Response of vegetation growth and productivity to spring climate indicators in the conterminous United States derived from satellite remote sensing data fusion

Page 1

Rir

Ya

b

c

d

a

ARRA

KFPSVGC

1

ila(ei

UT

h0

Agricultural and Forest Meteorology 194 (2014) 132–143

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology

j o ur na l ho me pag e: www.elsev ier .com/ locate /agr formet

esponse of vegetation growth and productivity to spring climatendicators in the conterminous United States derived from satelliteemote sensing data fusion

oungwook Kima,b,∗, J.S. Kimball a,b, K. Didanc, G.M. Henebryd

Flathead Lake Biological Station, The University of Montana, Polson, MT 59860, United StatesNumerical Terradynamics Simulation Group, The University of Montana, Missoula, MT 59812, United StatesDept. of Electrical and Computer Engineering, The University of Arizona, Tucson, AZ 85721, United StatesGeospatial Sciences Center of Excellence, South Dakota State University, Brookings, SD 57007, United States

r t i c l e i n f o

rticle history:eceived 6 December 2013eceived in revised form 11 March 2014ccepted 2 April 2014

eywords:reeze/thawhenologypring frostegetation indexPProp planting date

a b s t r a c t

Spring temperatures strongly influence plant phenology, including budburst, canopy development, andcrop planting period. Recent spring warming coincides with earlier and longer non-frozen season trends,and generally earlier spring canopy onset and vegetation productivity increases over the conterminousUS (CONUS). However, earlier spring onset increases frost damage risk, with potential negative impactsto productivity. Frost sensitivity and vulnerability is heterogeneous over the CONUS domain, while theoccurrence, intensity and regional impact of frost events are difficult to monitor from sparse weatherstations. To enhance regional frost risk monitoring capabilities, we developed spring frost day (SFD) andspring frost damage day (SFDD) metrics spanning a long-term (>30 year) record by integrating a satellitemicrowave remote sensing record of daily landscape freeze-thaw (FT) status and optical-IR sensor basedphenology record of start of season (SOS) and day of peak (DOP) canopy cover.

We find a decreasing regional SFD trend (−1.6 days decade−1; p < 0.1) coincident with spring warm-ing, while the SFDD is generally increasing (1.5 days decade−1; p < 0.1). Spring warming is reducing frostoccurrence, but an earlier SOS trend is paradoxically increasing vegetation frost damage risk. The eco-logical significance of the SFD and SFDD changes were evaluated using satellite derived vegetation gross

primary production (GPP) and vegetation greenness (EVI2) anomalies. Higher SFD and SFDD levels coin-cide with reduced vegetation growth in spring, but only the SFDD shows significant (p < 0.1) correlationwith EVI2 summer growth anomalies. Apparent vegetation sensitivity to the SFDD varies across regionalbiomes and elevation zones, while an increasing SFDD trend indicates potentially larger negative impactson regional vegetation growth with continued warming.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

Spring temperature in temperate and northern climate zoness a major environmental constraint to a range of plant pheno-ogy attributes, including canopy budburst, vegetation dormancynd cold hardiness, start of season (SOS), and crop planting period

Karlsen et al., 2007; Kucharik, 2008; Melaas et al., 2013; Vitasset al., 2009). Recent widespread warming trends are promot-ng earlier seasonal thawing and longer frost-free days in spring

∗ Corresponding author at: The University of Montana, FLBS/NTSG, DHC021, Theniversity of Montana 32 Campus, Missoula, MT 59812, United States.el.: +1 4062436706.

E-mail addresses: [email protected], [email protected] (Y. Kim).

ttp://dx.doi.org/10.1016/j.agrformet.2014.04.001168-1923/© 2014 Elsevier B.V. All rights reserved.

(Burrows et al., 2011; Martin et al., 2010; Yang et al., 2012). Pre-cocious dehardening and faster physiological recovery related towarmer springs are also leading to earlier seasonal onset of pho-tosynthetic activity (Beier et al., 2008; Ensminger et al., 2008;Schwartz and Reiter, 2000). In seasonally frozen environments, arapid rise in spring temperatures promotes snowmelt, landscapethawing and associated shifts in the surface radiation budget, whichhelp to determine leaf expansion (Ault et al., 2011; Morin et al.,2009; Wesolowski and Rowinski, 2006), flowering (Ellwood et al.,2013; Menzel, 2002), net photosynthesis (Hollinger et al., 2004;Richardson et al., 2009; Urbanski et al., 2007), crop planting dates

and seedling development (Mann et al., 2002; Shimono, 2011).

Recently, earlier spring onset and lengthening vegetation grow-ing seasons associated with regional warming trends have beenfound across the Northern Hemisphere (Cleland et al., 2007;

Page 2

rest M

S(b2Kswpia2i2csle2ra2dem

cpDStod(a2cscissfewafPufa2

pdtdUgcti2cact

Y. Kim et al. / Agricultural and Fo

chwartz et al., 2006; Tebaldi et al., 2006). In the continental USACONUS), surface air temperature records indicate a reduced num-er of frost days and earlier date of the last-spring freeze (Easterling,002), and a longer frost-free season (Cooter and Leduc, 1995;unkel et al., 2004). Several studies have reported earlier CONUSpring vegetation greening trends attributed to spring temperaturearming (Wang et al., 2011), and commensurate changes to forestroductivity (Augspurger, 2009; Hufkens et al., 2012), phenolog-

cal responses (Schwartz et al., 2012), and agricultural yields (Hund Buyanovsky, 2003; Schooley and Proctor, 2003; Tubiello et al.,007). However, reductions in spring frost days do not necessarily

ndicate reduced risk of frost damage to vegetation (Meehl et al.,000; Gu et al., 2008; Rixen et al., 2010) and associated fruit androp production (Shimono, 2011; Warmund et al., 2008). An earlierpring bud-burst may enhance risk of damage if frost occurs fol-owing dehardening and canopy leaf-out (Augspurger, 2013; Aultt al., 2013; Inouye, 2008; Norby et al., 2003; Rigby and Porporato,008). Unseasonably late spring frosts may lead to flaccid shoots,oot damage, plant cellular dehydration, low stomatal conductance,nd ice crystal formation within plant cells (Lamontagne et al.,000; Pearce, 2001; Warmund et al., 2008; Wipf et al., 2006). Frostamaged plants may also lose photosynthetic carbon and nutri-nts, resulting in reduced productivity and leaf area, and increasedortality risk (Gu et al., 2008; Martin et al., 2010).Previous studies have proposed a variety of spring climate indi-

ators in the CONUS domain using gridded estimates of climatearameters and in-situ field observations (Arguez et al., 2012;iamond and Lief, 2009; Di Luzio et al., 2008; McCabe et al., 2012;chwartz, 1997). Spring frost days determined from surface airemperatures have been used for studying temporal variabilityf the spring frost season (Easterling, 2002; Kunkel et al., 2004),etermining management routines for agriculture and gardeningKurtural and Wilson, 2008; USDA Plant Hardiness Zone Map 2012),nd analyzing frost impacts on spring phenology (Augspurger,009). Common limitations of weather station-based climate indi-ators in the CONUS domain include observation time differences,ampling artifacts, temporal inhomogeneity due to instrumenthanges, and regionally sparse observations, particularly in thentermountain west, along the west and east coasts, and in theoutheast (Easterling, 2002; Kunkel et al., 2004). Also, previouspring climate indicators have been produced using apparent sur-ace air temperature thresholds (Cittadini et al., 2006; Kunkelt al., 2004; Marino et al., 2011; Terando and Easterling, 2012)ithout considering different low temperature tolerances associ-

ted with vegetation health conditions, genetic differences, plantunctional types, and topographic conditions (Jolly et al., 2005;earce, 2001; Strimbeck et al., 2007). Global climate models oftensed to derive spring climate indicators can also show large dif-erences depending on the type of land parameters, observationnd assimilation schemes used (Kodra et al., 2011; Pierce et al.,009).

Satellite passive microwave remote sensing brightness tem-erature retrievals from lower frequency (≤37 GHz) sensors canetect landscape freeze-thaw (FT) state transitions associated withemporal changes in surface dielectric properties between pre-ominantly frozen and non-frozen conditions (Markus et al., 2009;laby et al., 1982). These data have been used to classify the aggre-ate frozen or non-frozen status of surface soil and vegetationanopy layers, with reduced sensitivity to potential signal degrada-ion from atmospheric aerosol and cloud contamination, and solarllumination effects (Han et al., 2010; Kim et al., 2011; Wang et al.,013). In this study, we apply a global satellite microwave daily FT

lassification record to define a new spring climate indicator (SCI)ssociated with the number of classified frost days. The FT record isombined with synergistic land surface phenology (LSP) informa-ion from satellite optical-IR remote sensing to define the number

eteorology 194 (2014) 132–143 133

of spring frost days occurring between spring canopy onset andpeak seasonal canopy development. The long-term trend patternsof spring crop planting dates for maize and soybean determinedfrom the United States Department of Agriculture (USDA), NationalAgricultural Statistics Service (NASS) are also used to evaluate therelative influence of the SCI on mean annual planting dates forthese major cropland types. We use satellite optical-IR remotesensing derived spectral vegetation greenness indices and vegeta-tion gross primary production (GPP) estimates to quantify annualproductivity and vegetation canopy growth responses to changesin the SCI metrics. Empirical correlation and regression analysis isused to quantify regional SCI trends, spring climate variability andfrost related impacts on vegetation productivity within the CONUSdomain.

2. Data and methods

2.1. Satellite data

The primary remote sensing data used for this study is a globaldaily landscape FT Earth System Data Record (FT-ESDR) developedfrom calibrated overlapping satellite passive microwave remotesensing time series extending from 1979 to 2010 (Kim et al., 2011,2013). The FT-ESDR provides global daily observations of the land-scape FT state at coarse spatial resolution (∼25 km) and was derivedfrom a temporal classification of calibrated, overlapping 37 GHz,vertically polarized brightness temperature records from SMMR(scanning multi-channel microwave radiometer) and SSM/I (spe-cial sensor microwave imager) sensors (Kim et al., 2012). Fourdiscrete FT classification levels are provided, including frozen (AMand PM frozen), non-frozen (AM and PM thawed), transitional (AMfrozen and PM thawed) and inverse transitional (AM thawed andPM frozen) status; the morning (AM) and evening (PM) FT designa-tions are defined relative to equatorial crossing times of the orbitalbrightness temperature retrievals from SMMR (∼12 AM/PM) andSSM/I (∼6 AM/PM). Kim et al. (2012) reported overall mean annualFT-ESDR spatial classification accuracies of 91.4 ± 1.05 (temporalstandard deviation) and 84.2 ± 0.92 percent for the PM and AMoverpass retrievals relative to in situ surface air temperature mea-surements from the global weather station network (Kim et al.,2012). The FT-ESDR has been used to quantify non-frozen sea-son impacts on vegetation productivity, evapotranspiration andthe terrestrial carbon cycle (Barichivich et al., 2013; Bi et al., 2013;Buermann et al., 2013; Xu et al., 2013; Zhang et al., 2011), and tovalidate the land component of the Community Earth System Model(CESM, Shi et al., 2013). The FT-ESDR product used for this investi-gation is publicly available (Kim et al., 2013) in a global cylindricalEqual-Area Scalable Earth grid (EASE-Grid; Brodzik and Knowles,2002) projection format.

We used the MODIS (Moderate Resolution Imaging Spectrora-diometer) MOD17A3 (Collection 055; Zhao and Running, 2010)global record to estimate annual GPP variability over the CONUSdomain. The 30-arcsec (approximately 1 km resolution) annualGPP data were reprojected from the WGS geographic format tothe 25 km global EASE-Grid of the FT-ESDR for this investiga-tion.

The satellite optical-IR vegetation greenness index (VI) datainclude the NDVI (normalized difference vegetation index; Tucker,1979) and EVI (enhanced vegetation index; Huete et al., 2002),which provide a relative measure of photosynthetic canopy cover.However, the NDVI shows a loss of sensitivity under high canopy

biomass (e.g., forests) levels (Chen et al., 2005; Vina et al., 2004;Wang et al., 2005, 2010) and from atmosphere aerosol contamina-tion (Kobayashi and Dye, 2005). Alternatively, the EVI is purportedto have improved canopy sensitivity in higher biomass (e.g.,

Page 3

1 rest M

fds2see(tac2(v(claVpSgrgtapEepF

aca(tvtspTpreasaaiwsTrsd5pF

2

mp

34 Y. Kim et al. / Agricultural and Fo

orested) regions and enhanced vegetation monitoring throughecoupling of canopy background effects and reduction in atmo-phere aerosol contamination (Huete et al., 2002; Xiao et al.,006). While global EVI data records are unavailable prior to thetart of the MODIS operational record in 2000, a long-term globalnvironmental data record employing a two band enhanced veg-tation index (EVI2) without the blue band has been developedJiang et al., 2008). The EVI2 was designed to provide indis-inguishable values from the EVI, while minimizing potentialtmosphere (e.g., aerosols) contamination effects from the bluehannel (Jiang et al., 2008; Kim et al., 2010; Guindin-Garcia et al.,012). The EVI2 has been used as a proxy for leaf area indexGuindin-Garcia et al., 2012; Rocha and Shaver, 2009), fractionalegetation cover (Gitelson, 2013) and gross primary productionPeng et al., 2011; Wu et al., 2010). A consistent, long-term andontinuous global EVI2 record was developed from calibrated over-apping satellite optical-IR sensor data, including MODIS, AVHRRnd SPOT vegetation (Didan, 2010), by the University of Arizonaegetation Index and Phenology Lab (VIP). The VIP VI record isublicly distributed under the NASA MEaSUREs (Making Earthystem Data Records for use in Research Environments) pro-ram (http://vip.arizona.edu). VIP bi-weekly 0.05◦ × 0.05◦ (5.6 km)esolution EVI2 records were used to compute mean vegetationreenness anomalies from March to May (hereafter spring), Juneo August (hereafter summer) and September to November (here-fter fall) for the 1982–2010 period, as the satellite optical-IR VIProduct does not extend to earlier years of record. The 5.6-kmVI2 data, excluding identified cloud and snow contaminated pix-ls, were reprojected from the VIP climate modeling grid (CMG)roduct projection format to the 25 km global EASE-Grid of theT-ESDR.

Bi-weekly 0.05◦ × 0.05◦ (5.6 km) resolution and 5-year aver-ge spring start of season (SOS) and day of peak (DOP) seasonalanopy greenness LSP metrics were defined; SOS was defineds the date corresponding to half of the annual maximum NDVIFisher et al., 2006; Pouliot et al., 2011; White et al., 1997), andhe DOP metric was defined as the day of the annual peak NDVIalue (Scheftic et al., 2014). These LSP metrics are included inhe VIP data record for the 1980–2010 period, even though theatellite optical-IR record only extends from 1982. The same NDVIhenology methods were applied equally to all land cover classes.he SOS retrieval method used on the VIP NDVI record is pur-orted to be more reliable in the northern latitudes and lesseliable in arid, tropical and Mediterranean ecoregions (Whitet al., 2009), while the VIP DOP estimates have generally favor-ble accuracy due to less cloud cover and a relatively stable NDVIhape pattern during summer (Scheftic et al., 2014). The 5-yearverage SOS and DOP metrics were processed to minimize cloudnd aerosol contamination, shadowing and extreme sensor view-ng effects. The VIP NDVI derived phenology metrics were used

ith the FT-ESDR to determine the number of frost days betweenpring canopy onset and peak seasonal canopy development.he VIP 5-year average NDVI phenology classification approacheduces the year-to-year noise effects and phenology algorithmensitivity to data outliers, including multiple NDVI peaks pro-uced by transient snowmelt and sudden precipitation events. The.6-km SOS and DOP data were reprojected from the VIP CMGroduct projection format to the 25 km global EASE-Grid of theT-ESDR.

.2. Defining satellite based spring climate indicators

We integrated synergistic satellite optical-IR and passiveicrowave sensor data records, including VIP derived SOS and DOP

henology metrics, and FT-ESDR derived FT dynamics into a new

eteorology 194 (2014) 132–143

set of SCI metrics representing spring frost impacts to vegetationproductivity. A metric for spring frost days (SFD) is defined as:

SFD =May∑

March

frozen and transitional days (1)

where the number of frost days was determined as the total num-ber of FT-ESDR defined frozen and transitional frost days in spring(March to May). The SFD metric (days) ranges from 0 (no frostdays) to 92 (complete frost days) for the spring period. Previ-ous studies have commonly used March, April and May to definethe spring period within the CONUS domain (Hu and Buyanovsky,2003; Hufkens et al., 2012; van der Velde et al., 2012). The SFDmetric is utilized in this study as a satellite measurement basedindicator of spring frozen temperature conditions affecting regionalvegetation growth.

A metric of spring frost damage days (SFDD) for vegetation isdefined as:

SFDD =DOP∑

SOS

frozen and transitional days (2)

where the SFDD metric (days) defines the number of FT-ESDR clas-sified frozen and transitional frost days between SOS and DOP asdetermined from the VIP phenology record. The number of daysbetween SOS and DOP varies according to geographical location,topography, climate zone, land cover characteristics, and weather.Comparable to the SFD range scale, the SFDD metric ranges from0 days to a maximum period defined by the SOS and DOP differ-ence. Unlike the SFD metric, which defines all frost days occurringin spring, the SFDD defines frost days only occurring duringactive canopy development, where vegetation dehardening follow-ing winter dormancy may occur up to several days after springonset, leading to increased frost vulnerability of new plant tissue(Strimbeck et al., 1995; Martin et al., 2010). Similar approacheshave been used to estimate spring frost damage risk using mini-mum daily air temperatures (Augspurger, 2013; Bennie et al., 2010;Leinonen and Hanninen, 2002), though frost sensitivity of photo-synthetic capacity and efficiency has been shown to be more closelyassociated with the frequency of frost events rather than mean sur-face air temperatures (Easterling, 2002; Inouye, 2008; Wang et al.,2011).

2.3. Ancillary data used to verify SCI metrics

Monthly climatic data compiled by the National Climate DataCenter (NCDC) were used to calculate annual mean spring sur-face air temperatures in the CONUS domain (Menne et al., 2009;http://www.ncdc.noaa.gov/cag/time-series). Annual variability ofspring temperatures was evaluated against corresponding SFDtemporal changes. A global 1-km resolution, 16 class IGBP landcover classification map for 2001 (MODIS MOD12Q1; Friedl et al.,2002) was used to determine land cover attributes within each 25-km resolution FT-ESDR grid cell. A 1-km resolution digital terrainmap (GLOBE Task Team et al., 1999) was used to evaluate the effectsof regional elevation gradients on the SFD pattern. The mean ele-vations over the CONUS domain were constructed by binning the1-km digital terrain map data at 200 m elevation gradient inter-vals.

Crop weekly progress data for agricultural fields at the nation-wide level are available from USDA NASS Quick Stats ProgressReports from 1980 to 2010 and were used to determine spring

planting dates (USDA-NASS, 2013) for evaluating the ecologi-cal significance of the satellite SCI metrics. Maize and soybeanrecords were used as a standard crop reference for this inves-tigation because they represent the largest cropland acreage in

Page 4

Y. Kim et al. / Agricultural and Forest Meteorology 194 (2014) 132–143 135

Table 1Ameriflux sites used for evaluating apparent spring temperature frost sensitivity of tower carbon (GPP, NEE) fluxes for representative CONUS vegetation classes.

Site ID US-SO4 US-Fuf US-FPe US-MOz

Site name Sky Oaks New Flagstaff Unmanaged Forest Fort Peck Missouri OzarkState CA AZ MT MOLocation (lat/lon) 33.3844, −116.6403 35.0890, −111.7620 48.3077,−105.1019 38.7441, −92.2Elevation (m) 1429 2180 634 219Vegetation type CS ENF GRS DBF

tdapmsht

Faa

Selected year 2006 2006

Canopy height 2.5 m 18 m

PI name W.C. Oechel S. Dore

he US and approximately 40% of global maize and soybean pro-uction (Schlenker and Roberts, 2009). Planting dates for maizend soybeans were determined as the day of occurrence forlanting progress at a 10% threshold (Kucharik, 2006). The SCI

etrics were extracted and summarized over all MOD12Q1 clas-

ified cropland area within the CONUS domain. Several studiesave reported similarities between MODIS classified cropland andhe USDA-NASS Cropland Data Layer (Stern et al., 2014), and

ig. 1. (a) Average annual spring frost day (SFD) metric (days) and (b) temporal SFD staverage SFD and SD variability (vertical bars) binned by elevation gradient within the COnd the analysis. Error bars in (c) denote one SD range around the mean SFD value.

2005 200720–40 cm 24.2 mT. Meyers L. Gu and S. Pallardy

other satellite derived cropland area classifications (Giri et al.,2005). However, the MOD12Q1 classification does not distin-guish maize and soybean (Doraiswamy et al., 2005), and maizeis rotated intensively with soybeans within the aggregate crop-

land category (Grant et al., 2007; Hollinger et al., 2005). We alsoassumed that widespread spring planting of maize and soybeansoccurred within the classified cropland area for every year of record(1980–2010).

ndard deviation (SD) derived from the FT-ESDR 32-year (1979–2010) record; (c)NUS domain. White areas in (a) and (b) were excluded from the FT-ESDR domain

Page 5

1 rest Meteorology 194 (2014) 132–143

dsmantC(grttttbokitFscse

2

tpaGicwds(Gt

odtfgd

s2Trp9af

3

3

bE

Fig. 2. Annual variation of mean spring frost day (SFD) and spring air temperature(Ta) anomalies over the CONUS domain and their associated Spearman correlation;

36 Y. Kim et al. / Agricultural and Fo

The ecological significance of the SCI metrics for vegetation pro-uctivity and terrestrial carbon exchange was evaluated at thetand level in relation to in situ tower eddy covariance measure-ent based estimates of vegetation gross primary production (GPP)

nd net ecosystem CO2 exchange (NEE) provided by the Amerifluxetwork (http://ameriflux.ornl.gov/). The tower sites selected forhis investigation are summarized in Table 1 and represent majorONUS land cover types, including deciduous broadleaf forestDBF), evergreen needleleaf forest (ENF), open shrubland (OS) andrassland (GRS), and spanning a large geographic area and diverseange of climate conditions. These data were used with suppor-ing in-situ biophysical and surface air temperature measurementso evaluate the functional response of terrestrial carbon fluxeso satellite observed spring frost events for the different vegeta-ion classes. The landscape FT retrievals determined from 37 GHzrightness temperatures are sensitive to the predominant frozenr non-frozen land surface conditions within the coarse (∼25-m) satellite footprint and correspond closely with FT transitionsndicated from surface air temperature measurements; surface airemperatures from global weather stations have also been used forT-ESDR validation (Kim et al., 2011, 2012). The 37 GHz derived FTignal is generally representative of shallow surface “skin” layeronditions and less sensitive to soil FT conditions, especially undernow, vegetation or surface organic layers (Green et al., 2012; Kimt al., 2011; McDonald et al., 2004).

.4. Data analysis methods

The Spearman’s correlation coefficient (r) was used to assesshe sign and strength of the relationships between the SCI (inde-endent variable) metrics and spring annual mean temperature,nnual mean crop planting date, seasonal mean EVI2, and annualPP anomalies. Temporal anomalies of the parameter series were

nitially computed as annual differences from average conditionsharacterized from the period of record; where a significant trendas identified (p < 0.1), the temporal anomalies were determined asifferences from the long-term detrended mean. The linear regres-ion model developed between the 11-year (2000–2010) MODISMOD17A3) annual GPP and SCI metrics was applied to estimatePP sensitivity to SCI variability for each land cover class within

he CONUS domain.Temporal trends in the annual mean SCI metrics were defined

ver the CONUS domain. The pre-whitened Kendall’s tau statisticerived from the ZYP package in R statistics was used to determinehe SCI trends. When trends were analyzed, outliers were screenedrom the analysis as a non-systematic variation distinguished on arid-cell-wise basis as quantities surpassing ±2 times the standardeviation of the 32-year (1979–2010) mean (Moore, 2006).

Several studies have documented an anomalous Eastern USpring frost event that occurred in April 2007 (Angel, 2007; Gu et al.,008; Polgar and Primack, 2013) across nine states (MO, IL, IN, KY,N, GA, AL, MS, and AR). We used this 2007 spring frost event as aegional case study to evaluate the relative SCI sensitivity. The dailyroportional frost affected area after SOS was estimated within the

states (approximately 1.2 million km2) for 2007 and evaluatedgainst the previous year (2006) and long-term mean climatologyrom the 1980–2010 record.

. Results

.1. Spatial and temporal variability in spring climate indicators

The SFD metric was derived for each EASE-grid cell on an annualasis over the 32-year period (1979–2010) defined from the FT-SDR; the resulting 32 SFD annual counts were averaged for each

temperatures are obtained from monthly NCDC climate data compiled within theCONUS domain. Annual SFD in 1994, where this value was identified as an outlier,is excluded in the statistics.

grid cell to obtain a mean annual SFD map of the CONUS domain.The SFD results (Fig. 1a and b) show a general latitudinal gradi-ent in spring frost days and associated SFD interannual variability(standard deviation, SD) over the CONUS domain, with generallygreater (fewer) frost days at higher (lower) elevations (Fig. 1c). Theresults show large regional SFD variability of ±20.7 days [spatialSD] relative to the regional mean of 24.8 ± 3.9 days [inter-annualSD]. The SFD was only determined for CONUS areas where seasonalfrozen temperatures were estimated to be a significant constraintto annual vegetation productivity, as defined by the FT-ESDR (Kimet al., 2011). Also, urban/built-up, open water and barren land areasidentified by the MOD12Q1 land cover classification were maskedfrom the FT-ESDR. As a result, the SFD record extends over 87% ofthe CONUS domain and was not determined for portions of Florida,Texas, Arizona, and California.

The satellite based SFD record was compared against monthlymean surface air temperatures determined from in-situ weatherstation records (Fig. 2). One year (1994) was excluded from theregional trend analysis as a non-systematic variation due to largeportions of missing SSM/I brightness temperature retrievals in July1994 within the CONUS domain that resulted in lower FT classifica-tion accuracy relative to other years of record (Kim et al., 2011). Theannual mean SFD anomaly was inversely proportional to the annualspring mean air temperature anomaly for the CONUS domain(r = −0.713; p < 0.1), implying that cooler (warmer) spring tempera-tures coincide with a larger (smaller) SFD period. Analysis of the SFDand planting date relations show direct correspondence for maize(r = 0.537; p < 0.1) and soybean (r = 0.421; p < 0.1) over MOD12Q1classified cropland areas (Fig. 3), where earlier (later) planting datesgenerally coincide with a lower (larger) SFD period. Relatively laterplanting dates can delay subsequent crop growth stages, includingsuccessional heading and maturity, resulting in a delayed harvest,potentially shorter growing season and lower yield.

3.2. Frost effects on vegetation growth and productivity

3.2.1. Relations between spring climate indicators and vegetationgrowth

The ecological significance of spring frost events was evalu-ated in relation to stand level carbon (CO2) fluxes by evaluatingtower eddy covariance measurement based daily GPP and NEE

observations, and in-situ air temperature measurements fromfour Ameriflux sites representing major CONUS vegetation classeswhere large spring temperature decreases occurred (Table 1). Meandaily air temperature measurements at each tower site were used

Page 6

Y. Kim et al. / Agricultural and Forest M

Fig. 3. Mean annual variation and Spearman correlations of crop planting dates andspring frost day (SFD) anomalies within classified (MOD12Q1) cropland areas for the1980–2010 period and CONUS domain. Crop planting dates for maize and soybeanwo

afsaidsn9da

The satellite results (Fig. 5) show a decreasing CONUS regional

FmiT

ere obtained from USDA Crop Progress reports and represent approximately 28.0%f the CONUS domain, and located primarily within the central US.

s an indicator of the timing, magnitude and duration of seasonalreezing events. At the CS site (Fig. 4a), in-situ air temperature mea-urements indicate two abrupt freezing events (DOY 50 and 69–71)fter photosynthesis begins (NEE < 0) in the spring, and coincid-ng with a decrease (17–31% relative to pre-freezing conditions) inaily GPP and an increase (5–10%) in daily NEE. A consistent sea-onal rise in GPP and NEE carbon sink activity at the CS site doesot occur until after the final spring temperature freeze event (DOY

5). At the ENF site (Fig. 4b), 6 ◦C mean daily air temperature con-itions from the previous week decrease below freezing to as lows −8 ◦C for six days (DOY 67–72) after spring vegetation uptake

ig. 4. Daily time-series of CO2 eddy covariance measurement based gross primary proean daily air temperature (Ta) measurements for selected Ameriflux tower sites represe

n 2006 (a); evergreen needleleaf forest in 2006 (b); grassland in 2005 (c), and deciduoa ≤ 0.0 ◦C.

eteorology 194 (2014) 132–143 137

initiation (NEE < 0). The freezing temperatures coincide with anabrupt ENF GPP decrease (72%) and corresponding NEE increase(∼1.6 fold). Other apparent spring frost events at the ENF site (e.g.,∼DOY 88, 95) coincide with similar, temporary reductions in GPPand NEE sink activity. At the GRS site (Fig. 4c), a substantial mid-spring (∼DOY 120) frost event (Ta < 0 ◦C) coincides with a large GPPreduction (74%) and NEE increase (24%). The DBF site (Fig. 4d) showstwo distinct freezing temperature periods (DOY 46–51 and DOY68–69), accompanied by reduced GPP (85–99%) and increased NEE(45–101%) relative to pre-freeze conditions. Similar to the CS andENF sites, a consistent spring rise in DBF GPP and NEE sink activ-ity does not occur until after the final spring freeze event (DOY69). However, an apparent late spring frost event (DOY 105) at theDBF site causes a large GPP decrease (35%) and NEE increase (35%);this anomalous event coincides with the end of consistent NEE sinkactivity for the rest of the season; GPP shows a relatively slow recov-ery with the subsequent rise in temperatures, but does not attainpre-freeze levels until approximately 25 days following the latefrost event. The transient reduction of photosynthetic activity at theselected sites from freezing conditions involves high maintenancerespiration for the repair of frost damage to the developing canopy(Ogren, 1997), and the direct constraints of colder temperatureson biological processes. The resulting decrease in photosynthesis(GPP) and increased maintenance respiration lead to a reductionin NEE carbon sink activity. The largest variations of GPP and NEEat the DBF site may be attributed to characteristically higher ratesof photosynthesis per unit leaf mass than other vegetation types(Givnish, 2002; Wesolowski and Rowinski, 2006).

SFD trend (−1.6 days decade−1; p < 0.1) over the 1979–2010 FT-ESDR period. The negative SFD trend contrasts with an increasingSFDD regional trend (1.5 days decade−1; p < 0.1) for the 1980–2010

duction (GPP) and net ecosystem exchange (NEE) fluxes and in-situ surface (2 m)nting the following CONUS vegetation classes and years of record: closed shrublandus broadleaf forest in 2007 (d). Arrows represent spring frost events indicated by

Page 7

138 Y. Kim et al. / Agricultural and Forest Meteorology 194 (2014) 132–143

Fig. 5. Kendall’s tau trends of mean annual spring frost day (SFD) (1979–2010) andspring frost damage day (SFDD) (1980–2010) metrics for the CONUS domain. Annualsiw

pdidgdeo

sit

Table 2Spearman correlations between seasonal mean EVI2 and spring frost day (SFD) andspring frost damage day (SFDD) anomalies for the 1982–2010 period and CONUSdomain. If p > 0.1, then correlations deemed not significant (NS).

SFD SFDD

EVI2 (Spring) −0.426 (p < 0.1) −0.459 (p < 0.1)

(Fig. 7). These results indicate that larger SFD and SFDD levels gen-

Fd

pring climate indicator (SCI) in 1994, where this value was identified as an outlier,s excluded in the statistics. Annual SFD is decreasing by −1.6 days decade−1 (p < 0.1),

hile annual SFDD is increasing by 1.5 days decade−1 (p < 0.1).

eriod encompassed by the VIP LSP record. The earlier SOS (−3.9ays decade−1; p < 0.1) and lower SFD trends produce a moderate

ncreasing regional trend in the associated SFDD metric. Regionalistribution of the SFDD trend (Spatial SD = 0.471) is more hetero-eneous than the SFD trend (Spatial SD = 0.220) over the CONUSomain (Fig. 6a and b), indicating that SFDD changes in the west-rn US are more spatially dispersed relative to the eastern portionf the US.

Spearman correlations between the SCI metrics and EVI2 sea-

onal growth anomalies for the period 1982–2010 are summarizedn Table 2 for spring, summer, and fall conditions. While the rela-ionships between SFD and SFDD metrics, and spring EVI2 growth

ig. 6. Regional Kendall’s tau trend patterns (days decade−1) and associated significant (ay (SFD) (1979–2010) and (b) spring frost damage day (SFDD) (1980–2010) metrics for

EVI2 (Summer) −0.030 (NS) −0.366 (p < 0.1)EVI2 (Fall) 0.073 (NS) −0.217 (NS)

anomalies indicate a significant negative influence of frost occur-rence on spring vegetation growth and productivity, only theSFDD shows a significant correlation with EVI2 summer growthchanges (r = −0.366; p < 0.1). The negative SFDD impact on vegeta-tion growth over the CONUS domain is associated with an earliervegetation greening trend accompanied by greater risk of frostdamage in early spring. There is no significant relationship betweenthe SCI metrics and fall EVI2 changes. The SFD metric largely influ-ences spring canopy growth and generally does not have a majorimpact on later seasonal vegetation development stages. In con-trast, the SFDD is a better predictor of seasonal vegetation growthchanges in both spring and summer.

3.2.2. Frost impacts to annual vegetation productivityGrid cell-wise correlation coefficients were determined within

the CONUS domain between MODIS annual GPP and SCI anoma-lies for the 2000–2010 record. The mean SFD and SFDD values forthe 2000–2010 MODIS record were compared against the meancorrelations between the SCI metrics and annual GPP anomalies

erally have a stronger negative impact on mean annual vegetationproductivity for the CONUS domain. These results also indicatethat SFD and SFDD levels below respective minimum thresholds

p < 0.1) trend areas (in black on adjacent inset maps) derived from (a) spring frostthe CONUS domain.

Page 8

Y. Kim et al. / Agricultural and Forest M

Fig. 7. Variation in mean Spearman correlations (r) between annual gross primaryproduction (GPP) (MODIS) and mean spring frost day (SFD) (a) and spring frostdo(

or

amcTd

TNbafmebs

Car

bs(

amage day (SFDD) (b) anomalies for different spring climate indicator (SCI) levelsver the 2000–2010 MODIS record. Error bars in (c) denote one standard deviationspatial) range around mean r.

f approximately 10 days and 2 days do not promote significanteductions in mean annual GPP for the domain.

The 11-year (2000–2010) average SCI metrics (days) and MODISnnual GPP (g C m−2 yr−1) estimates were linearly regressed to esti-

ate annual productivity sensitivity to the SCI for individual land

over classes. The sensitivity results are summarized in Table 3.hese results show that SFD increases coincide with annual pro-uctivity reductions for the land cover classes; the widespread

able 3ormalized frost sensitivity (%) derived from significant (p < 0.1) linear regressionsetween 11-year (2000–2010) average spring climate indicator (SCI) metrics (days)nd MODIS annual GPP (g C m−2 yr−1) for individual land cover classes. Normalizedrost sensitivity is determined as a proportional difference between maximum and

inimum annual GPP. Proportions (%) of cells analyzed are computed as the ratio ofach land cover area to total vegetated area within the CONUS domain and shown inrackets []. Values enclosed in parentheses represent mean SCI (days) and temporaltandard deviation values for the 2000–2010 record.

Land cover class* SFD SFDD

ENF [8.06] −0.6 (54.4 ± 17.21) 1.0 (8.5 ± 7.78)DBF [8.87] −1.3 (10.7 ± 8.34) −4.1 (1.3 ± 1.58)MF [5.90] −1.0 (18.8 ± 16.41) −2.0 (2.1 ± 3.22)OS [12.9] −0.2 (22.7 ± 21.24) −0.6 (7.4 ± 8.10)WS [4.48] −1.3 (5.3 ± 10.83) −2.1 (1.3 ± 3.33)GRS [26.8] −0.5 (30.6 ± 22.23) −1.4 (4.4 ± 5.84)CL [17.3] −0.6 (24.5 ± 13.38) −0.7 (2.4 ± 3.84)CL/NVM [10.7] −1.3 (16.0 ± 13.05) −0.9 (2.1 ± 2.47)

S, SVN and EBF areas are excluded in this analysis due to non-significant statisticsnd sparse CONUS coverage representing 0.15%, 0.23%, and 0.56% of the domain,espectively.

* Land cover classes include ENF (evergreen needleleaf forest), EBF (evergreenroadleaf forest), DBF (deciduous broadleaf forest), MF (mixed forests), CS (closedhrubland), OS (open shrubland), WS (woody Savannas), SVN (Savannas), GRSgrassland), CL (cropland), and CL/NVM (cropland/natural vegetation mosaic).

eteorology 194 (2014) 132–143 139

negative SFD impact on annual productivity is attributed to directassociations between frost events and reductions in photosyntheticcapacity, and associated shorter spring growth periods (Krameret al., 1996). The annual productivity sensitivity to the SFDD met-ric shows large regional variability in relation to different landcover classes, and attributed to characteristic differences in veg-etation cold tolerance (Strimbeck et al., 2007). Larger SFDD valuesproduce minimal reductions in ENF productivity even though thisvegetation class shows the largest (∼8.5 days) mean SFDD levelfor the 2000–2010 record; the apparent strong ENF frost resis-tance is attributed to characteristically greater cold hardinessadaptations by coniferous needleleaf forests, including expressinglow-temperate-sensitive metabolic enzymes and antifreeze pro-teins during frost events (Moffatt et al., 2006; Murata and Los,1997), and greater xylem resistance to frost embolism relative toother vegetation classes (Sperry and Sullivan, 1992; Givnish, 2002;Mohren et al., 1999; Strimbeck et al., 2007). Annual ENF sensitivityis dominantly constrained by the SFD metric (negative sensitivity)rather than the SFDD metric (positive sensitivity). Ensminger et al.(2008) also indicated that intermittent late spring frost events didnot severely limit photosynthetic recovery or influence chlorophylllevels in evergreen conifers. Spatial aggregation of topographi-cally heterogeneous conditions and climate zones in ENF areaswithin the Northwestern US may also conceal underlying vari-ability in annual productivity and SFDD relationships. A relativelylarge annual productivity reduction (approximately 2.0–4.1%) withincreasing SFDD is demonstrated in DBF and MF areas, which isconsistent with generally greater sensitivity of ring-porous anddiffuse-porous broadleaf deciduous vegetation to frost damage(Sperry and Sullivan, 1992; Givnish, 2002; Kramer et al., 1996;Polgar and Primack, 2013). Defoliation in DBF and MF areas maysuppress photosynthetic capacity whenever frost damage occurs.WS areas also show a large annual productivity reduction (2.1%)with SFDD increases; these results are consistent with other recentstudies indicating negative impacts of FT events on foliar biomassin WS areas (Ewers et al., 2003; Trujillo et al., 2012). OS and GRSareas show relatively low annual productivity reductions (0.6%and 1.4%) to increasing SFDD events; this response is consistentwith the characteristic larger component of below-ground vege-tation biomass of grassland and shrubland vegetation (Wilsey andPolley, 2006), which enables quick canopy regrowth of frost dam-aged above-ground foliar tissue (Kreyling et al., 2008; Wilsey et al.,2011). Relatively less annual productivity decline in cropland areas(0.7–0.9%) are attributed to agricultural management practices,including artificial fertilization, irrigation and crop protection prac-tices that may offset potential frost damage to annual productivity(Chmielewski, 2013; West et al., 2010). However, the MODIS GPPproduct also has generally larger uncertainty for cropland vegeta-tion (Zhao et al., 2005) that may mask potential frost impacts toproductivity.

3.3. SFDD case study of an anomalous spring frost event

Anomalously warm weather in March 2007 led to early springleaf-out and bloom for vegetation across eastern and central NorthAmerica (Polgar and Primack, 2013). Unseasonal frozen temper-atures from a cold Arctic air mass occurred over several days inearly April 2007 and affected a large portion of the central plainsand Southeast (Angel, 2007; Gu et al., 2008; Polgar and Primack,2013). Cold temperatures coupled with prior spring phenologyonset resulted in widespread frost damage to temperate deciduousforest, fruit crops, and natural vegetation (Angel, 2007; Augspurger,

2009; Gu et al., 2008). The daily areal proportion of frost affectedarea following SOS within the 9 states (MO, IL, IN, KY, TN, GA, AL,MS and AR) reporting unusual spring freeze activity in April 2007was derived for 2006, 2007 and the mean seasonal climatology

Page 9

140 Y. Kim et al. / Agricultural and Forest M

Fig. 8. Daily areal proportions (%) of frost affected areas following the start of season(SOS) within nine states (MO, IL, IN, KY, TN, GA, AL, MS and AR; grey on adjacentinset map) reporting an anomalous spring freeze event in April 2007. Daily areal pro-ptd

eralad2tttlroV

4

reanadCpeAnotipsc

tCIrTwr2

ortions are also presented for 2006 and the climatological mean determined fromhe long-term (1980–2010) satellite spring climate indicator record. Grey shadingenotes one standard deviation range around the long-term mean.

stablished from the long-term (1980–2010) satellite record. Theseesults indicate that the extent of frost affected areas is consider-bly larger in early April 2007 (∼DOY 97) relative to 2006 and theong-term record, and extending over more than 30% of the totalrea encompassed by the nine states (Fig. 8). Monthly mean climateata within the nine states compiled by NOAA indicates that March007 was approximately 3.2 ◦C and 3.6 ◦C warmer than 2006 andhe long-term average, respectively, which was followed by a largeemperature decrease of −4.1 ◦C and −1.4 ◦C in April 2007 relativeo the same month in 2006 and the long-term record. The anoma-ous large frost affected area following SOS in 2007 coincides withespective reductions in spring canopy greenness and annual GPPf 9 ± 10.3% [spatial SD] and 4 ± 4.2% [spatial SD] indicated by theIP EVI2 and MOD17A3 records, and relative to the previous year.

. Summary and conclusions

The results of this investigation document a new approach foregional monitoring of spring frost occurrence, and associated veg-tation growth and productivity impacts using satellite microwavend optical-IR remote sensing data fusion. The biophysical sig-ificance of the resulting SCI metrics for vegetation phenologynd productivity is assessed over the CONUS domain against aiverse set of environmental data records, including in-situ towerO2 eddy flux measurements and surface air temperatures, croplanting dates for maize and soybeans, and satellite derived veg-tation greenness (EVI2) and annual productivity (GPP) records.s a case study, an anomalous spring freeze in April 2007 affectedine states and was evaluated by estimating daily areal proportionsf frost affected areas after SOS. The resulting SCI analysis revealshat the SFD metric is sensitive to frozen temperature conditionsnfluencing canopy development, terrestrial carbon fluxes and croplanting schedules in the spring, while the SFDD metric provides atronger indicator of spring frost damage affecting regional canopyover and vegetation productivity.

The results of this study document significant long-term SCIrends consistent with spring warming and earlier SOS across theONUS domain. The SFD trend is decreasing with regional warming.

n contrast, the SFDD trend is increasing due to an earlier SOS trend,esulting in greater frost damage occurrence over the domain.

he apparent paradox of greater frost risk under regional climatearming is consistent with other recent studies representing a

ange of different plant species and climate zones (Augspurger,013; Jonsson et al., 2004; Rigby and Porporato, 2008), while the

eteorology 194 (2014) 132–143

current study indicates that this effect is widespread over theCONUS domain. However, these results are confined to near-termclimate trends in that spring frost (SFD and SFDD) occurrenceswould not occur over much of the CONUS domain if spring becomessufficiently warm.

Spearman correlation of VIP EVI2 responses to the SFD andSFDD metrics reveals significant negative SCI impacts on springvegetation growth and negative SFDD impacts on summer growth;however, we find no apparent SCI impact on fall vegetation growth.The SFD correlations with MODIS annual GPP anomalies indi-cate that warmer springs and associated SFD reductions generallypromote greater annual productivity across the CONUS domain,which is consistent with studies showing increased productivity inresponse to earlier and longer growing seasons (Kimball et al., 2004;Richardson et al., 2009; Dragoni et al., 2011). However, the SCIimpacts to annual productivity are minimal where the respectiveSFD and SFDD values are less than 14% and 3.3% of the spring season.The relative sensitivity of annual productivity to spring frost eventsalso varies according to individual land cover classes, which weattribute to differences in plant functional type characteristics andregional adaptations, including above and below ground biomassallocation, plant structure and cold hardiness. The resulting SCIinduced losses in mean annual productivity (2000–2010) are 0.4%and 1% for respective SFD and SFDD levels over the CONUS domain.These results imply that the SFDD provides a more robust metric ofspring frost damage to regional vegetation cover and productivity.Similar, but larger frost-induced losses in annual productivity havebeen reported from smaller regional studies, including 41–90%losses at Harvard Forest (Urbanski et al., 2007) and 7–14% lossesover the northeastern US (Hufkens et al., 2012).

This study focused on relations between spring frost occurrenceand vegetation phenology, and productivity as derived from land-scape level SCI metrics determined from satellite microwave andoptical-IR remote sensing data fusion. However, other environ-mental factors (e.g., soil moisture and photoperiod) also influencephenology and vegetation growth (Leinonen, 1996; Poirier et al.,2010) and may reinforce (e.g., summer drought) or offset negativespring frost related impacts to annual productivity. For example,relatively water-limited GRS areas are influenced by summer watersupply and potential drought impacts to GPP, which may influenceGPP-SFDD sensitivity (Craine et al., 2012; Inouye, 2008). In addition,in some nutrient (primarily N) limited systems, frost occurrencemay promote above-ground vegetation regrowth and productiv-ity by increasing plant-available soil N from microbial biomassturnover and soil cryoturbation (Kreyling et al., 2008; Cleland et al.,2006; Matzner and Borken, 2008). While it is difficult to isolatevegetation frost impacts from other environmental factors usingrelatively coarse satellite retrievals over a large geographic domain,the results of this study indicate that the SCI metrics are sensitiveto spring frost events affecting land surface phenology and poten-tially damaging to canopy development and annual productivity.The satellite derived SCI metrics may also provide an effective toolfor regional risk assessment and environmental monitoring, andprediction as climate indicators.

Appendix A. Appendix

Frequently used acronyms and abbreviations from themanuscript text:

CL/NVM cropland/natural vegetation mosaicCMG climate modeling grid

CONUS conterminous United StatesCS closed shrublandDBF deciduous broadleaf forestDOP day of peak

Page 10

rest M

EEEEEFGGILM

MNNNOSSSSSSUVW

R

A

A

A

A

A

A

B

B

B

B

B

B

B

C

C

Y. Kim et al. / Agricultural and Fo

ASE-Grid global cylindrical equal-area scalable Earth gridBF evergreen broadleaf forestNF evergreen needleleaf forestSDR Earth System Data RecordVI2 enhanced vegetation index 2T freeze-thawPP gross primary productionRS grassland

GBP International Geosphere-Biosphere ProgrammeSP land surface phenologyEaSUREs Making Earth System Data Records for use in Research

EnvironmentsODIS Moderate Resolution Imaging SpectroradiometerASS National Agricultural Statistics ServiceCDC National Climate Data CenterDVI normalized difference vegetation indexS open shrublandCI spring climate indicatorD standard deviationFD spring frost daysFDD spring frost damage daysOS start of seasonVN SavannasSDA United States Department of AgricultureIP Vegetation Index and Phenology LabS woody Savannas

eferences

ngel, J.R., 2007. The April 2007 hard freeze in Illinois: features and impacts.Illinois State Water Survey, Champaign, IL, pp. 22, ISWS DCS 2007-05http://www.isws.illinois.edu/pubdoc/DCS/ISWSDCS2007-05.pdf.

rguez, A., Durre, I., Applequist, S., Vose, R.S., Squires, M.F., Yin, X., Heim, R.R., Owen,T.W., 2012. NOAA’s 1981–2010 U.S. climate normal: an overview. Bull. Am.Meteorol. Soc. 93, 1687–1697.

ugspurger, C.K., 2013. Reconstructing patterns of temperature, phenology, andfrost damage over 124 years: spring damage risk is increasing. Ecology 94 (1),41–50.

ugspurger, C.K., 2009. Spring 2007 warmth and frost: phenology, damage andrefoliation in a temperate deciduous forest. Funct. Ecol. 23, 1031–1039.

ult, T.R., Henebry, G.M., de Beurs, K.M., Schwartz, M.D., Betancourt, J.L., Moore, D.,2013. The false spring of 2012, earliest in North American record. EOS Trans. Am.Geophys. Union 94 (20), 181–182.

ult, T.R., Macalady, A.K., Peterson, G.T., Betancourt, J.L., Schwartz, M.D., 2011. North-ern hemisphere modes of variability and the timing of spring in Western NorthAmerica. J. Clim. 24, 4003–4014.

arichivich, J., Briffa, K.R., Myneni, R.B., Osborn, T.J., Melvin, T.M., Ciais, P., Piao, S.L.,Tucker, C., 2013. Large-scale variations in the vegetation growing season andannual cycle of atmospheric CO2 at high northern latitudes from 1950 to 2011.Global Change Biol. 19, 3167–3183.

eier, C.M., Sink, S.E., Hennon, P.E., D’Amore, D.V., Juday, G.P., 2008. Twentieth-century warming and the dendroclimatology of declining yellow-cedar forestsin southeastern Alaska. Can. J. For. Res. 38, 1319–1334.

ennie, J., Kubin, E., Wiltshire, A., Huntley, B., Baxter, R., 2010. Predicting spatial andtemporal patterns of bud-burst and spring frost risk in north-west Europe: theimplications of local adaptation to climate. Global Change Biol. 16, 1503–1514.

i, J., Xu, L., Samanta, A., Zhu, Z., Myneni, R., 2013. Divergent arctic-boreal vegetationchanges between North America and Eurasia over the past 30 years. RemoteSens. 5, 2093–2112.

rodzik, M.J., Knowles, K.W., 2002. EASE-Grid: A versatile set of equal-area projec-tions and grids. In: Goodchild, M. (Ed.), Discrete Global Grids. National Centerfor Geographic Information and Analysis, Santa Barbara, CA, USA.

uermann, W., Bikash, P.R., Jung, M., Burn, D.H., Reichstein, M., 2013. Earlier springsdecrease peak summer productivity in North American boreal forests. Environ.Res. Lett. 8, 024027.

urrows, M.T., Schoeman, D.S., Buckley, L.B., Moore, P., Poloczanska, E.S., Brander,K.M., Brown, C., Bruno, J.F., Duarte, C.M., Halpern, B.S., Holding, J., Kappel, C.V.,Kiessling, W., O’Connor, M.I., Pandolfi, J.M., Parmesan, C., Schwing, F.B., Sydeman,W.J., Richardson, A.J., 2011. The pace of shifting climate in marine and terrestrialecosystems. Science 334, 652–655.

hen, X., Vierling, L., Deering, D., Conley, A., 2005. Monitoring boreal forest leaf area

index across a Siberian burn chronosequence: a MODIS validation study. Int. J.Remote Sens. 26 (24), 5433–5451.

hmielewski, F., 2013. Phenology in agriculture and horticulture. In: Schwartz,M.D. (Ed.), Phenology: An Integrative Environmental Science. Springer,pp. 539–562.

eteorology 194 (2014) 132–143 141

Cittadini, E.D., de Ridder, N., Peri, P.L., van Keulen, H., 2006. A method for assessingforst damage risk in sweet cherry orchards of South Patagonia. Agric. For. Mete-orol. 141, 235–243.

Cleland, E.E., Chuine, I., Menzel, A., Mooney, H.A., Schwartz, M.D., 2007. Shif-ting plant phenology in response to global change. Trends Ecol. Evol. 22 (7),357–365.

Cleland, E.E., Chiariello, N., Loarie, R., Mooney, H.A., Field, C.B., 2006. Diverseresponses of phenology to global changes in a grassland ecosystem. Proc. Nat.Acad. Sci. U.S.A. 103 (37), 13740–13744.

Cooter, E.J., Leduc, S.K., 1995. Recent frost data trends in the north-eastern USA. Int.J. Climatol. 15 (1), 65–75.

Craine, J.M., Nippert, J.B., Elmore, A.J., Skibbe, A.M., Hutchinson, S.L., Brunsell, N.A.,2012. Timing of climate variability and grassland productivity. Proc. Nat. Acad.Sci. U.S.A. 109 (9), 3401–3405.

Diamond, H.J., Lief, C.J., 2009. A comprehensive data portal for global climate infor-mation. EOS 90 (39), 341–342.

Didan, K., 2010. Multi-satellite earth science data record for studying global vegeta-tion trends and changes. In: Proceedings of the 2010 International Geoscienceand Remote Sensing Symposium, 25–30 July 2010, Honolulu, HI, USA.

Di Luzio, M.D., Johnson, G.L., Daly, C., Eischeid, J.K., Arnold, J.G., 2008. Construc-ting retrospective gridded daily precipitation and temperature datasets for theconterminous United States. J Appl. Meteorol. Climatol. 47, 475–497.

Doraiswamy, P.C., Sinclair, T.R., Hollinger, S., Akhmedov, B., Stern, A., Prueger, J.,2005. Application of MODIS derived parameters for regional crop yield assess-ment. Remote Sens. Environ. 97, 192–202.

Dragoni, D., Schmid, H.P., Wayson, C.A., Potter, H., Grimmond, C.S.B., Randolph, J.C.,2011. Evidence of increased net ecosystem productivity associated with a longervegetated season in a deciduous forest in south-central Indiana USA. GlobalChange Biol. 17, 886–897.

Easterling, D.R., 2002. Recent changes in frost days and the frost-free season in theUnited States. Bull. Am. Meteorol. Soc. 83, 1327–1332.

Ellwood, E.R., Temple, S.A., Primack, R.B., Bradley, N.L., Davis, C.C., 2013. Record-breaking early flowering in the Eastern United States. PLoS One 8 (1), e53788.

Ensminger, I., Schmidt, L., Lloyd, J., 2008. Soil temperature and intermittent frostmodulate the rate of recovery of photosynthesis in Scots pine under simulatedspring conditions. New Phytol. 177, 428–442.

Ewers, F.W., Lawson, M.C., Bowen, T.J., Davis, S.D., 2003. Freeze/thaw stress in Cean-othus of southern California chaparral. Oecologia 136, 213–219.

Fisher, J.I., Mustard, J.F., Vadeboncoeur, M.A., 2006. Green leaf phenology at Landsatresolution: scaling from the field to the satellite. Remote Sens. Environ. 100,265–279.

Friedl, M.A., McIver, D.K., Hodges, J.C.F., Zhang, X.Y., Muchoney, D., Strahler, A.H.,Woodcock, C.E., Gopal, S., Schneider, A., Cooper, A., Baccini, A., Gao, F., Schaaf,C., 2002. Global land cover mapping from MODIS: algorithms and early results.Remote Sens. Environ. 83, 287–302.

Giri, C., Zhu, Z., Reed, B., 2005. A comparative analysis of the Global Land Cover 2000and MODIS land cover data sets. Remote Sens. Environ. 94, 123–132.

Gitelson, A.A., 2013. Remote estimation of crop fractional vegetation cover: the useof noise equivalent as an indicator of performance of vegetation indices. Int. J.Remote Sens. 34 (17), 6054–6066.

Givnish, T.J., 2002. Adaptive significance of evergreen v. deciduous leaves: solvingthe triple paradox. Silva Fennica 36, 703–743.

GLOBE Task TeamHasting, D.A., Dunbar, P.K., Elphingstone, G.M., et al., 1999. TheGlobal Land One-Kilometer Base Elevation (GLOBE) Digital Elevation Model,Version 1.0. National Oceanic and Atmospheric Administration, National Geo-physical Data Center, Boulder, CO, U.S.A, Digital data base on the World WideWeb (〈http://www.ngdc.noaa.gov/mgg/toto/globe.html〉) and CD-ROMs.

Grant, R.F., Arkebauer, T.J., Dobermann, A., Hubbard, K.G., Schimelfenig, T.T., Suyker,A.E., Verma, S.B., Walters, D.T., 2007. Net biome productivity of irrigated andrainfed maize-soybean rotations: Modeling vs Measurements. Agron. J. 99,1404–1423.

Green, J., Kongoli, C., Prakash, A., Sturm, M., Duguay, C., Li, S., 2012. Quantifying therelationships between lake fraction, snow water equivalent and snow depth,and microwave brightness temperatures in an arctic tundra landscape. RemoteSens. Environ. 127, 329–340.

Gu, L., Hanson, P.J., Post, W.M., Kaiser, D.P., Yang, B., Nemani, R., Pallardy, S.G., Meyers,T., 2008. The 2007 Eastern US spring freeze: Increased cold damage in a warmingworld? BioScience 58 (3), 253–262.

Guindin-Garcia, N., Gitelson, A.A., Arkebauer, T.J., Shanahan, J., Weiss, A., 2012. Anevaluation of MODIS 8- and 16-day composite products for monitoring maizegreen leaf area index. Agric. For. Meteorol. 161, 15–25.

Han, L., Tsunekawa, A., Tsubo, M., 2010. Monitoring near-surface soil freeze-thawcycles in northern China and Mongolia from 1998 to 2007. Int. J. Appl. Earth Obs.Geoinf. 12, 375–384.

Hollinger, S.E., Bernacchi, C.J., Meyers, T.P., 2005. Carbon budget of mature no-tillecosystem in North Central region of the United States. Agric. For. Meteorol.130, 59–69.

Hollinger, D.Y., Aber, J., Dail, B., Davidson, E.A., Goltz, S.M., Hughes, H., Leclerc, M.Y.,Lee, J.T., Richardson, A.D., Rodrigues, C., Scott, N.A., Achuatavarier, D., Walsh,J., 2004. Spatial and temporal variability in forest-atmosphere CO2 exchange.Global Change Biol. 10, 1689–1706.

Hu, Q., Buyanovsky, G., 2003. Climate effects on corn yield in Missouri. J. Appl.Meteorol. 42, 1626–1635.

Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview ofthe radiometric and biophysical performance of the MODIS vegetation indices.Remote Sens. Environ. 83, 195–213.

Page 11

1 rest M

H

I

J

J

J

K

K

K

K

K

K

K

K

K

K

K

K

K

K

L

L

L

M

M

M

M

M

M

M

M

M

42 Y. Kim et al. / Agricultural and Fo

ufkens, K., Friedl, M.A., Keenan, T.F., Sonnentag, O., Bailey, A., O’Keefe, J., Richardson,A.D., 2012. Ecological impacts of a widespread frost event following early springleaf-out. Global Change Biol. 18, 2365–2377.

nouye, D.W., 2008. Effects of climate change on phenology, frost damage, and floralabundance of montane wildflowers. Ecology 89 (2), 353–362.

iang, Z., Huete, A.R., Didan, K., Miura, T., 2008. Development of a two-bandenhanced vegetation index without a blue band. Remote Sens. Environ. 112,3833–3845.

olly, W.M., Nemani, R., Running, S.W., 2005. A generalized, bioclimatic index topredict foliar phenology in response to climate. Global Change Biol. 11, 619–632.

onsson, A.M., Linderson, M., Stjernquist, I., Schlyter, P., Barring, L., 2004. Climatechange and the effect of temperature backlashes causing frost damage in Piceaabies. Global Planet. Change 44, 195–207.

arlsen, S.R., Solheim, I., Beck, P.S.A., Hogda, K.A., Wielgolaski, F.E., Tommervik, H.,2007. Variability of the start of the growing season in Fennoscandia, 1982–2002.Int. J. Biometeorol. 51, 513–524.

im, Y., Kimball, J.S., Glassy, J., McDonald, K.C., 2013. MEaSUREs Global Record ofDaily Landscape Freeze/Thaw Status. Version 2. NASA DAAC at the NationalSnow and Ice Data Center, Boulder, CO, USA [indicate subset used].

im, Y., Kimball, J.S., Zhang, K., McDonald, K.C., 2012. Satellite detection of increasingnorthern hemisphere non-frozen seasons from 1979 to 2008: implications forregional vegetation growth. Remote Sens. Environ. 121, 472–487.

im, Y., Kimball, J.S., McDonald, K.C., Glassy, J., 2011. Developing a global data recordof daily landscape freeze/thaw status using satellite passive microwave remotesensing. IEEE Trans. Geosci. Remote Sens. 49 (3), 949–960.

im, Y., Huete, A.R., Miura, T., Jiang, Z., 2010. Spectral compatibility of vegeta-tionindices across sensors: band decomposition analysis with Hyperion data.J. Appl. Remote Sens. 4, 043520.

imball, J.S., McDonald, K.C., Running, S.W., Frolking, S.E., 2004. Satellite radarremote sensing of seasonal growing seasons for boreal and subalpine evergreenforests. Remote Sens. Environ. 90, 243–258.

obayashi, H., Dye, D.G., 2005. Atmospheric conditions for monitoring the long-termvegetation dynamics in the Amazon using normalized difference vegetationindex. Remote Sens. Environ. 97, 519–525.

odra, E., Steinhaeuser, K., Ganguly, A.R., 2011. Persisting cold extremes under 21st-century warming scenarios. Geophys. Res. Lett. 38, L08705.

ramer, K., Friend, A., Leinonen, I., 1996. Modelling comparison to evaluate theimportance of phenology and spring frost damage for the effects of climatechange on growth of mixed temperate-zone deciduous forests. Clim. Res. 7,31–41.

reyling, J., Beierkuhnlein, C., Pritsch, K., Schloter, M., Jentsch, A., 2008. Recurrentsoil freeze-thaw cycles enhance grassland productivity. New Phytol. 177 (4),938–945.

ucharik, C.J., 2008. Contribution of planting date trends to increased Maize yieldsin the Central United States. Agron. J. 100 (2), 328–336.

ucharik, C.J., 2006. A multidecadal trend of earlier corn planting in the Central USA.Agron. J. 98, 1544–1550.

unkel, K.E., Easterling, D.R., Hubbard, K., Redmond, K., 2004. Temporal variationsin frost-free season in the United States: 1985–2000. Geophys. Res. Lett. 31,L03201.

urtural, S.K., Wilson, P.E., 2008. Kentucky Viticultural Regions and SuggestedCultivars, HO-88. University of Kentucky Agricultural Experiment Stationhttp://www.ca.uky.edu/agc/pubs/ho/ho88/ho88.pdf

amontagne, M., Bigras, F.J., Margolis, H.A., 2000. Chlorophyll fluorescence and CO2

assimilation of black spruce seedlings following frost in different temperatureand light conditions. Tree Physiol. 20, 249–255.

einonen, I., Hanninen, H., 2002. Adaptation of the timing of bud burst of NorwaySpruce to temperatue and boreal climates. Silva Fennica 36 (3), 695–701.

einonen, I., 1996. A simulation model for the annual forst hardiness and freezedamage of Scots pine. Ann. Bot. 78, 687–693.

ann, L., Tolbert, V., Cushman, J., 2002. Potential environmental effects of corn (Zeamays L.) stover removal with emphasis on soil organic matter and erosion. Agric.Ecosyst. Environ. 89, 149–166.

arino, G.P., Kaiser, D.P., Gu, L., Ricciuto, D.M., 2011. Reconstruction of false springoccurrences over the southern United States, 1901–2007: an increasing risk ofspring freeze damage? Environ. Res. Lett. 6, 024015.

arkus, T., Stroeve, J.C., Miller, J., 2009. Recent changes in Arctic sea ice melt onset,freezeup, and melt season length. J. Geophys. Res. 114, C12024.

artin, M., Gavazov, K., Korner, C., Hattenschwiler, S., Rixen, C., 2010. Reducedearly growing season freezing resistance in alpine treeline plants under elevatedatmospheric CO2. Global Change Biol. 16, 1057–1070.

atzner, E., Borken, W., 2008. Do freeze-thaw events enhance C and N losses fromsoils of different ecosystems? A review. Eur. J. Soil Sci. 59, 274–284.

cCabe, G.J., Ault, T.R., Cook, B.I., Betancourt, J.L., Schwartz, M.D., 2012. Influences ofthe El Nino southern oscillation and the pacific decadal oscillation on the timingof the North American spring. Int. J. Climatol. 32, 2301–2310.

cDonald, K.C., Kimball, J.S., Njoku, E., Zimmermann, R., Zhao, M., 2004. Variabilityin springtime thaw in the terrestrial high latitudes: monitoring a major controlon the biospheric assimilation of atmospheric CO2 with spaceborne microwaveremote sensing. Earth Interact. 8 (20), 1–23.

eehl, G.A., Karl, T., Eastering, D.R., et al., 2000. An introduction to trends in

extreme weather and climate events: observations, socioeconomic impacts, ter-restrial ecological impacts, and model projections. Bull. Am. Meteorol. Soc. 81,413–416.

elaas, E.K., Richardson, A.D., Friedl, M.A., Dragoni, D., Gough, C.M., Herbst, M., Mon-tagnani, L., Moors, E., 2013. Using FLUXNET data to improve models of springtime

eteorology 194 (2014) 132–143

vegetation activity onset in forest ecosystems. Agric. For. Meteorol. 171–172,46–56.

Menne, M.J., Williams, C.N., Vose, R.S., 2009. The US historical climatology networkmonthly temperature data, version 2. Bull. Am. Meteorol. Soc. 90, 993–1107http://www.ncdc.noaa.gov/cag/time-series

Menzel, A., 2002. Phenology: its importance to the global change community. Clim.Change 54, 379L 385.

Moffatt, B., Ewart, V., Eastman, A., 2006. Cold comfort: plant antifreeze proteins.Physiol. Plant. 126, 5–16.

Mohren, G.M.J., Bartelink, H.H., Kramer, K., Magnani, F., Sabaté, S., Loustau, D., 1999.Modelling long-term effects of CO2 increase and climate change on Europeanforests, with emphasis on ecosystem carbon budgets. In: Ceulemans, R.J.M.,Veroustraete, F., Gond, V., van Rensbergen, J.B.H.F. (Eds.), Forest EcosystemModelling, Upscaling and Remote Sensing. SPB Academic Publishing bv, TheNetherlands, pp. 179–192.

Moore, D.S., 2006. Introduction to the Practice of Statistics, fifth ed. W.H. Freemanand Company.

Morin, X., Lechowicz, M.J., Augspurger, C., O’Keefe, J., Viner, D., Chuine, I., 2009. Leafphenology in 22 North American tree species during the 21st century. GlobalChange Biol. 15, 961–975.

Murata, N., Los, D.A., 1997. Membrane fluidity and temperature perception. PlantPhysiol. 115, 875–879.

Norby, R.J., Hartz-Rubin, J.S., Verbrugge, M.J., 2003. Phenological responses in mapleto experimental atmospheric warming and CO2 enrichment. Global Change Biol.9, 1792–1801.

Ogren, E., 1997. Relationship between temperature, respiratory loss of sugar andpremature dehardening in dormant Scots pine seedlings. Tree Physiol. 17,47–51.

Pearce, R.S., 2001. Plant freezing and damage. Ann. Bot. 87, 417–424.Peng, Y., Gitelson, A.A., Keydan, G., Rundquist, D.C., Moses, W., 2011. Remote esti-

mation of gross production in maize and support for a new paradigm based ontotal crop chlorophyll content. Remote Sens. Environ. 115, 978–989.

Pierce, D.W., Barnett, T.P., Santer, B.D., Gleckler, P.J., 2009. Selecting global climatemodels for regional climate change studies. Proc. Nat. Acad. Sci. U.S.A. 106 (21),8441–8446.

Poirier, M., Lacointe, A., Ameglio, T., 2010. A semi-physiological model of cold hard-ening and dehardening in walnut stem. Tree Physiol. 30, 1555–1569.

Polgar, C.A., Primack, R.B., 2013. Leaf out phenology in temperate forests. Biodivers.Sci. 21 (1), 111–116.

Pouliot, D., Latifovic, R., Fernandes, R., Olthof, I., 2011. Evaluation of com-positing period and AVHRR and MERIS combination for improvement ofspring phenology detection in deciduous forests. Remote Sens. Environ. 115,158–166.

Richardson, A.D., Hollinger, D.Y., Dail, D.B., Lee, J.T., Munger, W., O’Keefe, J., 2009.Influence of spring phenology on seasonal and annual carbon balance in twocontrasting New England forests. Tree Physiol. 29, 321–331.

Rigby, J.R., Porporato, A., 2008. Spring frost risk in a changing climate. Geophys. Res.Lett. 35, L12703.

Rixen, C., Schwoerer, C., Wipf, S., 2010. Winter climate change at different temporalscales in Vaccinium myrtillus, an Arctic and alpine dwarf shrub. Polar Res. 29,85–94.

Rocha, A.V., Shaver, G.R., 2009. Advantages of a two band EVI calculated fromsolar photosynthetically active radiation fluxes. Agric. For. Meteorol. 149,1560–1563.

Scheftic, W., Zeng, X., Broxton, P., Brunke, M., 2014. Intercomparison of seven NDVIproducts over the United States and Mexico. Remote Sens. 6, 1057–1084.

Schlenker, W., Roberts, M.J., 2009. Nonlinear temperature effects indicate severedamages to U.S. crop yields under climate change. Proc. Nat. Acad. Sci. U.S.A.106 (37), 15594–15598.

Schooley, J., Proctor, J.T.A., 2003. Freeze damage to North American Ginseng. Hort-Technology 13 (4), 697–701.

Schwartz, M.D., Ault, T.R., Betancourt, J.L., 2012. Spring onset variations and trends inthe continental United States: past and regional assessment using temperature-based indices. Int. J. Climatol., http://dx.doi.org/10.1002/joc.3625.

Schwartz, M.D., Ahas, R., Aasa, A., 2006. Onset of spring starting earlier across theNorthern hemisphere. Global Change Biol. 12, 343–351.

Schwartz, M.D., Reiter, B.E., 2000. Changes in North American spring. Int. J. Climatol.20, 929–932.

Schwartz, M.D., 1997. Spring index models: an approach to connecting satellite andsurface phenology. In: Lieth, H., Schwartz, M.D. (Eds.), Phenology in SeasonalClimates I. Springer, pp. 23–38.

Shi, M., Yang, Z.-L., Landerer, F.W., 2013. Representing and evaluating the landscapefreeze/thaw properties and their impacts on soil impermeability: Hydrologicalprocesses in the community land model version 4. J. Geophys. Res. Atmos. 118,7542–7557.

Shimono, H., 2011. Earlier rice phenology as a result of climate change can increasethe risk of cold damage during reproductive growth in northern Japan. Agric.Ecosyst. Environ. 144, 201–207.

Sperry, J.S., Sullivan, J.E.M., 1992. Xylem embolism in response to freeze-thawcycles and water stress in ring-porous, diffuse-porous, and conifer species. PlantPhysiol. 100 (2), 605–613.

Stern, A.J., Doraiswamy, P.C., Hunt Jr., E.R., 2014. Comparison of different MODIS dataproduct collections over an agricultural area. Remote Sens. Lett. 5 (1), 1–9.

Strimbeck, G.R., Kjellsen, T.D., Schaberg, P.G., Murakami, P.F., 2007. Cold in the com-mon garden: comparative low-temperature tolerance of boreal and temperatureconifer foliage. Trees—Struct. Funct. 21, 557–567.

Page 12

rest M

S

T

T

T

T

T

U

U

UU

v

V

V

W

W

W

W

Zhao, M., Running, S.W., 2010. Drought-induced reduction in global terrestrial net

Y. Kim et al. / Agricultural and Fo

trimbeck, G.R., Schaberg, P.G., DeHayes, D.H., Shane, J.B., Hawley, G.J., 1995. Mid-winter dehardening of montane red spruce during a natural thaw. Can. J. For.Res. 25, 2040–2044.

ebaldi, C., Hayhoe, K., Arblaster, J.M., Meehl, G.A., 2006. Going to the extremes: anintercomparison of model-simulated historical and future changes in extremeevents. Clim. Change 79, 185–211.

erando, A., Easterling, W.E., 2012. Observed and modeled twentieth-century spatialand temporal patterns of selected agro-climate indices in North America. J. Clim.25, 473–490.

rujillo, E., Molotch, N.P., Goulden, M.J., Kelly, A.E., Bales, R.C., 2012. Elevation-dependent influence of snow accumulation on forest greening. Nat. Geosci. 5,705–709.

ubiello, F.N., Soussana, J., Howden, S.M., 2007. Crop and pasture response to climatechange. Proc. Nat. Acad. Sci. U.S.A. 104 (50), 19686–19690.

ucker, C.J., 1979. Red and photographic infrared linear combinations for monitoringvegetation. Remote Sens. Environ. 8, 127–150.

laby, F.T., Moore, R.K., Fung, A.K., 1982. Microwave Remote Sensing-Active andPassive, Vol. II. Addison-Wesley Publishing Company.

rbanski, S., Barford, C., Wofsy, S., Kucharik, C., Pyle, E., Budney, J., McKain, K., Fitz-jarrald, D., Czikowsky, M., Munger, J.W., 2007. Factors controlling CO2 exchangeon timescales from hourly to decadal at Harvard Forest. J. Geophys. Res. 112,G02020.

SDA-NASS, 2013. USDA-NASS, Quick Stats 〈http://quickstats.nass.usda.gov/〉.SDA Plant Hardiness Zone Map, 2012. Agricultural Research Service. U.S. Depart-

ment of Agriculture, Accessed from 〈http://planthardiness.ars.usda.gov〉.an der Velde, M., See, L., Fritz, S., Verheijen, F.G.A., Khabarov, N., Obersteiner, M.,

2012. Generating crop calendars with Web search data. Environ. Res. Lett. 7,024022.

ina, A., Henebry, G.M., Gitelson, A.A., 2004. Satellite monitoring of vegetationdynamics: sensitivity enhancement by the Wide Dynamic Range VegetationIndex. Geophys. Res. Lett. 31, L04503.

itasse, Y., Delzon, S., Dufrene, E., Pontailler, J., Louvet, J., Kremer, A., Michalet,R., 2009. Leaf phenology sensitivity to temperature in European trees: dowithin-species populations exhibit similar responses? Agric. For. Meteorol. 149,735–744.

armund, M.R., Guinan, P., Fernandez, G., 2008. Temperatues and cold damage tosmall fruit crops across the Eastern United States associated with the April 2007Freeze. HortScience 43 (6), 1643–1647.

ang, L., Derksen, C., Brown, R., Markus, T., 2013. Recent changes in pan-Arctic meltonset from satellite passive microwave measurements. Geophys. Res. Lett. 40,1–7.

ang, X., Piao, S., Ciais, P., Li, J., Friedlingstein, P., Koven, C., Chen, A., 2011. Springtemperature change and its implication in the change of vegetation growth inNorth America from 1982 to 2006. Proc. Nat. Acad. Sci. U.S.A. 108 (4), 1240–1245.

ang, Q., Adiku, S., Tenhunen, J., Granier, A., 2005. On the relationship of NDVI withleaf area index in a deciduous forest site. Remote Sens. Environ. 94, 244–255.

eteorology 194 (2014) 132–143 143

Wang, K., Dickinson, R. E., Wild, M., Liang, S., 2010. Evidence for decadal caria-tion in global terrestrial evapotranspiration between 1982 and 2002: 1. Modeldevelopment. J. Geophys. Res. 115, D20112.

Wesolowski, T., Rowinski, P., 2006. Timing of bud burst and tree-leaf developmentin a multispecies temperature forest. For. Ecol. Manage. 237, 387–393.

West, T.O., Brandt, C.C., Baskaran, L.M., Hellwinckel, C.M., Mueller, R., Bernacchi, C.J.,Bandaru, V., Yang, B., Wilson, B.S., Marland, G., Nelson, R.G., Ugarte, D.G., Post,W.M., 2010. Cropland carbon fluxes in the United States: increasing geospatialresolution of inventory-based carbon accounting. Ecol. Appl. 20 (4), 1074–1086.

White, M.A., Thornton, P.E., Running, S.W., 1997. A continental phenology modelfor monitoring vegetation responses to interannual climatic variability. GlobalBiogeochem. Cycles 11, 217–234.

White, M.A., de Beurs, K.M., Didan, K., Inouye, D.W., et al., 2009. Intercomparison,interpretation, and assessment of spring phenology in North America estimatedfrom remote sensing for 1982 to 2006. Global Change Biol. 15, 2335–2359.

Wilsey, B.J., Daneshgar, P.P., Polley, H.W., 2011. Biodiversity, phenology and tempo-ral niche differences between native- and novel exotic-dominated grasslands.Evol. Syst. 13, 265–276.

Wilsey, B.J., Polley, H.W., 2006. Aboveground productivity and root-shoot allocationdiffer between native and introduced grass species. Oecologia 150, 300–309.

Wipf, S., Rixen, C., Mulder, C.P.H., 2006. Advanced snowmelt causes shift towardspositive neighbor interactions in a subarctic tundra community. Global ChangeBiol. 12, 1496–1506.

Wu, C., Munger, J.W., Niu, Z., Kuang, D., 2010. Comparison of multiple models forestimating gross primary production using MODIS and eddy covariance data inHarvard Forest. Remote Sens. Environ. 114, 2925–2939.

Xiao, X., Hagen, S., Zhang, Q., Keller, M., Moore III, B., 2006. Detecting leaf pheno-logy of seasonally moist tropical forests in South America with multi-temporalMPODIS images. Remote Sens. Environ. 103, 465–473.

Xu, L., Myneni, R.B., Chapin III, F.S., Callaghan, T.V., Pinzon, J.E., Tucker, C.J., Zhu, Z.,Bi, J., Ciais, P., Tommervik, H., Euskirchen, E.S., Forbes, B.C., Piao, S.L., Anderson,B.T., Ganguly, S., Nemani, R.R., Goetz, S.J., Beck, P.S.A., Bunn, A.G., Cao, C., Stroeve,J.C., 2013. Temperature and vegetation seasonality diminishment over northernlands. Nat. Clim. Change, http://dx.doi.org/10.1038/nclimate1836.

Yang, X., Mustard, J.F., Tang, J., Xu, H., 2012. Regional-scale phenology modelingbased on meteorological records and remote sensing observations. J. Geophys.Res. 117, G03029.

Zhang, K., Kimball, J.S., Kim, Y., McDonald, K.C., 2011. Changing freeze-thaw sea-sons in northern high latitudes and associated influences on evapotranspiration.Hydrol. Processes 25, 4142–4151.

primary production from 2000 through 2009. Science 329, 940–943.Zhao, M., Heinsch, F.A., Nemani, R.R., Running, S.W., 2005. Improvements of the

MODIS terrestrial gross and net primary production global data set. RemoteSens. Environ. 95, 164–176.