Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

GEOPHYSICAL RESEARCH LETTERS, VOL. 31, L20502, doi:10.

1029/2004GL020927, 2004

Spatiotemporal patterns of cropland area and net primary production


in the central United States estimated from USDA agricultural
information
Jeffrey A. Hicke
Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado, USA

David B. Lobell
Department of Global Ecology, Carnegie Institution of Washington, Stanford, California, USA

Department of Geological and Environmental Sciences, Stanford University, Stanford, California, USA
Received 5 July 2004; accepted 15 September 2004; published 22 October 2004.

[1] The central United States, which is dominated by (see http://www.nass.usda.gov/census and http://www.nass.
agriculture, has been selected as the first North American usda.gov:81/ipedb, respectively), and croplands can be
Carbon Program intensive campaign. Data sets that describe intensively managed to enhance net C uptake [e.g., Ogle
spatiotemporal variability in carbon fluxes are needed to et al., 2003; Paustian et al., 1997]. The flat topography and
support this campaign. Here we report the behavior of knowledge of climate, soils, and land-use history facilitates
county cropland net primary production (NPP) in the first C flux measurements and modeling. For these reasons, the
intensive region derived using USDA information together first intensive campaign of the North American Carbon
with crop-specific parameters that convert agronomic data Program (NACP) [Wofsy and Harriss, 2002] will occur in
into carbon fluxes. Total cropland area in the eight-state the upper Midwest region of the United States. The ‘‘Mid-
region was 550,000 km2 (40% of total area), with some Continent NCAP Intensive Campaign’’ includes South
interannual variability but no temporal trend from 1972 to Dakota, Nebraska, Kansas, Missouri, Iowa, Minnesota,
2001. Regional production (P) was 0.3 Pg C yr1 in the Wisconsin, and Illinois (www.carboncyclescience.gov/
late 1990s, roughly 64% of the total US crop production. nacp-first-intensive-campaign.html). The campaign will
P was highest in the central counties (>1.2 Tg C yr1). In develop and test key approaches for reconciling estimates
contrast to area, both NPP (flux per unit area) and P of regional C sources and sinks derived from atmospheric
(spatially aggregated flux) increased during the study period measurements with those from other methods, including
(46 and 51%, respectively). Corn was the dominant crop inventories, remote sensing, and field measurements.
type grown in the region, contributing 58% of the total [3] One source of C flux information is the USDA
production, with soybeans second most productive but National Agricultural Statistics Service (NASS) [USDA,
substantially less (20%) despite similar harvested area. 2001] database of crop information. Annually, NASS
Maximum year-to-year variability in P was high, generally reports cropland area and yield by county across the US
greater than 30% for most counties, though exceeding 80% for different crop types based on surveys and inventories of
for some counties. INDEX TERMS: 0315 Atmospheric farms. Cropland information is available back to the early
Composition and Structure: Biosphere/atmosphere interactions; 1900s, although the most complete reporting began around
0322 Atmospheric Composition and Structure: Constituent 1970. Patterns and trends can be studied with the agronomic
sources and sinks; 1615 Global Change: Biogeochemical data reported by NASS, though these analyses are limited in
processes (4805) Citation: Hicke, J. A., and D. B. Lobell usefulness with respect to the C cycle for several reasons.
(2004), Spatiotemporal patterns of cropland area and net primary First, agronomic units (e.g., ‘‘bushels’’) cannot be included
production in the central United States estimated from USDA in C cycle studies without conversion to biomass units (g C).
agricultural information, Geophys. Res. Lett., 31, L20502, Second, top-down studies of C fluxes, such as aircraft and
doi:10.1029/2004GL020927. tower measurements, require net fluxes. Therefore, values
from all crop types must be combined to determine the total
1. Introduction effect across crop types. To study cropland C fluxes,
reported yields can be converted to production by utilizing
[2] In the central United States, croplands are important harvest indices, moisture content variables, and other
components of the carbon (C) cycling. Extensive area parameters to compute NPP [Lobell et al., 2002; Prince et
within this region is cultivated (some counties exceed al., 2001].
75% cropland area [Lobell et al., 2002; Prince et al., [4] NASS data were used by Prince et al. [2001] to
2001]), and production is comparable to surrounding wood- investigate NPP of major crop types in the Midwest for
lands [Prince et al., 2001]. Yields and therefore C uptake in 1992, and for 1982 – 1996 for Iowa. The authors found that
crops have increased over time [Hicke et al., 2004; U.S. the areas with the highest NPP were associated with corn
Department of Agriculture (USDA), 1997a; USDA, 2001] and soybeans. In Iowa, NPP over the 15 years varied by
100%. Lobell et al. [2002] found good agreement between
Copyright 2004 by the American Geophysical Union. NPP estimated with a similar approach and that computed
0094-8276/04/2004GL020927 using satellite observations during 1982 – 1998. Hicke et al.

L20502 1 of 5
L20502 HICKE AND LOBELL: CROPLAND NPP IN THE CENTRAL US L20502

Table 1. Parameters Used for Converting Yields to Production for tantly, hay. Here we analyze NPP from only crop types
the 17 Crops Used in this Study for which all years were complete (as opposed to all
Mass per Moisture Harvest Fraction of available information). Incomplete cropland information
Reported Yield Content Index Production Allocated was reported for South Dakota and southern and eastern
a b c
Crop (MRY) (MC), % (HI) Aboveground (fAG) Iowa. Although including all available crop information will
Corn for grain 25.401 11 0.45 0.85d give a more complete picture in some years, doing so would
Soybeans 27.216 10 0.40 0.87d have introduced spurious errors to trend and interannual
Wheat, winter 27.216 11 0.40 0.83d variability calculations that resulted from simply incomplete
Hay, other 907.000 15 1.00 0.53d
Hay, alfalfa 907.000 15 1.00 0.53d reporting as opposed to actual changes in agricultural
Wheat, spring 27.216 11 0.40 0.83d numbers. See Hicke et al. [2004] and Lobell et al. [2002]
Cotton upland 217.700 8 0.40 0.80e for additional details of the processing.
Sorghum for grain 25.401 10 0.40 0.80e [8] In this paper, we refer to ‘‘NPP’’ as having units of
Barley 21.772 12 0.40 0.67d
Corn for silage 907.000 65 1.00 0.85d
g C m2 yr1, whereas ‘‘production’’ has units of g C yr1
Wheat durum 27.216 12 0.40 0.83d (i.e., production is spatially aggregated NPP). To compute
Rice 50.800 9 0.40 0.80e NPP, we divide P by the crop area A.
Sunflower seed 0.453 10 0.35 0.94d
Oats 14.515 11 0.40 0.71d
Peanuts 0.453 9 0.40 0.80e 3. Results and Discussion
Potatoes 50.800 80 0.50 0.80e
Sugarbeets 907.000 85 0.40 0.80e [9] Mean cropland P (across years) for the eight-state
a
USDA [1997b].
region is 0.31 Pg C yr1, about two-thirds of the total US
b
NRC [1982]. cropland production [Hicke et al., 2004] and 11% of the
c
Hay [1995]. total coterminous US NPP [Hicke et al., 2002]. Production
d
Prince et al. [2001]. patterns were similar to cropland area patterns, with highest
e
Lobell et al. [2002]. values in Illinois, Iowa, Minnesota, and Nebraska (Figure 1).
A few counties had extremely high values, >1.2 Tg C yr1,
[2004] discussed nationwide cropland area and production whereas in many of the remaining counties with substantial
trends derived with NASS information over thirty years. cropland area, P ranged from 0.4– 1.2 Tg C yr1.
Although overall US cropland area changed little and [10] Regional cropland area averaged about 550,000 km2
production increased dramatically during this time, substan- (Figure 2), or 40% of the eight-state region. County crop-
tial spatial variability existed in the trends. land areas ranged from near zero in the forested regions of
[5] Here we present detailed cropland production infor- Minnesota, Wisconsin, and Missouri to over 80% in south-
mation for the eight-state region encompassing the first ern Minnesota, Iowa, southeastern South Dakota, eastern
NACP intensive campaign. Spatial patterns of production Nebraska, and Illinois [see also Lobell et al., 2002; Prince et
and year-to-year variability across the NCAP first intensive al., 2001]. Little trend is evident over the 30-year study
area are analyzed. We discuss year-to-year variability and period, though year-to-year variations of 100,000 km2, or
the contribution by different crop types from 1972– 2001. nearly 20% of the mean, occurred. Similar behavior from
1982– 1997 (no change or slight decrease) was reported in
the region by the USDA National Resources Inventory, an
2. Methods independent statistical survey conducted every five years by
the National Resources Conservation Service [USDA
[6] NASS harvested area and yield in agronomic units NRCS, 1997] (see http://www.nrcs.usda.gov/technical/NRI/
were reported annually by county [USDA, 2001]. For 1997). The lack of a trend suggests that all land potentially
an individual county and crop type, we converted yields
(PC) to production P (g C yr1) using the following
equation:

PC  MRY  ð1  MC Þ  C
P¼ : ð1Þ
HI  fAG

MRY is the mass per reported yield, and MC is the harvest


moisture content (mass water/mass harvest, g/g). C
(=0.45 gC/g) converts mass to carbon. The harvest index
HI specifies the ratio of yield mass to aboveground biomass,
and fAG is the fraction of production allocated aboveground,
and is directly related to the root:shoot ratio. The numerator
of equation (1) converts reported yields to C of the
harvested plant part; the denominator accounts for addi-
tional aboveground plant parts (HI) and belowground
production (fAG). MRY, MC, HI, and fAG vary by crop type
and were taken from published values (Table 1).
[7] We selected the 17 crops with the largest areas
(>1 million acres) for analysis. Some of these crop types
were incompletely reported for some years, most impor- Figure 1. Cropland production in 2001 (Tg C yr1).

2 of 5
L20502 HICKE AND LOBELL: CROPLAND NPP IN THE CENTRAL US L20502

effect on yields in years following El Niño events derived


in this study, past studies have reported a higher correlation
of yields with ENSO indices during years prior to ENSO
events [Phillips et al., 1999; Wannebo and Rosenzweig,
2003]. Prince et al. [2001] suggested that different mech-
anisms drove low Iowa NPP in 1983, 1988, and 1993 (a wet
spring coupled with management changes, a summer
drought, and flooding, respectively). Future studies are
needed to explore forcing mechanisms in more detail.
[13] Corn accounted for 35% of the 2001 cropland area in
the region and 58% of the production, and corn production
increased rapidly during the time period of interest
(Figure 3). Soybeans were the second most productive crop,
with 20% of the total regional P occurring on 34% of the
cropland area. Wheat, hay (only counties with complete
reporting), and sorghum each contributed <10%, and
other crops contributed only minor amounts. As evident
in Figure 3, the fraction of production attributable to
individual crops varies by year, which may have implica-
tions for the seasonal timing of C uptake based on pheno-
logical differences between crops.
[14] Characterizing the interannual variability of C fluxes
is important for interpreting the results of the NACP
intensive campaign. The regional time series (Figures 2
and 3) demonstrate the high degree of temporal variability
in area, NPP, and production. The coefficient of variability
(COV; standard deviation of detrended values divided by
the mean) for regional production was 0.13. The maximum
year-to-year change in production was 40% of the mean and
occurred in 1993 –1994. The COV of corn was 0.24, and for
soybeans was 0.30. Maximum year-to-year variability of
regional corn production was 80 Tg C yr1 (57%), while
soybean production varied by 15.4 Tg C yr1 (40%), both
occurring in 1993 – 1994. Negative deviations from the
Figure 2. Time series of (a) cropland area, (b) NPP (g C detrended mean were fewer and larger than positive devia-
m2 yr1), and (c) production (Pg C yr1) aggregated for tions in all three variables (A, NPP, P). Approximately one-
the eight-state region showed in Figure 1. third of the years had negative deviations, but the mean of
these deviations was twice that of the positive deviations.
[15] The map of maximum year-to-year P variability
reveals that many counties exceeded 50% variability, with
available for agriculture was converted before 1972. Inter- values 80% in some counties (Figure 4). A few counties
annual variations in area were likely due to climate and/or
economic drivers.
[11] NPP and P show similar year-to-year variability,
both in magnitude and phase. However, in contrast to area,
the C fluxes increased substantially from 1972– 2001. Over
the thirty-year study period, NPP averaged 470 g C m2
yr1, increasing 46% from 380 to 550 g C m2 yr1. Mean
production was 0.26 Pg C yr1, increasing 51% from 0.2 to
0.3 Pg C yr1. Possible drivers of long-term increased NPP
in the region include improved cultivars, better fertilizer and
pest management, more favorable climate, shifts to produc-
tive crop types, and economic influences [Duvick and
Cassman, 1999; Evans, 1997; Lobell and Asner, 2003].
Similar to cropland area, several of these factors could drive
interannual variations, including climate and economics.
[12] Particularly low flux values occurred in 1983, 1988,
and 1993, years following large El Niño events. The effect
of the El Niño/Southern Oscillation (ENSO) on crop yields
in the US has been noted previously from analysis of yields
[e.g., Phillips et al., 1999], model results [e.g., Izaurralde et
al., 1999; Legler et al., 1999], and remote sensing [e.g., Figure 3. Cropland production (Tg C yr1) for the eight-
Wannebo and Rosenzweig, 2003]. Despite the apparent state region by crop type for the five most productive crops.

3 of 5
L20502 HICKE AND LOBELL: CROPLAND NPP IN THE CENTRAL US L20502

acterizing C uptake in the Midwest. First, although crop-


lands are dominant across the region, other land cover types
are present in the area, including grasslands and forest,
and must be considered when evaluating fluxes. Second,
harvested yields sometimes result in underreporting
of annual production. For example, Lobell et al. [2002]
discussed severe weed problems and flooding in the late
growing season in the Midwest that resulted in deviations
between satellite- and NASS-derived NPP. Third, the coarse
spatial (county) and temporal (annual) scales of this data set
imply that additional methods must be developed to capture
finer scales that will be necessary for analysis within the
context of the NACP. Fourth, the surveys and inventories
performed by NASS at annual timescales require some
subsampling, conferring some uncertainty to the estimates
(95% confidence intervals for 2s uncertainty are estimated
for the most productive crops to be 5 – 10% [USDA, 2001]).
Figure 4. Maximum year-to-year variability (for thirty Improved accuracy can be achieved by integrating Census
years, 1972 – 2001) in production (as percentage of mean of Agriculture data [USDA, 1997a], more complete but
production). available only every five years, with the NASS information.
[ 19 ] The cropland database is available online at
www.nrel.colostate.edu/~jhicke/usda_croplands/usda_
cropland_npp.php.
may be artificially high since we analyzed only crop types
with complete reporting and so may have missed shifting [20] Acknowledgments. We thank Mark Easter, Steve Williams, and
production associated with, for example, hay. However, Keith Paustian for assistance with the NASS database, and Dennis Ojima
high variability also occurred in counties without missing for his support. We are grateful to Rich Conant and Robert Harriss for their
comments and reviews. This research was partially funded by NASA IDS
data. Irrigation provides a buffer against year-to-year vari- grant NNG04GH63G.
ability associated with droughts. As an example, irrigated
corn in central Nebraska produced large NPP but with low References
variability. Duvick, D. N., and K. G. Cassman (1999), Post-green revolution trends in
yield potential of temperate maize in the north-central United States,
Crop Sci., 39, 1622 – 1630.
4. Conclusions Evans, L. T. (1997), Adapting and improving crops: The endless task,
Philos. Trans. R. Soc. London, Ser. B, 352, 901 – 906.
[16] We estimated county-level NPP from cropland area Hay, R. K. M. (1995), Harvest index—A review of its use in plant-breeding
and harvested yields reported by the USDA NASS for and crop physiology, Ann. Appl. Biol., 126, 197 – 216.
1972 – 2001 in the eight-state region of the first NACP Hicke, J. A., G. P. Asner, J. T. Randerson et al. (2002), Satellite-derived
increases in net primary productivity across North America, 1982 – 1998,
intensive campaign. We found that cropland area remained Geophys. Res. Lett., 29(10), 1427, doi:10.1029/2001GL013578.
constant over the 30-year study period, with some interan- Hicke, J. A., D. B. Lobell, and G. P. Asner (2004), Cropland area and net
nual variability. NPP over that time increased by roughly primary production computed from 30 years of USDA agricultural har-
50%, and was also subject to large year-to-year variability. vest data, Earth Interact., 8, 1 – 20.
Izaurralde, R. C., N. J. Rosenberg, R. A. Brown et al. (1999), Modeled
Total regional production was two-thirds and 11% of the effects of moderate and strong ‘Los Ninos’ on crop productivity in North
conterminous US cropland and total production, respectively, America, Agric. For. Meteorol., 94, 259 – 268.
with large year-to-year variability, in some cases exceeding Legler, D. M., K. J. Bryant, and J. J. O’Brien (1999), Impact of ENSO-
related climate anomalies on crop yields in the US, Clim. Change, 42,
80% of the mean. Negative deviations were less frequent and 351 – 375.
larger than positive deviations, suggesting that drivers of Lobell, D. B., and G. P. Asner (2003), Climate and management contri-
interannual variability that reduced yields, such as flooding, butions to recent trends in US agricultural yields, Science, 299, 1032 –
1032.
had a strong influence on production. Corn dominated Lobell, D. B., J. A. Hicke, G. P. Asner et al. (2002), Satellite estimates of
production values. Despite a similar harvest area, soybeans, productivity and light use efficiency in United States agriculture, 1982 –
the second most productive crop, accounted for only 20% of 1998, Global Change Biol., 8, 1 – 15.
the regional production. National Research Council (NRC) (1982), United States – Canadian Tables
of Feed Composition, Natl. Acad. Press, Washington, D. C.
[17] Our cropland NPP is useful to the NACP Mid- Ogle, S. M., F. J. Breidt, M. D. Eve, and K. Paustian (2003), Uncertainty in
Continent Intensive Campaign in several ways. The analysis estimating land use and management impacts on soil organic carbon
of spatiotemporal patterns of NPP improves our understand- storage for US agricultural lands between 1982 and 1997, Global Change
Biol., 9, 1521 – 1542.
ing of C cycling in the region. In addition, cropland NPP Paustian, K., H. P. Collins, and E. A. Paul (1997), Management controls on
provides estimates of C inputs to C cycle models and soil carbon, in Soil Organic Matter in Temperate Agroecosystems, edited
provides a means of validating model outputs. Finally, the by E. A. Paul et al., pp. 15 – 49, CRC Press, Boca Raton, Fla.
Phillips, J., B. Rajagopalan, M. Cane, and C. Rosenzweig (1999), The role
NPP based on surveys and inventories allows for compar- of ENSO in determining climate and maize yield variability in the US
isons with other spatially extensive observations such as cornbelt, Int. J. Climatol., 19, 877 – 888.
remotely sensed data [e.g., Lobell et al., 2002]. Prince, S. D., J. Haskett, M. Steininger et al. (2001), Net primary produc-
[18] Although this data set is a useful resource for tion of US Midwest croplands from agricultural harvest yield data, Ecol.
Appl., 11, 1194 – 1205.
reconciling top-down and bottom-up estimates of regional U.S. Department of Agriculture (USDA) (1997a), Census of agriculture,
C cycling, several issues prevent it from completely char- Natl. Agric. Stat. Serv., Washington, D. C.

4 of 5
L20502 HICKE AND LOBELL: CROPLAND NPP IN THE CENTRAL US L20502

U.S. Department of Agriculture (USDA) (1997b), Statistical bulletin Wofsy, S., and R. Harriss (2002), The North American Carbon Program
number 947a, Natl. Agric. Stat. Serv., Washington, D. C. (NACP), report, U.S. Global Change Res. Prog., Washington, D. C.
U.S. Department of Agriculture (USDA) (2001), Published estimates
database, Natl. Agric. Stat. Serv., Washington, D. C.
U.S. Department of Agriculture (USDA) NRCS (1997), National resources 
inventory, Washington, D. C. J. A. Hicke, Natural Resource Ecology Laboratory, Colorado State
Wannebo, A., and C. Rosenzweig (2003), Remote sensing of US cornbelt University, Fort Collins, CO 80523, USA. (jhicke@nrel.colorado.edu)
areas sensitive to the El Nino – Southern Oscillation, Int. J. Remote Sens., D. B. Lobell, Department of Global Ecology, Carnegie Institution of
24, 2055 – 2067. Washington, Stanford, CA 94305, USA.

5 of 5

You might also like