Global Interannual Variations in Sea Surface Temperature and Land Surface Vegetation, Air Temperature, and Precipitation

Sietse O. Los Science Systems and Applications, Inc., and Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt, Maryland

Search for other papers by Sietse O. Los in
Current site
Google Scholar
PubMed
Close
,
G. James Collatz Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt Maryland

Search for other papers by G. James Collatz in
Current site
Google Scholar
PubMed
Close
,
Lahouari Bounoua Department of Meteorology, University of Maryland at College Park, College Park, and Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt, Maryland

Search for other papers by Lahouari Bounoua in
Current site
Google Scholar
PubMed
Close
,
Piers J. Sellers Johnson Space Center, Houston, Texas

Search for other papers by Piers J. Sellers in
Current site
Google Scholar
PubMed
Close
, and
Compton J. Tucker Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt Maryland

Search for other papers by Compton J. Tucker in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Anomalies in global vegetation greenness, SST, land surface air temperature, and precipitation exhibit linked, low-frequency interannual variations. These interannual variations were detected and analyzed for 1982–90 with a multivariate spectral method. The two most dominant signals for 1982–90 had periods of about 2.6 and 3.4 yr. Signals centered at 2.6 years per cycle corresponded to variations in the El Niño–Southern Oscillation index and explained about 28% of the variance in anomalies of SST, land surface air temperature, precipitation, and vegetation; these signals were most pronounced in 1) SST anomalies in the eastern equatorial Pacific Ocean, 2) land surface vegetation and precipitation anomalies in tropical and subtropical regions, and 3) land surface vegetation, precipitation, and temperature anomalies in North America. Signals at 3.4 years per cycle corresponded to variations in the North Atlantic oscillation index and explained 8.6% of the variance in the combined datasets; their occurrence was most pronounced in 1) Atlantic SST anomalies, 2) in land surface temperature and vegetation anomalies in Europe and eastern Asia, and 3) in precipitation and vegetation anomalies in sub-Saharan Africa, southern Africa, and eastern North America. Anomalies in vegetation were positively related to anomalies in precipitation throughout the Tropics and subtropics and in midlatitudes in the central parts of continents. Anomalies in vegetation and temperature were positively linked in coastal temperate climates such as in Europe and eastern Asia. These associations between temperature and vegetation may be explained by the sensitivity of the length of growing season to variations in temperature.

Corresponding author address: Sietse O. Los, Science Systems and Applications Inc., and Biospheric Sciences Branch, Code 923, NASA/GSFC, Greenbelt, MD 20771.

Email: sietse@scirocco.gsfc.nasa.gov

Abstract

Anomalies in global vegetation greenness, SST, land surface air temperature, and precipitation exhibit linked, low-frequency interannual variations. These interannual variations were detected and analyzed for 1982–90 with a multivariate spectral method. The two most dominant signals for 1982–90 had periods of about 2.6 and 3.4 yr. Signals centered at 2.6 years per cycle corresponded to variations in the El Niño–Southern Oscillation index and explained about 28% of the variance in anomalies of SST, land surface air temperature, precipitation, and vegetation; these signals were most pronounced in 1) SST anomalies in the eastern equatorial Pacific Ocean, 2) land surface vegetation and precipitation anomalies in tropical and subtropical regions, and 3) land surface vegetation, precipitation, and temperature anomalies in North America. Signals at 3.4 years per cycle corresponded to variations in the North Atlantic oscillation index and explained 8.6% of the variance in the combined datasets; their occurrence was most pronounced in 1) Atlantic SST anomalies, 2) in land surface temperature and vegetation anomalies in Europe and eastern Asia, and 3) in precipitation and vegetation anomalies in sub-Saharan Africa, southern Africa, and eastern North America. Anomalies in vegetation were positively related to anomalies in precipitation throughout the Tropics and subtropics and in midlatitudes in the central parts of continents. Anomalies in vegetation and temperature were positively linked in coastal temperate climates such as in Europe and eastern Asia. These associations between temperature and vegetation may be explained by the sensitivity of the length of growing season to variations in temperature.

Corresponding author address: Sietse O. Los, Science Systems and Applications Inc., and Biospheric Sciences Branch, Code 923, NASA/GSFC, Greenbelt, MD 20771.

Email: sietse@scirocco.gsfc.nasa.gov

1. Introduction

Near-cyclical interannual variations in global precipitation and temperature result in cycles of increased and decreased vegetation photosynthesis. These climate-driven variations in vegetation photosynthesis affect the hydrological cycle, the carbon cycle, and the energy balance. Thus, understanding the couplings between climate and vegetation should lead to improved projections of the impacts of climate change on vegetation-mediated changes in land surface water, carbon, and energy fluxes. Seasonal oscillations in land surface photosynthesis and respiration result in seasonal oscillations in atmospheric carbon dioxide (CO2; Tucker et al. 1986; Fung et al. 1987). A link between interannual variations in plant photosynthesis and atmospheric CO2 therefore seems plausible.

Several studies found climate-related interannual variations in photosynthetic activity of vegetation. One line of evidence is from atmospheric CO2 observations; e.g., Keeling et al. (1995) inferred 3–7-yr and decadal cycles in carbon uptake by land surface vegetation. They related the 3–7-yr cycles to alternating states of the El Niño–Southern Oscillation. Tans et al. (1990) found evidence to support long-term sequestration of atmospheric CO2 by land surface vegetation, possibly driven by trends in climate.

Another line of evidence for climate-driven interannual variations in vegetation was established from multiyear data records collected by the Advanced Very High Resolution Radiometer (AVHRR) aboard the National Oceanic and Atmospheric Administration (NOAA) satellites. The AVHRR normalized difference vegetation index (NDVI) is near proportional to the amount of visible light absorbed for photosynthesis by land surface vegetation (Asrar et al. 1984; Sellers 1985). Climate-driven interannual variations in the NDVI were detected in the Sahel (Nicholson et al. 1998; Tucker and Nicholson 1999), in Africa (Anyamba and Eastman 1996), and in South America and Australia (Myneni et al. 1996).

Potter and Klooster (1998) suggested that large interannual variations in tropical vegetation over time should lead to increased sequestration of carbon from atmospheric CO2. Braswell et al. (1997) found lags of 2–3 yr between atmospheric CO2 concentrations, AVHRR NDVI, and satellite retrievals of tropospheric temperatures for the 1980s, and suggested that these lags were related to nutrient feedbacks and temperature sensitivity of decomposition. Randerson et al. (1999) found that differences in the rates of decomposition and photosynthesis could explain the terrestrial CO2 sink suggested by Tans et al. (1990). These studies illustrate that the interannual variation in land surface photosynthesis is likely to be an important factor in the distribution of land surface CO2 sources and sinks.

Detection of interannual variation in land surface vegetation from AVHRR NDVI data is not straightforward. Variations in AVHRR NDVI result not only from variability in land surface vegetation, but also from cloud effects, viewing geometry, variations in sensor sensitivity, solar illumination geometry, and atmospheric effects (Los et al. 1994; Sellers et al. 1996a; Gutman and Ignatov 1995; Gutman 1999). Some authors therefore question the realism of interannual variation in land surface vegetation derived from AVHRR observations (e.g., Brest et al. 1997). Others assume that the interannual signal from vegetation is significantly larger than that of interferences due to other sources. For example, Myneni et al. (1998) assumed that in high northern latitudes the effects of volcanic aerosols during 1982–83 and of solar zenith angle variations during 1982–91 were smaller than their estimates of increased vegetation greenness and lengthening of the growing season.

Malmström et al. (1997) found evidence in surface observations to support interannual variations in vegetation data. The increase in AVHRR NDVI during the 1980s was confirmed by ground measurements in some areas but not in others. They also found that corrections for sensor degradation, volcanic aerosols, and solar zenith angle effects improved realism of interannual variations in AVHRR NDVI data, and that these corrections reduced the increase in NDVI during the 1980s that was reported by Myneni et al. (1998).

The hydrological and carbon cycle impacts of interannual variations in vegetation can be assessed only if accurate estimates of this variability can be observed. The purpose of the current paper is to 1) detect statistically significant interannual variations in vegetation that are related to interannual variations in climate, and 2) explore mechanisms that explain interannual variations in vegetation. Specifically, we provide an assessment of interannual variations in the global AVHRR NDVI data as they are related to interannual variations in land surface air temperature, land surface precipitation, and sea surface temperature (SST). The analysis is based on a multivariate frequency domain decomposition (Mann and Park 1994, 1996, 1999).

2. Data

The following six global datasets were analyzed: an AVHRR NDVI dataset (Los et al. 2000); two land surface precipitation datasets: a gridded dataset and a point dataset (Eischeid et al. 1991; Baker et al. 1994); two land surface air temperature datasets: a gridded dataset and a point dataset (Jones et al. 1998); and an SST dataset (Reynolds 1988; Reynolds and Marsico 1993). Because data are from different sources—from satellite and ground measurements—we expect that errors among these datasets are uncorrelated. Analysis of common modes of interannual variation in these data should thus enhance shared signals and suppress errors.

a. FASIR NDVI

We used the 1° by 1° FASIR NDVI dataset by Los et al. (2000) [FASIR is a long acronym having to do with corrections that involve Fourier adjustment, solar zenith angle, interpolation, and reconstruction—see Los et al. (2000) and the following list]. This dataset is adjusted for 1) effects of sensor degradation in the AVHRRs aboard NOAA-7, -9, and -11 (Los 1998a); 2) volcanic aerosol effects from April 1982 until December 1983 caused by the El Chichón eruption; 3) outliers in NDVI time series that result from aerosols and clouds;4) variations in solar zenith angle; 5) missing data; and 6) severe cloud effects over tropical forests (Los 1998b;Los et al. 2000). This dataset reflects interannual variation in vegetation greenness more realistically than datasets without these corrections (Malmström et al. 1997;Los et al. 2000). Effects related to the period of sensor operation and to changes in data collection from one satellite to the next are strongly reduced. FASIR NDVI anomalies were calculated for 1982–90 as departures from the monthly mean.

b. Land surface air temperature

Monthly mean land surface air temperature anomaly data at 5° by 5° and monthly mean land surface temperature station data were obtained from the NOAA National Climate Data Center (NCDC) for 1982–90 (Jones et al. 1998). The 5° by 5° data were resampled to 1° by 1° by repeating the 5° by 5° value for 25 1° by 1° grid cells. Anomalies were recalculated to obtain zero mean for 1982–90. To increase spatial resolution in areas with a dense observation network such as in North America, Europe, and parts of East Asia, station data were averaged over 1° by 1°. Several grid cells contained more than one station, others contained one station, and a large number contained none. Anomalies were calculated from the station data for 1982–90 and, where available, they replaced the 5° by 5° data.

c. Land surface precipitation

Monthly land surface precipitation anomaly data at 5° by 5° resolution and monthly land surface precipitation station data were obtained from the NOAA NCDC for 1982–90 (Eischeid et al. 1991). The 5° by 5° data and station data were treated similar to the land surface air temperature data.

Dai et al. (1997) showed that the correlation radius for global precipitation anomalies is approximately 250 km. This will affect the results of our regression analysis in regions where we use a 5° by 5° value to represent a 1° by 1° grid cell. In our dataset, these regions are in the central parts of Asia, in deserts, and in tropical forests. The multivariate spectral analysis that is based on a singular value decomposition in the spectral domain (section 3a) is not affected by the resampling of 5° by 5° data to 1° by 1°, because the observations are simply repeated. The alternative, regridding all data to 5° by 5°, is less desirable since averaging will reduce or in extreme cases cancel interannual variations in 1° by 1° datasets (SST and NDVI), and in the precipitation and temperature data where a dense network of stations exists.

d. Sea surface temperature

The Reynolds monthly mean SST dataset for 1982–90 at 1° by 1° resolution (Reynolds 1988; Reynolds and Marsico 1993) was obtained from the Data Support Section at the National Center for Atmospheric Research. This dataset is a blend of sea surface observations from buoys, ships, and satellite. Monthly SST anomalies were calculated for 1982–90.

e. Merging global data

The global time series of FASIR NDVI, land surface air temperature, land surface precipitation, and SST were combined in three arrays (NDVI, precipitation, and temperature) of 108 monthly time steps, 360° longitude, and 180° latitude. FASIR NDVI, land surface air temperature, and land surface precipitation had one realization in one array each. SSTs had a realization in all three arrays (see Figs. 2 and 3 for an example of the mapping of the arrays). The arrays were combined in one large two-dimensional matrix of N = 108 rows (months) and M = 19440 columns (combined ocean and land surface datasets × degrees longitude × degrees latitude = 3 × 360 × 180).

Data were normalized according to
x(m)nX(m)nLaϕσaM,
where X(m)n refers to the original data, n = 1, . . . , (N = 108), m = 1, . . . , (M = 19 440), cos(ϕ) is the cosine of the latitude, and La is the number of grid cells for population a (FASIR NDVI, SST, land surface temperature, and precipitation). Scaling by the factor La/M gives each population equal weighting, regardless of size. The standard deviation σa for the entire population is given by
i1520-0442-14-7-1535-e2
where Xa are the zero-mean anomalies for population a.

A more common approach is to normalize data by grid cell, by month, or by both. This normalization is less desirable for the analysis of vegetation and rainfall data because these data have zero signals in large areas (deserts) and during long time periods (e.g., winter for the vegetation index); division by the standard deviation of a close to zero value reduces the signal-to-noise ratio.

3. Analysis

In the current section, we identify statistically significant joint interannual variation in the anomalies of FASIR NDVI, land surface air temperature, land surface precipitation, and SST (section 3a). We then add these signals and identify spatially variations in dependencies of interannual variation in vegetation anomalies on precipitation and temperature anomalies (section 3b).

a. Multispectral singular value decomposition

We used a multispectral singular value decomposition method (Mann and Park 1996) to identify areas with common frequencies in global anomaly time series of FASIR NDVI, land surface air temperature, land surface precipitation, and SST. The method by Mann and Park (1996) identifies common frequencies in spatiotemporal data regardless of phase; it also provides levels of significance for each frequency. The technique has been used in several studies to detect joint spatiotemporal modes of interannual variation (see, e.g., Park and Maasch 1993; Mann and Park 1994, 1996, 1999; Lall and Mann 1995; Koch and Mann 1996; Tourre et al. 1999). In this section we discuss parts of the techniques that are relevant for our analysis. A comprehensive discussion of the multispectral singular value decomposition is given in Mann and Park (1999).

The multispectral singular value decomposition has five components:

  1. Multitapering of the time series. In a single taper approach, leakage effects in Fourier transforms are reduced by multiplying time series with weights that gradually decrease to zero near the tails. Single tapers have the disadvantage that the signal at the tails is lost. Application of multiple, orthogonal tapers avoids leakage effects, and at the same time retains a larger proportion of the signal in the Fourier transform (Thomson 1982). We select K = 3 tapers as in Mann and Park (1996).

  2. Calculate spectra for each time series. A fast Fourier transform is applied to each of the tapered time series.
    i1520-0442-14-7-1535-e3
    where w(k)n (n = 1, . . . , N) is the kth taper, and Δt is the time step (1/12 year). The transforms of each of the tapered time series are organized by frequency in a matrix A. Matrix A has M rows (one for each location) and K columns (one for each taper):
    i1520-0442-14-7-1535-e4
  3. Isolate spatially coherent signals. Spatially coherent signals centered around a particular frequency are isolated with a complex singular value decomposition that is applied to matrix A for each frequency f (Boisvert et al. 1988):
    i1520-0442-14-7-1535-e5
    where vector uk represents the spatial empirical orthogonal functions (EOFs); vk represents the spectral EOFs, or principal modulations (Mann and Park 1999); and λk are the eigenvalues. For each narrow band around frequency f, we can express the variance explained by the kth mode relative to the variance explained by the other modes. This is referred to as local fractional variance (Mann and Park 1996). The equation is given by
    i1520-0442-14-7-1535-e6
    Figure 1a shows the local fractional variance spectrum. The local fractional variance spectrum is an extension of a single spectrum for one time series to a combined spectrum for multiple time series. In this case, it reflects a joint spectrum of interannual variation in SST, land surface NDVI, precipitation, and air temperature.
  4. Identify significant modes in the local fractional variance spectrum. Peaks in the local fractional variance spectrum above noise levels indicate potentially significant narrowband spatiotemporal signals in the data. Statistical significance of modes of interannual variation can be tested with a bootstrapping procedure (Efron 1990). The spatial data fields were left intact, but the monthly fields were permuted into semirandom sequences of the same length. This procedure was repeated 100 times, and the multitaper singular value decomposition was applied to each of the 100 realizations. The confidence levels indicate statistically significant peaks in the local fractional variance spectrum at frequencies of f = 0.375 (2.6 years per cycle) and f = 0.281 (3.4 years per cycle; see Fig. 1a). The decadal signal centered at f = 0.094 or 10.4 years per cycle falls in the secular band; that is, because of the short time series, the decadal variation cannot be distinguished from modes with lower frequencies such as a secular trend.

  5. Reconstruct statistically significant narrow band signals. At each grid point, a time domain signal was estimated from the complex valued principal modulations vk centered at frequencies f = 0.375 and f = 0.281. Linear combinations of the principal modulations vk were fitted on the original time series data with an iterative procedure (Mann and Park 1996). The statistically significant signals at frequencies f = 0.375 and f = 0.281 were estimated consecutively because they could otherwise not be separated. First the signal at (f = 0.375) was estimated (highest level of significance) and reconstructed. The reconstructed signal was subtracted from the data. A second local variance spectrum was calculated to verify that a significant signal at f = 0.281 existed after the signal at f = 0.375 was removed. This signal at f = 0.281 was then estimated and removed.

  6. Converting reconstructed signals to original units. The reconstructed signals were converted to their original units by rearranging Eq. (1):
    X(m,f)n[x(m,f)nσaMLaϕ
    with f referring to the central frequency of the reconstructed signal.

The reconstructed time domain signals for selected sites are shown in Figs. 1b and 1c. Reconstructed time domain signals at other locations may vary in magnitude and phase from these examples. Areas with small or zero magnitudes have an insignificant response at these frequencies. Note that for different sites the amplitude of the signal may be modulated; signals with otherwise the same phase may be larger or smaller for a particular time period relative to one another. The progression of the reconstructed signals of all variables investigated is shown in Figs. 2 and 3.

b. Regression analysis

The combined signal summed the signals with 2.6- and 3.4-yr cycles. A regression analysis was applied on these combined time series. The NDVI anomalies of the combined signal, ΔNDVI, were assumed to be linearly dependent on precipitation and temperature anomalies
β0β1Pβ2Tϵ,
where β0...2 are regression coefficients, ΔP are precipitation anomalies (mm), ΔT are temperature anomalies (K), and ϵ is the residual. Figure 4 shows the single-degree-of-freedom effects on the NDVI anomalies explained by anomalies in precipitation and temperature for each grid cell. The square of the single-degree-of-freedom effects is equivalent to the sum of squares explained by a particular variable.

The dependencies of vegetation variations on temperature and precipitation anomalies as found in the regression analysis are illustrated with time series of selected sites (Figs. 5–9).

4. Results

The aim of the current paper is 1) to detect common modes of interannual variation in SST and land surface NDVI, precipitation, and air temperature data and 2) to investigate how these interannual variations in climate and vegetation are related.

Two dominant modes of interannual variation were detected during the 1980s in the SST and land surface NDVI, precipitation, and air temperature (Fig. 1a). The largest mode of interannual variation with a period of about 2.6 yr explained 28% of the total variance in these datasets. This mode was strong in SSTs in the eastern equatorial Pacific Ocean and corresponds well with variations in the Southern Oscillation index during the 1980s (Fig. 1b). Because of this close resemblance during the 1980s, we refer to the 2.6-yr oscillation as ENSO oscillation.

A second mode with a period of about 3.4 yr explained about 8.6% of the total variance in these datasets. Regionally this variance can be much larger. The amplitude of the signal provides an indication of the variance explained in a region (Fig. 3). The time series from the eastern equatorial Pacific SSTs at this frequency corresponds well with cycles in the North Atlantic oscillation (NAO) index during the 1980s (Fig. 1c). We will refer to the 3.4-yr oscillation as NAO, although over longer time periods, the NAO index exhibits modes of interannual variation at timescales of 2–7 yr and at decadal and longer timescales (Hurrel 1995 1996; Van Loon and Rogers 1987). Signals with a period of about 3.4 yr were especially strong in land surface temperatures over Eurasia during the 1980s (section 4b).

There are indications for a decadal mode from other analysis of longer time series of land surface air temperature and sea level pressure that is especially strong during the 1980s (Mann and Park 1996). Because the NDVI data span a period of 9 yr, it is not reasonable to draw conclusions about decadal variations from the current analysis. Analysis of this mode is therefore not discussed further, other than to point out that it was present in SSTs in the eastern equatorial Pacific Ocean, in land surface air temperatures throughout mid and high Northern latitudes, and in NDVI over temperate regions (Europe, East Asia, and northwestern and eastern North America). The positive relationship between NDVI and temperature in these land surface areas was also found in the ENSO and NAO frequencies (section 3b). Lau and Weng (1999) found a strong decadal component in ENSO signals that explained occurrence of weaker and stronger ENSOs. Our results indicate that the “decadal mode” affected SSTs in the eastern equatorial Pacific Ocean in a similar fashion.

a. El Niño–Southern Oscillation

Figure 2 shows a sequence through the first half of a 2.6-yr cycle. ENSO-type oscillations in the Walker circulation are reflected in alternations of positive and negative SST anomalies at the equator that are contiguous with precipitation and NDVI anomalies over land. In various cases, anomalies over land have similar sign as the adjacent SST anomalies [see, e.g., SSTs in Canada and the northwest Atlantic Ocean, SSTs northwest of Australia, and NDVI (and precipitation) anomalies in central and southeast Australia in Fig. 2]. In other cases, signs of land surface anomalies and adjacent SST anomalies are opposite (e.g., positive SST anomalies in the equatorial Pacific Ocean and negative precipitation and NDVI anomalies in the Amazon). Note that anomalies in the equatorial Atlantic Ocean are negative and are contiguous with NDVI and precipitation anomalies in the Amazon (Fig. 2e). This result is in agreement with the effect of Atlantic SST variations on rainfall intensity found by Moura and Shukla (1981).

Over land, the 2.6-yr cycle had strong signals in precipitation in tropical and subtropical regions in South America, Australia, the horn of Africa, Southern Africa, and in western and central parts of North America. Vegetation anomalies are closely linked to these precipitation anomalies. Signals at 2.6 yr in air temperature were strong throughout North America and in northern parts of Eurasia. Correlations between temperature and vegetation anomalies vary from one region to the next throughout North America. The coupling between temperatures and NDVI is moderately strong in eastern and western Eurasia.

b. North Atlantic oscillation

The NAO has modes at different timescales ranging from 2–7 yr to decadal and longer cycles (Hurrel 1995, 1996). In our analysis, the 3.4-yr cycle corresponds closely to variations in the NAO during the 1980s (Fig. 1c). The land surface temperature component in Eurasia is the dominant signal at this frequency; variations in SSTs are smaller, especially when compared with the signals at 2.6 yr. Anomalies in 3.4-yr cycles of SST appear contiguous with those of precipitation and NDVI over land.

Strong relationships appear between temperature and NDVI in Europe, weaker linkages appear in east Asia;insignificant relationships between temperature and NDVI were found in central Asia. The eastern parts of North America show 3.4-yr oscillations in rainfall and NDVI, but 2.6-yr oscillations in temperature.

c. Dependencies of vegetation on precipitation and temperature

The signals at 2.6 and 3.4 yr were added to investigate the dependencies of NDVI variations on precipitation and temperature in more detail. The dependency of NDVI variations on precipitation and temperature variations was estimated with linear regression (section 3b). The spatial distribution of these dependencies are shown in Fig. 4. We highlight the following results:

  • Several high northern-latitude regions and mountainous regions (Alps, Himalayas) show a decrease in NDVI with increased precipitation (Fig. 4a) and a weak dependency of NDVI anomalies on temperature (Fig. 4b). The negative correlations between NDVI and precipitation could indicate a link between precipitation and snow cover extent, dependent on the seasonality of precipitation. Increased precipitation could lead to increased snow cover, and this could lead to decreased NDVI values because snow has an NDVI value near −0.3. A decrease in snow cover as observed during the 1980s (Groisman 1994) could thus lead to an increase in NDVI. Snow effects should be accounted for in the analysis of dependencies between vegetation and climate.

  • In temperate latitudes, regions with both positive and negative relationships between NDVI and temperature exist. In Europe NDVI increases with temperature, but also with precipitation. Weaker positive relationships between NDVI and temperature were found in coastal regions of North America, west of the Mississippi, and in western Eurasia. Negative relationships between NDVI and temperature were found in North America (central east coast and Texas) and in central Asia. These regions tend to have positive relationships between NDVI and precipitation. The positive relationship with precipitation and NDVI and negative relationship between NDVI and temperature suggests that moisture availability is limiting vegetation growth. In areas where moisture may be less limiting (e.g., in parts of Europe), NDVI tends to show strong positive relationships with temperature. It is surprising that NDVI increases are most closely linked to temperature increases in temperate latitudes and are less dependent on temperature increases in high northern latitudes reported by Myneni et al. (1998). The higher sensitivity in temperate latitudes suggests a nonlinearity in the NDVI response to temperature variations (see section 5b).

  • In tropical and subtropical regions, NDVI variations are parallel with precipitation anomalies and antiparallel with temperature. Associated with increased precipitation are increased cloudiness, which reduces incoming solar radiation, and increased moisture available for plants, which leads to increased evapotranspiration. Both effects lead to decreased temperature, which may, at least in part, explain the negative relationship between NDVI and temperature for large parts of the Tropics and subtropics. Braswell et al. (1997) also found that, in tropical areas, NDVI decreased with increasing temperature.

d. Interannual variations at selected sites

The dependencies of NDVI variations on temperature and precipitation found in Fig. 4 are discussed for selected regions.

1) Eurasia

The central northern parts of Russia (Figs. 5a–d) show inverse relationships between NDVI and precipitation anomalies at both ENSO and NAO frequencies. The anomalies in the combined temperature signals tend to be negatively related to those in the NDVI during the first part of the 1980s and positively related during the second part. The consistent relationship between NDVI and precipitation and inconsistent relationship with temperature supports our earlier notion that NDVI could be affected by snow cover, rather than by variations in vegetation cover. If true, we expect variations in precipitation (snow) during spring and fall to be causing this effect.

In Europe a very strong connection between anomalies in NDVI and land surface air temperature exists in 3.4-yr (NAO) signals (Figs. 5e–h). A positive correlation exists with smaller 2.6-yr signals in land surface temperature and precipitation. The 3.4-yr cycles explain most of the variance in the combined NDVI and temperature anomalies. The close association between temperature and NDVI, and weaker association with precipitation indicates that moisture is in general not limiting vegetation growth in Europe.

2) North America

The east coast of North America exhibits variations in temperature anomalies at ENSO timescales (Figs. 6a–d). However, unlike Europe where NDVI variations were closely linked to temperature variations, temperature variations in the East Coast seem unrelated to NDVI variations. Variations in NDVI anomalies are closely linked to those in precipitation instead, both at ENSO and NAO timescales, with NAO signals dominating (Fig. 6d).

3) South America

Nordeste (Brazil) has strong ENSO signals in anomalies of NDVI, precipitation, and temperature (Figs. 7a–d). NDVI and precipitation anomalies are positively linked (Myneni et al. 1996); temperature anomalies are negatively linked with both precipitation and NDVI anomalies. These relationships are consistent with a cooling effect of precipitation (through increased cloudiness and increased evapotranspiration) in tropical regions. The NAO component of NDVI and precipitation is similar in magnitude to the ENSO component in line with results by Moura and Shukla (1981), who suggested that rainfall in Nordeste was affected by SST variations in the Atlantic Ocean. The NAO component of temperature variations is weaker. Interference of the ENSO and NAO signals leads to stronger precipitation and NDVI effects during the first half of the 1980s and weaker effects during the second half.

4) Africa

The African continent has fairly strong signals with 3.4-yr cycles (Fig. 3). These 3.4-yr signals are also apparent in SST anomalies in the Atlantic and Indian Oceans (Fig. 3). The 3.4-yr anomalies in precipitation and NDVI are modified by signals at ENSO frequencies. This is illustrated with time series of NDVI and precipitation in the Horn of Africa (Figs. 8a–d). Combination of NAO and ENSO signals (Fig. 8d) reflects the typical drought pattern in sub-Saharan Africa during the 1980s (Tucker and Nicholson 1999).

5) Southeast Australia

Figures 9a–d indicates that both ENSO and oscillations affect precipitation and temperature anomalies. ENSO and NAO combined produce a large positive rainfall anomaly in 1984 and a drought in 1987 and cancel each other’s influence in the late 1980s and 1990. Interference of these two oscillations could provide an explanation for the unpredictability of ENSO effects in this continent. Incorporation of an NAO-type oscillation could lead to improved forecasts of droughts or rainfall events.

5. Discussion

In this paper we report statistically significant modes of interannual variation in anomalies of SST, land surface NDVI, precipitation, and air temperature and explore relationships between vegetation and climate in these signals. Several shortcomings have been reported in the NDVI data that could in principle limit the validity of the analysis. To overcome these potential problems, we used an NDVI dataset in which interferences, that is, signals from sources other than vegetation, were reduced. To increase the signal-to-noise ratio, we analyzed data from different sources—from satellite and ground stations—that are subject to different types of errors.

The observed positive relationship between temperature and NDVI anomalies in temperate coastal regions (e.g., Europe) could be an artifact in the data. Increased temperatures could be associated with decreased cloud cover, decreased cloud cover could reduce interference with the land surface signal, and this could increase the NDVI. If this were true, we would expect a negative relationship between precipitation and NDVI in these areas. However, several of the areas with a positive NDVI temperature relationship also show a positive relationship with precipitation. We do find a negative relationship between NDVI and precipitation in some tropical areas. This relationship could be caused by cloud effects, but could also be the result of a low measurement density of the precipitation data. The accuracy of the analysis is limited in areas with a low density of measurements such as in deserts and the Tropics.

a. Interannual variation

We detected two statistically significant modes of interannual variation in anomalies of SST, land surface NDVI, precipitation, and air temperature during the 1980s, one with a period of 2.6 yr that closely corresponds to the Southern Oscillation index, and one with a period of 3.4 yr that closely corresponds to the NAO. In several regions we found that these two signals interfered (South America, Africa, and Australia) and amplified or dampened interannual variations in local climate and vegetation. The ability to predict occurrences of more than one climatic oscillation could improve predictions of climatic extremes and their impact on vegetation and the hydrological and carbon cycle.

The interannual variations with periods of 2.6 and 3.4 yr detected in the multispectral analysis should result in autocorrelations with peaks at 2–3-yr lags, corresponding to the lags between temperature and NDVI reported by Braswell et al. (1997) in global time series. Hence the cross correlations between NDVI and temperature reported by Braswell et al. (1997) can be the result of oscillations in the climate system. Furthermore, the time series at the selected sites suggest smaller lags between temperature and NDVI, at least at the regional scale. Area-dependent temperature and NDVI relationship translate into biome-dependent correlations between these parameters that were found by Braswell et al. (1997). The area-dependent response of NDVI to temperature and precipitation could complicate relationships between climate, photosynthesis, and decomposition.

b. Length of growing season

The positive relationship between temperature and NDVI anomalies found in the regression analysis suggests a mechanism for increased vegetation greenness in temperate latitudes (Europe and eastern Asia, the northwest coast, and to a lesser extent the east coast of North America) because temperatures in these areas have increased over the past couple of decades (Hansen et al. 1999). The regression analysis also indicates a much smaller sensitivity of vegetation variability to substantial variations in temperature in high northern latitudes. This suggests that greenness in high latitudes may not respond as strongly to climate warming as in midlatitudes.

Several authors have indicated that variations in the length of the growing season and an earlier spring explain variability in vegetation greenness (Goulden et al. 1996; Menzel and Fabian 1999; Spiecker et al. 1996). We designed a simple test to see how length of growing season varied with temperature. We calculated the change in the number of days with mean temperature above 5°C as a result of an overall increase of 1°C (Fig. 10). The sensitivity of the length of the growing season should similarly depend on the slope of the temperature cycle during spring and fall. The areas that are sensitive to variations in the length of growing season also show a strong dependency of vegetation on temperature variations (Fig. 4b). By the same token, areas in high northern latitudes with less sensitivity in the length of growing season and a lower correlation between vegetation and temperature are expected to have a smaller increase in vegetation greenness in response to warming.

Increased greenness and increased photosynthetic capacity of vegetation results in increased gross uptake of atmospheric CO2. Changes in the rate of organic decomposition, which depends on moisture, temperature, soil carbon, and nutrient availability, can amplify, diminish, or negate a net carbon flux. For example, Goulden et al. (1996) showed that in a North American eastern temperate forest, net carbon uptake by the land surface increased in years with increased temperatures and longer growing seasons. In a boreal forest Goulden et al. (1998) found no substantial increase in photosynthetic capacity with increased temperatures—similar to our findings—but did find instead that decomposition rates increased with temperature.

In some cases decomposition lags behind an increase in vegetation greenness. This discrepancy results in a net uptake of carbon by the land surface (Randerson et al. 1999). As a result of this mechanism, areas that show a positive relationship between NDVI and temperature could, when temperatures increase, in principle contribute to interannual variations in carbon sequestration or to terrestrial carbon sink.

Increased photosynthesis as a result of increased temperatures could result in increased evapotranspiration, which counteracts the initial increase in temperature (Bounoua et al. 2000). This effect, through evapotranspirational cooling and increased cloudiness, could contribute to the smaller increases in temperatures in Europe as compared with the central parts of Eurasia (Hansen et al. 1999).

Acknowledgments

This research was funded by two NASA Earth Observing System-Interdisciplinary Science grants (Biosphere–Atmosphere Interactions project and Impacts of Interannual Climate Variability on Agro-Ecosystems project), Contracts NAS-531732 and NAS5-99085. We thank Dr. Fritz Hasler, who provided us with computing resources to perform the multivariate spectral analysis; Drs. M. Mann and J. Park, who provided computing code; Drs. W. Cramer and R. Leemans, who provided temperature data; and Dr. I. Fung and an anonymous reviewer for their comments. Dr. S. Denning and Dr. Mann made useful suggestions to improve an earlier version of this paper. Modules for the complex singular value decomposition were provided by Statlib and the Center for Applied Mathematics.

REFERENCES

  • Anyamba, A., and R. Eastman, 1996: Interannual variability of NDVI over Africa and its relation to El Niño/Southern Oscillation. Int. J. Remote Sens.,17, 2533–2548.

  • Asrar, G., M. Fuchs, E. T. Kanemasu, and J. L. Hatfield, 1984: Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. Agron. J.,76, 300–306.

  • Baker, C. B., J. K. Eischeid, T. R. Karl, and H. F. Diaz, 1994: The quality control of long-term climatological data using objective data analysis. Preprints, Ninth Conf. on Applied Climatology, Dallas, TX, Amer. Meteor. Soc., 150–155.

  • Boisvert, R. F., S. E. Howe, and D. K. Kahaner, 1988: NIST core math library (CMLIB). Center for Applied Mathematics, National Bureau of Standards, Gaithersburg, MD. [Available online at ftp://ftp.nist.gov/pub/cmlib.].

  • Bounoua, L., G. J. Collatz, S. O. Los, P. J. Sellers, D. A. Dazlich, C. J. Tucker, and D. A. Randall, 2000: Sensitivity of climate to changes in NDVI. J. Climate,13, 2277–2292.

  • Braswell, B. H., D. S. Schimel, E. Linder, and B. Moore, 1997: The response of global terrestrial ecosystems to interannual temperature variability. Science,278, 870–872.

  • Brest, C. L., W. B. Rossow, and M. D. Roiter, 1997: Update of radiance calibrations for ISCCP. J. Atmos. Oceanic Technol.,14, 1091–1109.

  • Dai, A., I. Y. Fung, and A. D. DeGenio, 1997: Surface observed global land precipitation variations during 1900–88. J. Climate,10, 2943–2962.

  • Efron, B., 1990: The Jacknife, the Bootstrap and Other Resampling Plans. Society for Applied and Industrial Mathematics, 92 pp.

  • Eischeid, J. K., H. F. Diaz, R. S. Bradley, and P. D. Jones, 1991: A comprehensive precipitation data set for global land areas. Carbon Dioxide Research Program Tech. Rep. TR051, Dept. of Energy, Office of Energy Research, 82 pp.

  • Fung, I. Y., C. J. Tucker, and K. C. Prentice, 1987: Application of Advanced Very High Resolution Radiometer Vegetation Index to study atmosphere–biosphere exchange of CO2. J. Geophys. Res.,92, 2999–3015.

  • Goulden, M. L., J. W. Munger, S. M. Fan, B. C. Daube, and S. C. Wofsy, 1996: Exchange of carbon dioxide by a deciduous forest:Response to interannual climate variability. Science,271, 1576–1578.

  • ——, and Coauthors, 1998: Sensitivity of boreal forest carbon balance to soil thaw. Science,279, 214–217.

  • Groisman, P. Y., T. R. Karl, and R. W. Knight, 1994: The impact of snow cover on the heat-balance and the rise of continental spring temperatures. Science,263, 198–200.

  • Gutman, G. G., 1999: On the use of long-term global data of land reflectances and vegetation indices derived from the Advanced Very High Resolution Radiometer. J. Geophys. Res.,104, 6241–6255.

  • ——, and A. Ignatov, 1995: Global land monitoring from AVHRR: Potentials and limitations. Int. J. Remote Sens.,16, 2301–2309.

  • Hansen, J., R. Ruedy, J. Glascoe, and M. Sato, 1999: GISS analysis of surface temperature change. J. Geophys. Res.,104, 30 997–31 022.

  • Hurrel, J. W., 1995: Decadal trends in the North Atlantic oscillation regional temperatures and precipitation. Science,269, 676–679.

  • ——, 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett.,23, 665–668.

  • Jones, P. D., D. E. Parker, T. J. Osborn, and K. R. Briffa, 1998: Global and hemispheric temperature anomalies—land and marine instrumental records. Trends: A Compendium of Data on Global Change, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory. [Available online at http://cdiac.esd.ornl.gov/trends/trends.htm.].

  • Keeling, C. D., T. P. Whorf, M. Whalen, and J. van der Plicht, 1995:Interannual extremes in the rate of rise of atmospheric carbon dioxide since 1980. Nature,375, 666–670.

  • Koch, D. M., and M. E. Mann, 1996: Spatial and temporal variability of Be-7 surface concentrations. Tellus,48B, 387–396.

  • Lall, U., and M. E. Mann, 1995: The Great Salt Lake—A barometer of low-frequency climatic variability. Water Resour. Res.,31, 2503–2515.

  • Lau, K. M., and H. Y. Weng, 1999: Interannual, decadal-interdecadal, and global warming signals in sea surface temperature during 1955–97. J. Climate,12, 1257–1267.

  • Leemans, R., and W. Cramer, 1991: The IIASA database for mean monthly values of temperature, precipitation and cloudiness of global terrestrial grid. International Institute for Applied Systems Analysis (IIASA), RR-91-18, Laxenburg, Austria.

  • Los, S. O., 1998a: Estimation of the ratio of sensor degradation between NOAA-AVHRR channels 1 and 2 from monthly NDVI composites. IEEE Trans. Geosci. Remote Sens.,36, 206–213.

  • ——, 1998b: Linkages between global vegetation and climate: An analysis based on NOAA-Advanced Very High Resolution Radiometer data. NASA Report GSFC/CR-1998-206852, 199 pp. [Available from NASA-CASI, Hanover, MD 21076-1320.].

  • ——, C. O. Justice, and C. J. Tucker, 1994: A global 1° by 1° NDVI dataset for climate studies derived from the GIMMS continental NDVI. Int. J. Remote Sens.,15, 3493–3518.

  • ——, and Coauthors, 2000: A global 9-year biophysical land-surface data set from NOAA AVHRR data. J. Hydrometeor.,1, 183–199.

  • Malmström, C. M., M. V. Thompson, G. Juday, S. O. Los, J. T. Randerson, and C. B. Field, 1997: Interannual variation in global-scale primary production: Testing model estimates. Glob. Biochem. Cycles,11, 367–392.

  • Mann, M. E., and J. Park, 1994: Global scale modes of surface temperature variability on interannual to century time scales. J. Geophys. Res.,99, 25 819–25 833.

  • ——, and ——, 1996: Joint spatio-temporal modes of surface temperature and sea-level pressure variability in the Northern Hemisphere during the last century. J. Climate,9, 2137–2162.

  • ——, and ——, 1999: Oscillatory spatiotemporal signal detection in climate studies: A multiple taper spectral domain approach. Advances in Geophysics, Academic Press, Vol. 41, 1–131.

  • Menzel, A., and P. Fabian, 1999: Growing season extended in Europe. Nature,397, 659–659.

  • Moura, J. D., and J. Shukla, 1981: On the dynamics of droughts in northeast Brazil. Observations, theory and numerical experiments with a general circulation model. J. Atmos. Sci.,38, 2653–2675.

  • Myneni, R. B., S. O. Los, and C. J. Tucker, 1996: Satellite-based identification of linked vegetation index and sea-surface temperature anomaly areas from 1982–1990 for Africa, Australia and South America. Geophys. Res. Lett.,23, 729–732.

  • ——, C. J. Tucker, G. Asrar, and C. D. Keeling, 1998: Interannual variations in satellite-sensed vegetation index data from 1981–1991. J. Geophys. Res.,103, 6145–6160.

  • Nicholson, S. E., C. J. Tucker, and M. B. Ba, 1998: Desertification, drought, and surface vegetation: An example from the West African Sahel. Bull. Amer. Meteor. Soc.,79, 815–829.

  • Park, J., and K. A. Maasch, 1993: Plio-pleistocene time evolution of the 100-kyr cycle in marine paleo-climatic records. J. Geophys. Res.,98, 447–461.

  • Potter, C. S., and S. A. Klooster, 1998: Interannual variability in soil trace gas (CO2, N2O, NO) fluxes and analysis of ontrollers on regional to global scales. Glob. Biochem. Cycles,12, 621–635.

  • Randerson, J. T., C. B. Field, I. Y. Fung, and P. P. Tans, 1999: Increase in early ecosystem uptake explain recent changes in the seasonal cycle of atmospheric CO2 at high northern latitudes. Geophys. Res. Lett.,26, 2765–2768.

  • Reynolds, R. W., 1988: A real-time global sea-surface temperature analysis. J. Climate,1, 75–86.

  • ——, and D. C. Marsico, 1993: An improved real-time global temperature analysis. J. Climate,6, 114–119.

  • Sellers, P. J., 1985: Canopy reflectance, photosynthesis and transpiration. Int. J. Remote Sens.,6, 1335–1372.

  • ——, S. O. Los, C. J. Tucker, C. O. Justice, D. A. Dazlich, G. J. Collatz, and D. A. Randall, 1996a: A revised land-surface parameterization (SiB-2) for atmospheric GCMs. Part II: The generation of global fields of terrestrial biophysical parameters from satellite data. J. Climate,9, 706–737.

  • Spiecker, H., K. Mielikäinen, M. Köhl, and J. Skovgaards, Eds., 1996:Growth Trends in European Forests; Studies from 12 Countries. Springer-Verlag, 372 pp.

  • Tans, P. P., I. Y. Fung, and T. Takahashi, 1990: Observational contraints on the global atmospheric CO2 budget. Science,247, 1431–1438.

  • Thomson, D. J., 1982: Spectrum estimation and harmonic analysis. IEEE Proc.,70, 1055–1096.

  • Tourre, Y. M., B. Rajagopalan, and Y. Kushnir, 1999: Dominant patterns of climate variability in the Atlantic Ocean during the last 136 years. J. Climate,12, 2285–2299.

  • Tucker, C. J., and S. E. Nicholson, 1999: Variations in the size of the Sahara Desert from 1980 to 1997. Ambio,28, 587–591.

  • ——, I. Y. Fung, C. D. Keeling, and R. H. Gammon, 1986: The relationship of global green leaf biomass to atmospheric CO2 concentrations. Nature,319, 195–199.

  • Van Loon, H., and J. C. Rogers, 1987: The seasaw in winter temperatures between Greenland and northern Europe, Part I: General description. Mon. Wea. Rev.,106, 296–310.

Fig. 1.
Fig. 1.

Summary of spectral analysis. (a) Local fractional variance spectrum of global anomalies in SST, land surface air temperature, FASIR NDVI, and precipitation indicates spatiotemporal modes centered at 2.6- and 3.4-yr cycles that are significant above the 95% level. A third, 10.3-yr cycle is indicated at the 90% confidence level in the secular band; this cycle is indistinguishable from a secular trend because of the short time series investigated. (b) Interannual variation centered at 2.6 years per cycle in SST anomalies at 0° lat, 90°W long (solid line) as compared with interannual variation in the Southern Oscillation index (dashed line). The 2.6-yr signal in NDVI anomalies from Nordeste (dots; same signal is shown in Fig. 7b) is shown for comparison. The SST time series shown has a phase of θ = −8.6° or a lag of −0.07 yr. The NDVI 2.6-yr anomaly signal in Nordeste shows opposite phase or a lag of about 1.3 yr. (c) Interannual variation centered at 3.4 years per cycle in SST anomalies at 0° lat, 90°W long (solid line) as compared with interannual variation in the NAO index (dashed line). The 3.4-yr signal in the NDVI anomalies from Nordeste (dots; same signal is shown in Fig. 7c) is shown for comparison. The SST time series has a phase of θ = 3.2° or a lag of 0.03 yr. The NDVI 3.4-yr signal has opposite phase or a lag of about 1.7 yr

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 2.
Fig. 2.

Progression of interannual variation at ENSO frequencies in SST anomalies over the oceans, and in anomalies of NDVI, land surface precipitation, and land surface air temperature over the land surface during the first 17 months of the 2.6-yr ENSO cycle. Each map has the same SST anomalies; over land, anomalies in NDVI, land surface precipitation, and temperature are shown, respectively. Color bar indicates anomalies normalized by the standard deviations of the respective populations: (a) t = 1 month, (b) t = 9 months, and (c) t = 17 months

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 3.
Fig. 3.

Progression of interannual variation at NAO frequencies in anomalies over oceans in SST, and in anomalies over land in NDVI, precipitation, and air temperature during the first 21 months of the 3.4-yr NAO cycle. Color bar indicates anomalies normalized by the standard deviations of the respective populations: (a) t = 1 month, (b) t = 11 months, and (c) t = 21 months

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 4.
Fig. 4.

Single-degree-of-freedom effects in reconstructed (2.6 and 3.4 yr) NDVI anomalies explained by (a) reconstructed land surface precipitation anomalies and (b) land surface air temperature anomalies. Color bar indicates an increase (red) or decrease (blue) in NDVI with an increase in either (a) precipitation or (b) temperature. White areas indicate oceans and missing data. A 3 by 3 spatial median filter was applied to enhance the image

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 5.
Fig. 5.

Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and land surface temperature anomalies (dots) from selected sites in Eurasia. (a) (b) Location of sites. (b) (f) NDVI, land surface precipitation, and temperature anomalies at ENSO (2.6 yr) cycles. (c) (g) Same as (b) and (f) but for NAO (3.4 yr) cycles. (d) (h) ENSO (b) and (f) and NAO (c) and (g) cycles combined

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 6.
Fig. 6.

Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in North America. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 7.
Fig. 7.

Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in South America. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 8.
Fig. 8.

Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from a region in eastern Africa. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 9.
Fig. 9.

Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in Australia. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Fig. 10.
Fig. 10.

Sensitivity test of the length of growing season to variations in temperature. A cubic spline was fitted through mean monthly temperature data [W. Cramer 2000, personal communication, updated dataset from Leemans and Cramer (1991)]. From the interpolated data, the difference in number of days with mean temperature above 5°C and 4°C was established. This difference in number of days indicates the sensitivity in length of growing season for an increase of 1°C. In areas where moisture is not a limiting factor, this increase provides an explanation for the sensitivity of NDVI on temperature variations (see Fig. 4)

Citation: Journal of Climate 14, 7; 10.1175/1520-0442(2001)014<1535:GIVISS>2.0.CO;2

Save
  • Anyamba, A., and R. Eastman, 1996: Interannual variability of NDVI over Africa and its relation to El Niño/Southern Oscillation. Int. J. Remote Sens.,17, 2533–2548.

  • Asrar, G., M. Fuchs, E. T. Kanemasu, and J. L. Hatfield, 1984: Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. Agron. J.,76, 300–306.

  • Baker, C. B., J. K. Eischeid, T. R. Karl, and H. F. Diaz, 1994: The quality control of long-term climatological data using objective data analysis. Preprints, Ninth Conf. on Applied Climatology, Dallas, TX, Amer. Meteor. Soc., 150–155.

  • Boisvert, R. F., S. E. Howe, and D. K. Kahaner, 1988: NIST core math library (CMLIB). Center for Applied Mathematics, National Bureau of Standards, Gaithersburg, MD. [Available online at ftp://ftp.nist.gov/pub/cmlib.].

  • Bounoua, L., G. J. Collatz, S. O. Los, P. J. Sellers, D. A. Dazlich, C. J. Tucker, and D. A. Randall, 2000: Sensitivity of climate to changes in NDVI. J. Climate,13, 2277–2292.

  • Braswell, B. H., D. S. Schimel, E. Linder, and B. Moore, 1997: The response of global terrestrial ecosystems to interannual temperature variability. Science,278, 870–872.

  • Brest, C. L., W. B. Rossow, and M. D. Roiter, 1997: Update of radiance calibrations for ISCCP. J. Atmos. Oceanic Technol.,14, 1091–1109.

  • Dai, A., I. Y. Fung, and A. D. DeGenio, 1997: Surface observed global land precipitation variations during 1900–88. J. Climate,10, 2943–2962.

  • Efron, B., 1990: The Jacknife, the Bootstrap and Other Resampling Plans. Society for Applied and Industrial Mathematics, 92 pp.

  • Eischeid, J. K., H. F. Diaz, R. S. Bradley, and P. D. Jones, 1991: A comprehensive precipitation data set for global land areas. Carbon Dioxide Research Program Tech. Rep. TR051, Dept. of Energy, Office of Energy Research, 82 pp.

  • Fung, I. Y., C. J. Tucker, and K. C. Prentice, 1987: Application of Advanced Very High Resolution Radiometer Vegetation Index to study atmosphere–biosphere exchange of CO2. J. Geophys. Res.,92, 2999–3015.

  • Goulden, M. L., J. W. Munger, S. M. Fan, B. C. Daube, and S. C. Wofsy, 1996: Exchange of carbon dioxide by a deciduous forest:Response to interannual climate variability. Science,271, 1576–1578.

  • ——, and Coauthors, 1998: Sensitivity of boreal forest carbon balance to soil thaw. Science,279, 214–217.

  • Groisman, P. Y., T. R. Karl, and R. W. Knight, 1994: The impact of snow cover on the heat-balance and the rise of continental spring temperatures. Science,263, 198–200.

  • Gutman, G. G., 1999: On the use of long-term global data of land reflectances and vegetation indices derived from the Advanced Very High Resolution Radiometer. J. Geophys. Res.,104, 6241–6255.

  • ——, and A. Ignatov, 1995: Global land monitoring from AVHRR: Potentials and limitations. Int. J. Remote Sens.,16, 2301–2309.

  • Hansen, J., R. Ruedy, J. Glascoe, and M. Sato, 1999: GISS analysis of surface temperature change. J. Geophys. Res.,104, 30 997–31 022.

  • Hurrel, J. W., 1995: Decadal trends in the North Atlantic oscillation regional temperatures and precipitation. Science,269, 676–679.

  • ——, 1996: Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett.,23, 665–668.

  • Jones, P. D., D. E. Parker, T. J. Osborn, and K. R. Briffa, 1998: Global and hemispheric temperature anomalies—land and marine instrumental records. Trends: A Compendium of Data on Global Change, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory. [Available online at http://cdiac.esd.ornl.gov/trends/trends.htm.].

  • Keeling, C. D., T. P. Whorf, M. Whalen, and J. van der Plicht, 1995:Interannual extremes in the rate of rise of atmospheric carbon dioxide since 1980. Nature,375, 666–670.

  • Koch, D. M., and M. E. Mann, 1996: Spatial and temporal variability of Be-7 surface concentrations. Tellus,48B, 387–396.

  • Lall, U., and M. E. Mann, 1995: The Great Salt Lake—A barometer of low-frequency climatic variability. Water Resour. Res.,31, 2503–2515.

  • Lau, K. M., and H. Y. Weng, 1999: Interannual, decadal-interdecadal, and global warming signals in sea surface temperature during 1955–97. J. Climate,12, 1257–1267.

  • Leemans, R., and W. Cramer, 1991: The IIASA database for mean monthly values of temperature, precipitation and cloudiness of global terrestrial grid. International Institute for Applied Systems Analysis (IIASA), RR-91-18, Laxenburg, Austria.

  • Los, S. O., 1998a: Estimation of the ratio of sensor degradation between NOAA-AVHRR channels 1 and 2 from monthly NDVI composites. IEEE Trans. Geosci. Remote Sens.,36, 206–213.

  • ——, 1998b: Linkages between global vegetation and climate: An analysis based on NOAA-Advanced Very High Resolution Radiometer data. NASA Report GSFC/CR-1998-206852, 199 pp. [Available from NASA-CASI, Hanover, MD 21076-1320.].

  • ——, C. O. Justice, and C. J. Tucker, 1994: A global 1° by 1° NDVI dataset for climate studies derived from the GIMMS continental NDVI. Int. J. Remote Sens.,15, 3493–3518.

  • ——, and Coauthors, 2000: A global 9-year biophysical land-surface data set from NOAA AVHRR data. J. Hydrometeor.,1, 183–199.

  • Malmström, C. M., M. V. Thompson, G. Juday, S. O. Los, J. T. Randerson, and C. B. Field, 1997: Interannual variation in global-scale primary production: Testing model estimates. Glob. Biochem. Cycles,11, 367–392.

  • Mann, M. E., and J. Park, 1994: Global scale modes of surface temperature variability on interannual to century time scales. J. Geophys. Res.,99, 25 819–25 833.

  • ——, and ——, 1996: Joint spatio-temporal modes of surface temperature and sea-level pressure variability in the Northern Hemisphere during the last century. J. Climate,9, 2137–2162.

  • ——, and ——, 1999: Oscillatory spatiotemporal signal detection in climate studies: A multiple taper spectral domain approach. Advances in Geophysics, Academic Press, Vol. 41, 1–131.

  • Menzel, A., and P. Fabian, 1999: Growing season extended in Europe. Nature,397, 659–659.

  • Moura, J. D., and J. Shukla, 1981: On the dynamics of droughts in northeast Brazil. Observations, theory and numerical experiments with a general circulation model. J. Atmos. Sci.,38, 2653–2675.

  • Myneni, R. B., S. O. Los, and C. J. Tucker, 1996: Satellite-based identification of linked vegetation index and sea-surface temperature anomaly areas from 1982–1990 for Africa, Australia and South America. Geophys. Res. Lett.,23, 729–732.

  • ——, C. J. Tucker, G. Asrar, and C. D. Keeling, 1998: Interannual variations in satellite-sensed vegetation index data from 1981–1991. J. Geophys. Res.,103, 6145–6160.

  • Nicholson, S. E., C. J. Tucker, and M. B. Ba, 1998: Desertification, drought, and surface vegetation: An example from the West African Sahel. Bull. Amer. Meteor. Soc.,79, 815–829.

  • Park, J., and K. A. Maasch, 1993: Plio-pleistocene time evolution of the 100-kyr cycle in marine paleo-climatic records. J. Geophys. Res.,98, 447–461.

  • Potter, C. S., and S. A. Klooster, 1998: Interannual variability in soil trace gas (CO2, N2O, NO) fluxes and analysis of ontrollers on regional to global scales. Glob. Biochem. Cycles,12, 621–635.

  • Randerson, J. T., C. B. Field, I. Y. Fung, and P. P. Tans, 1999: Increase in early ecosystem uptake explain recent changes in the seasonal cycle of atmospheric CO2 at high northern latitudes. Geophys. Res. Lett.,26, 2765–2768.

  • Reynolds, R. W., 1988: A real-time global sea-surface temperature analysis. J. Climate,1, 75–86.

  • ——, and D. C. Marsico, 1993: An improved real-time global temperature analysis. J. Climate,6, 114–119.

  • Sellers, P. J., 1985: Canopy reflectance, photosynthesis and transpiration. Int. J. Remote Sens.,6, 1335–1372.

  • ——, S. O. Los, C. J. Tucker, C. O. Justice, D. A. Dazlich, G. J. Collatz, and D. A. Randall, 1996a: A revised land-surface parameterization (SiB-2) for atmospheric GCMs. Part II: The generation of global fields of terrestrial biophysical parameters from satellite data. J. Climate,9, 706–737.

  • Spiecker, H., K. Mielikäinen, M. Köhl, and J. Skovgaards, Eds., 1996:Growth Trends in European Forests; Studies from 12 Countries. Springer-Verlag, 372 pp.

  • Tans, P. P., I. Y. Fung, and T. Takahashi, 1990: Observational contraints on the global atmospheric CO2 budget. Science,247, 1431–1438.

  • Thomson, D. J., 1982: Spectrum estimation and harmonic analysis. IEEE Proc.,70, 1055–1096.

  • Tourre, Y. M., B. Rajagopalan, and Y. Kushnir, 1999: Dominant patterns of climate variability in the Atlantic Ocean during the last 136 years. J. Climate,12, 2285–2299.

  • Tucker, C. J., and S. E. Nicholson, 1999: Variations in the size of the Sahara Desert from 1980 to 1997. Ambio,28, 587–591.

  • ——, I. Y. Fung, C. D. Keeling, and R. H. Gammon, 1986: The relationship of global green leaf biomass to atmospheric CO2 concentrations. Nature,319, 195–199.

  • Van Loon, H., and J. C. Rogers, 1987: The seasaw in winter temperatures between Greenland and northern Europe, Part I: General description. Mon. Wea. Rev.,106, 296–310.

  • Fig. 1.

    Summary of spectral analysis. (a) Local fractional variance spectrum of global anomalies in SST, land surface air temperature, FASIR NDVI, and precipitation indicates spatiotemporal modes centered at 2.6- and 3.4-yr cycles that are significant above the 95% level. A third, 10.3-yr cycle is indicated at the 90% confidence level in the secular band; this cycle is indistinguishable from a secular trend because of the short time series investigated. (b) Interannual variation centered at 2.6 years per cycle in SST anomalies at 0° lat, 90°W long (solid line) as compared with interannual variation in the Southern Oscillation index (dashed line). The 2.6-yr signal in NDVI anomalies from Nordeste (dots; same signal is shown in Fig. 7b) is shown for comparison. The SST time series shown has a phase of θ = −8.6° or a lag of −0.07 yr. The NDVI 2.6-yr anomaly signal in Nordeste shows opposite phase or a lag of about 1.3 yr. (c) Interannual variation centered at 3.4 years per cycle in SST anomalies at 0° lat, 90°W long (solid line) as compared with interannual variation in the NAO index (dashed line). The 3.4-yr signal in the NDVI anomalies from Nordeste (dots; same signal is shown in Fig. 7c) is shown for comparison. The SST time series has a phase of θ = 3.2° or a lag of 0.03 yr. The NDVI 3.4-yr signal has opposite phase or a lag of about 1.7 yr

  • Fig. 2.

    Progression of interannual variation at ENSO frequencies in SST anomalies over the oceans, and in anomalies of NDVI, land surface precipitation, and land surface air temperature over the land surface during the first 17 months of the 2.6-yr ENSO cycle. Each map has the same SST anomalies; over land, anomalies in NDVI, land surface precipitation, and temperature are shown, respectively. Color bar indicates anomalies normalized by the standard deviations of the respective populations: (a) t = 1 month, (b) t = 9 months, and (c) t = 17 months

  • Fig. 3.

    Progression of interannual variation at NAO frequencies in anomalies over oceans in SST, and in anomalies over land in NDVI, precipitation, and air temperature during the first 21 months of the 3.4-yr NAO cycle. Color bar indicates anomalies normalized by the standard deviations of the respective populations: (a) t = 1 month, (b) t = 11 months, and (c) t = 21 months

  • Fig. 4.

    Single-degree-of-freedom effects in reconstructed (2.6 and 3.4 yr) NDVI anomalies explained by (a) reconstructed land surface precipitation anomalies and (b) land surface air temperature anomalies. Color bar indicates an increase (red) or decrease (blue) in NDVI with an increase in either (a) precipitation or (b) temperature. White areas indicate oceans and missing data. A 3 by 3 spatial median filter was applied to enhance the image

  • Fig. 5.

    Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and land surface temperature anomalies (dots) from selected sites in Eurasia. (a) (b) Location of sites. (b) (f) NDVI, land surface precipitation, and temperature anomalies at ENSO (2.6 yr) cycles. (c) (g) Same as (b) and (f) but for NAO (3.4 yr) cycles. (d) (h) ENSO (b) and (f) and NAO (c) and (g) cycles combined

  • Fig. 6.

    Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in North America. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

  • Fig. 7.

    Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in South America. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

  • Fig. 8.

    Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from a region in eastern Africa. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

  • Fig. 9.

    Cycles at ENSO and NAO frequencies during the 1980s in NDVI anomalies (solid lines), land surface precipitation anomalies (bars), and temperature anomalies (dots) from selected sites in Australia. (a) Location of site. (b) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO (2.6 yr) cycles. (c) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at NAO (3.4 yr) cycles. (d) NDVI anomalies (solid line), land surface precipitation anomalies (bars), and temperature anomalies at ENSO and NAO cycles combined

  • Fig. 10.

    Sensitivity test of the length of growing season to variations in temperature. A cubic spline was fitted through mean monthly temperature data [W. Cramer 2000, personal communication, updated dataset from Leemans and Cramer (1991)]. From the interpolated data, the difference in number of days with mean temperature above 5°C and 4°C was established. This difference in number of days indicates the sensitivity in length of growing season for an increase of 1°C. In areas where moisture is not a limiting factor, this increase provides an explanation for the sensitivity of NDVI on temperature variations (see Fig. 4)

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1190 248 8
PDF Downloads 789 79 8