adjusted by mul-
tiplying them by the ratio of actual to predicted
A hypothetical farm based on characteristics of based on average observations for the 1980s.
southern Minnesota (Redwood County) is used Second, GAPS generated field hours for stage
394 May 1993 Amer. J. Agr. Econ.

1 and stage 2 periods were reduced by 30% and water-stress because of the excellent water hold-
10%, respectively, to prevent all stage 1 oper- ing capacity of this soil and the wetter climate.
ations from being completed before June, which The percentage change in average crop yields
is unrealistic for this region. GAPS also over- from 1980 under the mildly warmer and drier
predicted grain moisture contents for com and climate scenario 3 is presented in figure 1 (mid-
sorghum. Using actual 1980s observations to dle graph). As was the case for the wetter cli-
calibrate, drying costs based on GAPS moisture mate scenario, this drier climate scenario ap-
contents were reduced by 35%. Similar adjust- pears to have no adverse effects on soybean or
ments in both field hours and moisture contents sorghum yields. However, corn yields decrease
were required in Kaiser's (1985) study of this marginally over time. In this case the drier cli-
region. mate has two impacts on crop yields. First, the
growing season is lengthened due in part to the
drier climate, allowing greater flexibility in ac-
Climate Change Impacts on Crop Yields cess to the field. This has a positive effect since
later maturing cultivars can be adopted. Second,

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


The climatic, agronomic, and economic com- the decrease in precipitation causes some water
ponents are solved for the four climate scenar- stress, which has a negative effect on yields.
ios. Recall that scenario 1 is the no-climate- The percentage change in average crop yields
change-situation, scenario 2 is the mildly warmer from 1980 for the most severe climate change
(2.5~ and wetter case (10%), scenario 3 is the scenario is presented at the bottom of figure 1.
mildly warmer (2.5~ and drier (10%) situa- In this scenario, the average decrease in precip-
tion, and scenario 4 is about twice as warm and itation is twice as large as the previous scenario
dry as scenario 3. The yield states of nature, and the increase in temperature is 4.2~ rather
averaged by decade for presentation purposes than 2.5~ in year 2060. Average soybean and
(1980-2079) for climate scenarios 2, 3, and 4, sorghum yields are not adversely impacted by
are presented in figure 1.5 This figure displays even this relatively severe change in climate.
average crop yields over the simulation period Average com yields, on the other hand, trend
a s a percent of the 1980s average for each cli- downward. Moreover, the magnitude of de-
mate scenario. crease is larger than the previous scenario.
The mildly warmer and wetter scenario 2 has Two observations emerge from these results.
no adverse impact on crop yields at this rela- First, the model predicts that soybean and
tively cool location (top graph, figure 1). While sorghum yields at this relatively cool location
the climate is gradually warming, the accom- are not adversely affected by these three climate
panying increase in precipitation prevents the change scenarios. Second, com yields are some-
crops from experiencing water stress for the what adversely affected by the more severe
simulation period. In fact, sorghum and soybean warmer and drier scenarios, but not by the mild
yields increase over time, while com yields re- warmer and wetter scenario. At this location,
main relatively stable, increasing slightly from com appears to be the most climate-sensitive of
2000-2060. For all three crops, the model pre- the three crops.
dicts adoption of later-maturing, higher-yielding
cultivars over time as field time availability in-
creases because of climate warming. The ro- Climate Change Impacts on Crop Mix and
bustness of yields to climate change in this sce- Profitability
nario is a result of the relatively mild change in
climate assumed, the relatively cool location, the The crop price states of nature, averaged by de-
introduction of later-maturing cultivars later on cade, for each of the three changed climate sit-
in the simulation period, and the absence of plant uations are presented in figure 2. In all three
scenarios, the real com price is increasing, real
sorghum price is decreasing, and the real soy-
5 The crop yield averages in this figure are weighted averages of bean price is relatively constant. The average
yields over the six possible planting-harvesting dates. The weights prices are quite similar between scenarios 2 and
are using the results of the economic model, which give the optimal
proportion of planting and harvesting dates. For example, if the
3 because yields are comparable between the two
economic results for 1980 indicated that 75% of the com acreage cases. However, the real com price path is higher
should be planted in period 1 and harvested in period 7, then the in the most severe climate scenario compared to
average crop yield over the 30 states for planting period l, har-
vesting period 7 should comprise 75% of the overall weighted av- scenarios 2 and 3 because com yields are lower
erage for the 1980 decade. in climate scenario 4.
Kaiser, Riha, Wilks, Rossiter, and Sampath Farm-Level Effects of Climate Warming 395

Clirnate Scenario 2 ( M i l d l y Warmer and Mildly Wetter)

120

115

~ 110
~ ~
== 1o5.&~._~..=./ -'~. .J- \ ~ = ~ Com

100 / -~." ". '~ Soyoeans


,,i..
o 95
~" ~ Sorghum
"~ 90

no 65
80t I 1 1 1 ,= ;
1980 1990 2000 2010 2020 2030 2040 2050 2060 2070
Decade

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


Climate Scenario 3 ( M i l d l y Warmer and Mildly Drier)

120

115

~ 110
~176
o 105 ~ ' ~ Com
o
,,- 100 Soybeans
"8 95
~ ~ Sorghum
~Q 9o
L.
o
D. 85
80 I t I [ ..= | ,I . ; ;
1980 1990 2000 2010 2020 2030 2040 2050 2060 2070
Dec=de

Climate Scenario 4 (Significantly Warmer and Drier)

120

115
P
:~ 110

o 105 ~= "' Com

,P 100 -----a Soyoeans


"6 95
~ ' ~ Sorghum
~ 9o
85

80 I I I I I I I I t
198( 1990 2000 2010 2020 2030 2040 2050 2060 2070
Decade

Figure 1. Percentage change in average crop yieids, climate scenarios 2, 3, and 4, 1980-
2070

For the risk-neutral case, expected net farm revenue between scenario 1 and the three cli-
revenue for the three climate change scenarios mate change scenarios tends to widen in later
is shown in figure 3. In the no-climate-change decades of the simulation. The optimal crop mix
scenario (scenario i), average net revenue de- changes little over time for all three scenarios
clines slightly from 1980 to 2000, but thereafter with corn acreage representing about 70% of to-
consistently increases because of increases in real tal acreage planted and soybean acreage com-
corn prices (caused by increases in population). prising the rest of the acreage. Under none of
For all three climate change scenarios, net rev- these scenarios does sorghum become profitable
enue is somewhat higher than the no-climate- enough to replace com production.
change case. Moreover, the difference in net The standard deviation of net farm revenue is
396 May 1993 Amer. J. Agr. Econ.

Climate Scenario 2 (Mildly Warmer and Mildly Wetter)

6.50
6.00
5.50
5.00
4.50 - - - ' - - Com
4.00
3.50 Soybeans
3.00
2.50
2.00 --'-- Sorghum
1.50
1.00
0.50
0.00 I I I I I I I I I
1980 1990 2000 2010 2020 2030 2040 2050 2060 2070
Decade

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


Climate Scenario 3 (Mildly Warmer and Mildly Drier)

6.50
6.00
5.50
5.00
4.50 --=-- Com
4.00
3.50 Soybeans
3.00
2.50
~ . ~ u ~ " / '-- Sorghum
2.00
1.50 ~
~ =""--=- .=.=.=.. 9
1.00
0.50
0.00 I I I I I I I I I
1 980 1990 2000 2010 2020 2030 2040 2050 2060 2070
Decade

Climate Scenario 4 (Significantly Warmer and Drier)

6.50 ~
6.00 t
5.50 ,,,~=J

4.50 --" Com


4.00
3.50 ----o Soybeans
$ 3"00
I,...,-""'~./2.50 --'--Sorghum
2.00
1.50
1.00
i ~ ~'--" ~ ~. . . . .
0.50
0.00 / I I I I I I : : :
1980 1990 2000 2010 2020 2030 2040 2050 2060 2070
Decade

Figure 2. Average crop-price states of nature, climate scenarios 2, 3, and 4, 1980-2070

lower for the two mi]d climate change scenarios warmer, 20% drier climate (scenario 4) results
(scenarios 2 and 3) relative to no-climate change in higher fluctuations in revenue risk. Under this
for all decades. The increased stability of net scenario, revenue risk is higher in five decades
farm revenue is due to a decline in yield vari- and lower in four decades, compared to the no-
ability for all crops compared to the baseline no- climate change scenario.
climate change case. Crop yield variability, in To represent a risk-averse solution, a risk-
turn, is lower than in the no-climate change sce- aversion coefficient of 1.25 is used, which is in
nario because of the assumed decrease in the the range Brink and McCarl found representa-
variance of temperature. However, the 4.2 ~ tive for Cornbelt farmers. In this case, less com
Kaiser, Riha, Wilks, Rossiter, and Sampath Farm-Level Effects of Climate Warming 397

160000

140000

120000 i
~ I ~=~Scenario 1
100000

I
~~J "----'0 Scenario 2
80000
~ * Scenario 3
60000
O Scenario 4
40000

20000

0 I I I I ! ! ! I I
1980 1990 2000 2010 2020 2030 2040 2050 2060 2070

Figure 3. Average net revenue for the risk-neutral case, climate scenarios 1-4, 1980-2070

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


and more soybeans are grown as compared to changing climate scenarios considered in this
the risk neutral case. The difference is due to study (warmer and either wetter or drier cli-
soybeans being less risky than corn in terms of mates). Adaptive strategies include adopting later
net revenue va¡ However, under the risk- matu¡ cultivars, changing crop mix, and al-
averse case the share of corn as a percentage of tering the timing of field operations to take
total acreage increases over time under all three advantage of a longer growing season due to
scenarios, perhaps due to the accompanying de- climate warming.
crease in variability of com yields for the cli- There are several shortcomings of this anal-
mate change scenario. Sorghum, which is the ysis. First, the assumption that increased at-
most stable of all three crop yields, is still not mospheric concentrations of CO2 will have no
grown under any scenario. Expected revenue is effect on crop yields is questionable. Additional
lower in the risk-averse case than in the risk- research on the magnitude of this enhancement
neutral case because of the positive trade off be- will be beneficial. Second, the reliability of cli-
tween risk and income. mate models needs to be improved. Currently,
there is no consensus as to the magnitude of
change in climatic variables due to increases in
Summary greenhouse gases. Third, this analysis has ig-
nored the role of technological progress in mit-
Climate change will affect the agricultural eco- igating some or all of the potential negative ef-
system in a variety of ways. The growth and fects of climate change on agriculture. Clearly
yield of crops will change, as well as the periods technological innovation plays a major role in
of time available for field operations. These shaping agricultural production. Finally, the
changes will be reflected in decisions made by manner in which the price states of nature are
farmers, and more broadly, in changes in crop generated in the analysis needs improvement.
production, geography, and economics. This Linking micro models with sector-level models
paper has reported the results of a farm-level could improve the process of generating price
analysis to address the effects of climate change paths for climate change scenarios.
on farm operations and profitability.
[Received October 1991. Final revision
The results presented in this paper are not in-
received May 1992.]
tended a s a basis for general conclusions about
climate change and agriculture across the na-
tion. For example, while the relatively cool lo-
References
cation chosen for illustration generally benefits
from warming climates, more southern loca-
Adams, R. M., C. Rosenzweig, R. M. Peart, J. T. Ritchie,
tions could suffer considerable declines in yields
B. A. McCarl, J. D. Glyer, R. B. Curry, J. W. Jones,
and revenue. Instead the results ate intended to K. J. Boote, and L. H. Allen, Jr. "Global Climate
illustrate the importance of adaptive strategies in Change and U.S. Agriculture." Nature 345(1990):219-
predicting outcomes. The results indicate that 24.
grain farmers in the southern region of Minne- Adams, R. M., B. A. McCarl, D. J. Dudek, and J. D.
sota can effectively adapt to the gradually Glyer. "Imptications of Global Climate Change for
398 May 1993 Amer. J. Agr. Econ.

Hazell, P. B. R. "A Linear Alternative to Quadratic and tion, Texas A&M University, 1989.
Semivariance Programming for Farm Planning Under Sonka, S. T., and P. J. Lamb. "On Climate Change and
Uncertainty." Amer. J. Agr. Econ. 53(1971):153-62. Economic Analysis." Climatic Change 11 (1987):291 -
Kaiser, H. M. An Analysis of Farm Commodity Programs 311.
as Risk Management Strategies for Minnesota Com and Stockle, C., and G. S. Campbell. "A Simulation Model for
Soybean Producers. Ph.D. Dissertation, University of Predicting Effect of Water Stress on Yield: An Ex-
Minnesota, 1985. ample Using Com." Advances in lrrigation
Kaiser, H. M. "Climate Change and Agriculture." North- 3(1985):283-311.
east. J. Agr. and Res. Econ. 20(1991):152-63. Waggoner, P. E. "Anticipating the Frequency Distribution
Kaiser, H. M., and J. Apland. "DSSP: A Model of Pro- of Precipitation if Climate Change Alters its Mean."
duction and Marketing Decisions on a Midwestern Crop Agricultural and Forest Meteorology 47(1989):321-37.
Farm. North Cent. J. Agr. Econ. 11(1989):157-69. Wilkerson, G. G., J. W. Jones, K. J. Boote, K. T. Ingram,
Kane, S., J. Reilly, and R. Bucklin. "Implications of the and J. W. Mishoe. "Modeling Soybean Growth for Crop
Greenhouse Effect for World Agricultural Commodity Management." Transactions of the American Society
Markets." Paper presented at the Western Economic of Agricultural Engineers 26(1983):63-73.
Association Conference, June 1989. Western Ag¡ West. J. Agr. Econ. 13

Downloaded from http://ajae.oxfordjournals.org/ at Cambridge University on January 12, 2015


Katz, R. W. "Probabilistic Models." Probability, Statis- (1988):348-56.
tics, and Decision Making in the Atmospheric Sci- Arthur, L. M., and F. Abizadeh. "Potential Effects of Cli-
ences, A. H. Murphy, and R. W. Katz (eds), pp. 261- mate Change on Agriculture in the Prairie Region of
88. Boulder, Colorado: Westview Press, 1985. Canada." West. J. Agr. Econ. 13(1988):216-24.
Katz, R. W., and B. A. Brown "Extreme Events in a Brink, L., and B. McCarl. "The Tradeoff Between Ex-
Changing Climate: Variability Is More Important Than pected Return and Risk Among Cornbelt Farmers."
Averages." Climatic Change 21 (1992):289-302. Amer. J. Agr. Econ. 60(1978):259-63.
van Keulen, H., and J. Wolf (eds). Modeling of Agricul- Buttler, I. W., and S. J. Riha. "GAPS: A General Purpose
tural Production: Weather, Soils, and Crops. Wa- Simulation Model of the Soil-Plant-Atmosphere Sys-
geningen: Pudoc, 1986. tem." Version 1.1 User's Manual. Department of
Kokoski, M. F., and V. K. Smith. "A General Equilibrium Agronomy, Cornell University, 1989.
Analysis of Partial Equilibrium Welfare Measures: The Chapman, D., and T. Drennen. "Equity and Effectiveness
Case of Climate Change." Amer. Econ. Rey. 77 of Possible CO2 Treating Proposals." Contemporary
( 1987):331-41. Policy Issues 8(1990):1-28.
Lewandrowski, J., and R. Brazee. "Government Farm Pro- Cocks, K. D. "Discrete Stochastic Programming." Man-
grams and Climate Change: A First Look" in J. M. agement Science 15(1968):72-79.
Reilly and M. Anderson, eds., Economic lssues in Crosson, P. R. "Climate Change and Mid-latitudes Agri-
Global Climate Change: Agriculture, Forestry, and culture: Perspectives on Consequences and Policy Re-
Natural Resources, pp. 132-47. Boulder, Colorado: sponses." Climatic Change 15(1989):51-73.
Westview Press, 1992. Easterly, W. E., P. R. Crosson, N. J. Rosenberg, M. S.
Murray, J. J. Soil Survey of Redwood County, Minnesota. McKenney, and K. D. Frede¡ "Methodology for
Washington: U.S. Department of Agriculture, 1985. Assessing Regional Economic Impacts of and Re-
Nordhaus, W. D., and G. W. Yohe. "Future Paths of En- sponses to Climate Change: The MINK Study." Eco-
ergy and Carbon Dioxide Emissions." Changing Cli- nomic Issues in Global Climate Change: Agriculture,
mate. Washington, DC: National Research Council, Forestry, and Natural Resources, J. M. Reilly and M.
1983. Anderson, eds., pp. 1688-99. Boulder, Colorado:
Onal, H., and Y. Fang. "The Effects of Future Climate Westview Press, 1992.
Change Upon the Agriculture of Illinois." Paper pre- Edmonds, J., and J. Reilly. "Global Energy and CO2 to the
sented at Global Warming Conference, Chicago, April Year 2050." Energy J. 4(1983):21-46.
1991. Freund, R. J. "The Introduction of Risk into a Program-
Rae, A. N. "Stochastic Programming, Utility and Sequen- ming Model." Econometrica 24(1956):253-63.
tial Decision Problems in Farm Management." Amer. Hansen, J., I. Fung, A. Lacis, D. Rind, S. Lebedeff, R.
J. Agr. Econ. 53(1971):448-60. Ruedy, G. Russel, and P. Stone. "Global Climate
Richardson, C. W. "Stochastic Simulation of Daily Pre- Changes as Forecast by Goddard Institute for Space
cipitation, Temperature, and Solar Radiation." Water Studies Three-Dimensional Model." J. Geophysical
Resources Research. 17(1981): 182-90. Research D93(1988):9341-64.
Rind, D., R. Goldberg, and R. Ruedy. "Change in Climate Wilks, D. S. "Adapting Stochastic Weather Generator AI-
Variability in the 21st Century." Climatic Change gorithms for Climate Change Studies. Climatic Change
14(1989):5-37. 22(1992):67-84.
Rosenthal, W. D., R. L. Vanderlip, B. S. Jackson, and G. Wilson, C. A., and J-F. B. Mitchell. "Simulated Climate
F. Arkin. "SORKAM: A Grain Sorghum Crop Model." and COz-Induced Climate Change Over Western Eu-
Texas Agficultural Experiment Station, College Sta- rope." Climatic Change 10(1987): 11-42.

You might also like