Abstract

The influence of the El Niño–Southern Oscillation (ENSO) on the terrestrial energy profile over North America is studied using a 15-yr model simulation. A large-area basin scale (LABs) land surface model is driven using the European Centre for Medium-Range Weather Forecasts 15-yr Re-Analyses (1979–93) dataset. It is found that the fluctuations of the soil temperature anomalies at different soil depths, in certain geographic regions, are correlated with the ENSO signal. In other words, the temperature anomaly can penetrate into the deeper soil layers due to the long wavelength associated with the ENSO signal. Using a simplified theoretical method, it is shown that the propagation of the ENSO-related long-wavelength temperature anomaly from the land surface to deep soil needs several months. In addition, it is found that the variation of the anomaly of the terrestrial enthalpy, consisting of the soil water enthalpy and soil particle enthalpy, in the shallow soil zone is dominated by the variation of the soil water storage, while that in the deep soil zone is determined by the variation of the soil temperature.

1. Introduction

Recent studies (Dirmeyer 2001; Koster and Suarez 2001) have illustrated the role of land surface moisture storage in modulating the climate anomalies induced by the sea surface temperature (SST) dynamics. Anomalous soil moisture states (wet or dry) take several weeks or longer to dissipate and impact the partitioning of energy at the land–atmosphere interface. These issues pertain to the identification of the strength of coupling between the terrestrial and the atmospheric processes and how this coupling enhances or dissipates the persistence of anomalies in the atmosphere forced externally by the SST. Short-time-scale (1 month) simulations (Dirmeyer 2001) have shown that “the impacts of suppressed ocean variability on the variability within an ensemble of monthly climate simulations are stronger than for suppressed land surface variability, suggesting the coupling between the land and atmosphere is weak in comparison with that between ocean and atmosphere.” However, even a weak coupling can modulate the anomalies driven by SST and may influence atmospheric processes over land. Such analyses are suggestive of the need for more extensive study to explore the coupling strength at different temporal and spatial scales and for longer period of time.

There is also a need to identify how the strength of the coupling varies with geographic and topographic attributes. Given that modulation of anomalies is due to the coupled responses of land surface and atmospheric processes, it is also of interest to understand how this coupling influences the terrestrial hydrologic processes, not just in terms of near-surface soil moisture, but also other hydrologically relevant variables such as total soil moisture, streamflow, temperature, etc. Using the large-area basin scale (LABs) land surface model, Chen and Kumar (2001) have established that the dynamics of the entire soil moisture profile, not just near-surface moisture, plays an extremely important role in the partitioning of precipitation into various moisture reservoirs and, consequently, that of radiation into various energy components. In a subsequent study Chen and Kumar (2002), using a 15-yr simulation with the European Centre for Medium-Range Weather Forecasts (ECMWF) 15-yr Re-Analysis (ERA-15) data for 1979–93, have identified that the moisture in the deeper soil layer plays a very active role in the terrestrial manifestation of climatic anomalies, such as El Niño–Southern Oscillation (ENSO). They hypothesized that the relatively slower dynamics of terrestrial moisture serves as a memory in causing a delayed peak in runoff as compared to the precipitation. That is, the high rainfall during an ENSO episode, may not always result in extreme runoff. The infiltrated water could get stored as soil moisture, and a lesser intense precipitation at a later time (possibly 1 month to a season delay) may cause high runoff due to the reduced soil water deficit.

Although the control of soil moisture on the near surface and atmosphere has been well explored, the relationship between climate anomalies and terrestrial temperature profile, particularly in deep soil, has received little attention. Given the large heat capacity of water it should be expected that the soil moisture dynamics plays an important role in the ENSO impacts of terrestrial temperature. By zonally and meridionally averaging surface air temperature over the continental area, Bradley et al. (1987) found that the spatial distributions of surface temperature anomalies have a tendency that is opposite of the warm and cold phases of ENSO. Using monthly mean temperatures of surface meteorological station data in North America for 1875–1980, Ropelewski and Halpert (1986) found three regions, namely, northwestern North America, the southeastern United States, and eastern Canada, that respond to ENSO events. The ENSO-related temperature anomalies in the three regions indicate that northwestern North America and the southeastern United States have well-defined “seasons” of a potentially significant ENSO-related response, but eastern Canada does not (Ropelewski and Halpert 1986). Also, Ropelewski and Halpert (1986) found that most of the ENSO-related temperature anomaly in northwestern North America shows a positive correlation with ENSO, but that in the southeastern United States it shows a negative correlation. The land surface temperature correlation pattern with ENSO, obtained using observed data from 1944 to 1996 on North America (Green et al. 1997), reveals that there exists distinct coherent correlated regions in different seasons for warm and cold ENSO events. For instance, in the warm spring of ENSO, positive temperature anomalies (warmer than normal) occur in the middle-latitude area of the continent, and negative anomalies (colder than normal) occur in Alaska, southwestern United States, Mexico, and northeastern Canada. In the cold spring of ENSO, negative anomalies appear in most of the middle-latitude area of the continent, and positive anomalies appear in the southeastern United States.

However, the relationship of ENSO with the various terrestrial energy components, such as shallow and deep soil temperature, soil enthalpy (the products of soil heat capacity, soil temperature, and soil depth), and sensible and latent heat fluxes, has not been explored in depth. In an earlier presentation, Kumar and Chen (2002) showed that the deep soil temperature anomaly, in certain geographic regions over North America, is related to the ENSO extreme events. Recent study (Xue et al. 2002) has shown that the deep soil temperature most likely has profound impacts on local climatology. Using a coupled regional climate model, they found that late spring deep soil temperature probably influences the summer precipitation over the western United States. Furthermore, Hu and Feng (2003, manuscript submitted to J. Climate) indicated that “the soil enthalpy variations may serve the role to record the winter precipitation and temperature anomalies and release their effects on summer monsoon rainfall through interactions of soil enthalpy with the surface and lower troposphere temperatures.” Prediction at the seasonal or longer time scales relies on components of the climate system that have longer memory and, therefore, can be predicted well in advance. Deep-layer terrestrial impacts may provide a basis to take advantage of such characteristics of terrestrial climate for prediction.

In this paper, we focus on the study of the impact of ENSO on temperature and enthalpy profiles of the soil over North America using the LABs land surface model (Chen and Kumar 2001) driven by the ERA-15 (1979–93) dataset (Gibson et al. 1999). We also investigate the relationship between ENSO and the land surface heat fluxes over the continent. We study how the variability of moisture affects the ENSO impact on the terrestrial enthalpy profile. This paper is organized as follows. Section 2 gives the background, which introduces the LABs model, the datasets, and the methodology for this study. The results are presented in section 3, and a summary and conclusions are given in section 4.

2. Background

a. LABs model and terrestrial enthalpy components

LABs (Chen and Kumar 2001) is a basin-scale model that implements the vertical transport of moisture and energy fluxes using the soil–vegetation–atmosphere transfer (SVAT) scheme of the National Center for Atmospheric Research (NCAR) Land Surface Model (LSM; Bonan 1996), and runoff generation using the Topmodel framework (Beven and Kirkby 1979; Beven et al. 1995). This model has been implemented over entire North America using the basin delineations of HYDRO1k data (Verdin and Verdin 1999) and validated over the Mississippi River basin using observed streamflow (Chen 2001; Chen and Kumar 2001). Each basin is represented as a single column with six soil layers (Fig. 1) for modeling the vertical moisture and energy fluxes in the soil column. The dynamics of variable source runoff generation is simulated by parameterization of the spatial distribution of the topographic index (Beven et al. 1995) using a three-parameter gamma distribution (Sivapalan et al. 1987; Kumar et al. 2000). The spatial pattern of the topographic index allows prediction of the saturated contributing area, where all rainfall contributes to runoff without any infiltration loss, as well as the unsaturated regions where Hortonian infiltration excess runoff mechanism prevails. Four land cover categories, along with a category for water body, are used to represent the heterogeneity of surface types. Each of the four land cover categories can be both in the saturated and the unsaturated regions where demand and supply limited evapotranspiration, respectively, may take place. Therefore, we have nine possible subregions to represent various land covers in each basin. As stated in Chen and Kumar (2001), the soil texture, soil color, and soil type of each basin in North America are obtained by interpolation from NCAR LSM's data file that uses the surface types of Olson et al. (1983), the soil colors of Biosphere–Atmosphere Transfer Scheme (BATS) T42 data (Dickinson et al. 1993), and the soil textures of Webb et al. (1993).

Fig. 1.

Schematic diagram of the soil column profile with six soil layers (after Bonan 1996)

Fig. 1.

Schematic diagram of the soil column profile with six soil layers (after Bonan 1996)

In LABs, the soil temperature is computed using the following one-dimensional energy diffusion equation (Hillel 1980; Bonan 1996):

 
formula

where T is the soil temperature (K), C(θ) and kT(θ) are the volumetric soil heat capacity (product of the soil heat capacity and soil density, with units J m−3 K−1) and the thermal conductivity (W m−1 K−1), respectively, and are functions of soil moisture content θ. Equation (1) is solved numerically by discretizing over the six layers (Fig. 1) with the ground heat flux, obtained from the solution of the surface energy balance, as the upper boundary condition and zero heat flux at the bottom of the soil column (see Bonan 1996 for details).

In each soil layer, the soil enthalpy consists of the soil water enthalpy (SWH) ΩSWH and the soil particle enthalpy (SPH) ΩSPH (J m−2). For soil layer l, the components of the enthalpy are computed as

 
formula

where Cwat, Cice, and Cpart are the heat capacities of water, ice, and soil particle (see Table 1), respectively, while θt(l) and Tt(l) are the volumetric soil water content and soil temperature of soil layer l at time t, and Δz(l) is the thickness of soil layer l. The volumetric water content at saturation is θsat and is computed based on the percentage of sand in the soil (Bonan 1996), and hfus is the latent heat of fusion of water content (J kg−1); ρwat is the bulk densities of water (kg m−3). The terrestrial enthalpy ΩTHl(l) for soil layer l at time t is then given as

 
formula

and the total terrestrial enthalpy obtained with six soil layers is computed as follows:

 
formula

where ΩSNHt is the land snowpack enthalpy obtained as

 
formula

Here, Csnow is the snow heat capacity (see Table 1), HSWEt is the snow water equivalent (m), and ρsnow is the bulk densities of water and snow (kg m−3). The land surface temperature assumed same as the snow temperature in the model is Tsurft. In our study, we focus on the ENSO effects on the terrestrial enthalpy (TH) in the various soil layers and, consequently, the snow enthalpy is not included but is given here for completeness. The soil air enthalpy is neglected because it is significantly smaller than the other terrestrial enthalpy components.

Table 1.

The heat capacities of water, ice, snow, and soil particle used in the LABs model (adapted from Bonan 1996): (%sand) and (%clay) are the percentages of soil texture sand and clay, respectively

The heat capacities of water, ice, snow, and soil particle used in the LABs model (adapted from Bonan 1996): (%sand) and (%clay) are the percentages of soil texture sand and clay, respectively
The heat capacities of water, ice, snow, and soil particle used in the LABs model (adapted from Bonan 1996): (%sand) and (%clay) are the percentages of soil texture sand and clay, respectively

Because the emphasis of this paper is on terrestrial energy profile, it is very important to properly model the snow-cover insulation. The snow scheme in LABs is the heritage of the NCAR LSM (Bonan 1996). In the NCAR LSM snow scheme, when snow is on the ground, the thermal properties of snow cover are blended with the first soil layer to create a snow/soil layer, and the thermal conductivity and heat capacity of this layer are decided by the depth of the snow cover and the first soil layer thickness, and the properties of snow and soil (see Bonan 1996, p. 77, for details). However, better modeling of the snow insulation is still an open research topic.

b. Methodology

In this study the LABs model is driven using the ERA-15 dataset (1979–93), and the 15-yr terrestrial hydrologic processes are simulated for all of the basins in North America. Chen and Kumar (2002) compared gridded precipitation data (Higgins et al. 1996) with the ERA-15 precipitation and found that the data quality is acceptable. Similarly, Roads and Betts (2000) also found that the ECMWF reanalysis precipitation is much closer to the observations over the Mississippi River basin than that of the (National Centers for Environmental Prediction) NCEP–NCAR reanalysis, and the NCEP–NCAR and ECMWF Re-Analysis seasonal energy components are more similar to each other. Therefore, we believe that the ERA-15 data are acceptable for our study.

The basin delineation is obtained using the HYDRO1k data based on GTOPO30 digital elevation model (DEM) data (Gesch et al. 1999; Jenson and Domingue 1988). According to them the arcs for the stream network over a continent are sorted into three classes: those that drain directly to the sea, those that drain directly into closed basins, and those that are a tributary to arcs in these first two cases (Verdin and Verdin 1999; Chen and Kumar 2001). Then, the basins are hierarchically delineated into five levels (Verdin and Verdin 1999). There are a total of 5020 level-5 basins with an average area of 3255 km2 over all of North American [for details of basin delineations, please see Verdin and Verdin (1999) and Chen and Kumar (2001)]. LABs simulation is performed for all level-5 basins. The parameters of the three-parameter gamma distribution characterizing the spatial variability of the topographic index for each level-5 basin are obtained from the GTOPO30 DEM. Using topographic index obtained from 90-m DEM data for several 1° latitude × 1° longitude grid boxes, Kumar et al. (2000) found that a simple linear relationship between the L-moments obtained at the 1-km and 90-m resolutions can be developed. Using this scheme, we use the downscaled parameters for the three-parameter gamma distribution from 1-km to 90-m resolutions [see Kumar et al. (2000) and Chen and Kumar (2001) for details]. The model time step is 30 min [for a detailed description of the application of LABs over North America, see Chen (2001) and Chen and Kumar (2001, 2002)].

The LABs model outputs are first combined into monthly aggregates. Then, the monthly time series y is separated into a monthly mean time series and an anomaly time series. The monthly mean series for a 15-yr time period is obtained using

 
formula

where, k and j are indices for year and month, respectively. The anomaly time series for the entire 15-yr period is obtained as the departure from the monthly mean series, that is,

 
formula

The sample size of the anomaly series is 180 for the 15-yr simulation.

As described in Chen and Kumar (2002), to locate the significant-correlation areas, we use the test statistic, S, to identify the confidence interval γ of the correlations between ENSO and the terrestrial hydrologic variables. The test statistic, S, is defined as (Hirsch et al. 1993)

 
formula

where S(ρ; n*) is the point on the Student's t test distribution for the effective number of degrees of freedom n* and coefficient ρ. The effective number of degrees of freedom n*, when we have two autocorrelated time series, is given as follows (Emery and Thomson 2001):

 
formula

where n is the sample size, and ρEE(τ) and ρyy(τ) are the autocorrelation coefficients of the monthly anomaly of terrestrial variable y′ and ENSO index E, respectively. Here ρEy′(τ) and ρy′E(τ) are the cross correlations between E and y′, τ is the time lag of the second series comparing to the first one, and both time series are used to compute the autocorrelation or cross correlation. In this study, the ENSO index is derived from the raw data of the National Oceanic and Atmospheric Administration (NOAA) Niño-3 (5°N–5°S, 90°–150°W) sea surface temperature (Climate Prediction Center 2001) for the base climatologic period of 1979–93 by subtracting the monthly means for 1979–93 (see Chen and Kumar 2002 for details).

In reality, the computation of n* requires the substitution of sample estimates over finite lags for the correlation (Emery and Thomson 2001). Considering the characteristics of annual cycle of terrestrial variables, we use the maximum monthly lag of 12 to estimate the effective number of degrees of freedom. If each of the n values in given samples E and y′ is statistically independent, the value of the denominator of the right-hand side of Eq. (9) is equal to 1, and n* is equal to the sample size n.

The confidence interval γ for S(ρ; n*) is then computed as

 
formula

where f(s; n*) is the probability density function of the Student's t test distribution. Combining Eqs. (8) and (10), the confidence interval γ × 100% are computed (see Table 2 for the sample size of 15). A confidence interval γ indicates that (1 − γ) × 100% of the variability is attributable to random variability.

Table 2.

Confidence intervals and associated cross correlations for 15 degrees of freedom

Confidence intervals and associated cross correlations for 15 degrees of freedom
Confidence intervals and associated cross correlations for 15 degrees of freedom

For the complete monthly anomaly series, the extremal values of the cross correlations and their associated time lags are obtained using the following method. The cross correlations are computed for lags of 0–24 months. At different lags the computations are based on (nτ) samples, where n is the maximum number of samples in a series. Consequently, the confidence interval γ for cross correlation ρ(τ) is different for each τ. We, therefore, revise the ρ(τ) to obtain ρm(τ), which has the same confidence intervals for all τ with the degree of freedom of m. This is obtained as [see appendix A of Chen and Kumar (2002) for derivation]

 
formula

The extremal value ρm(τmax) (maximum absolute cross-correlation coefficient) and the corresponding lag τmax are identified from this series. The value of m = 15 is used as the degree of freedom to study the significance of cross correlation between ENSO and each of the terrestrial variables.

3. Results

a. Teleconnection between ENSO and profiles of terrestrial temperature and enthalpy

To understand the teleconnection of ENSO with terrestrial energy components we use a two-step process. We first compute the adjusted extremal (maximum or minimum value) cross-correlation coefficient ρm(τmax) and the associated time lag τmax for each of the 5020 basins using the complete 15-yr monthly anomaly series. The regions of potential impact of ENSO are located as areas where the ρm is generally higher than surrounding regions and show a spatially coherent pattern, that is, they are not randomly distributed. Each of these regions is then used for a detailed investigation.

Figure 2 shows the spatial pattern of the extremal cross correlation between ENSO and soil temperature anomaly at each of the six layers. We see that there are six core regions around which the coherent regions are centered. These are the regions west of the Great Lakes (GL), southern Canada (SC), the region of the Pacific Northwest (PN) of the United States, the area around the Gulf of Mexico (GM), American Rocky Mountain region (RM), and western Mexico (WM; see Fig. 3). The soil temperature anomaly in first three regions (GL, SC, and PN) are positively correlated with ENSO, but in the other three regions (GM, RM, and WM) they are negatively correlated. However, the sixth-layer temperature anomaly in a small part of GM is positively correlated with ENSO. Detailed study shows that this unusual pattern of the sixth-layer temperature anomaly in the small part of GM is due to the complex interaction between the soil moisture fluctuation and the soil temperature propagation (discussed later). Three core regions (SC, GL, and GM) are located in the same area where the total soil moisture anomalies show a strong correlation with the ENSO signal (see the bottom panel of Fig. 4). It should be noticed that the propagation of the fluctuation of temperature anomalies is mostly governed by two factors [see Eq. (1)], namely, soil moisture variability and thermal conductivity of the soil. Therefore, we should expect that regions with significant soil moisture anomalies may result in temperature anomalies. However, not all regions with soil moisture anomalies may show significant temperature anomalies, partly due to the different time scales of the dynamics of soil moisture and temperature; while soil moisture dynamics is driven by gravity and evapotranspiration demand, the temperature propagation is governed by a diffusive process (Hillel 1980). In addition, it is worth noting that the area showing the deep soil temperature with long-term impacts on the land surface climate in the study of Xue et al. (2002) is around the core region PN, one of the six core regions identified here.

Fig. 2.

Extremal cross correlation between ENSO and soil temperature anomaly (STA) at each soil layer (SL). In the above six panels, the −0.5, −0.3, 0.3, and 0.5 contour lines are given. The confidence interval for the regions inside the −0.5 and 0.5 contours is greater than 95% (see Table 2)

Fig. 2.

Extremal cross correlation between ENSO and soil temperature anomaly (STA) at each soil layer (SL). In the above six panels, the −0.5, −0.3, 0.3, and 0.5 contour lines are given. The confidence interval for the regions inside the −0.5 and 0.5 contours is greater than 95% (see Table 2)

Fig. 3.

Six ENSO-related core regions for soil temperature anomalies. Three of them, PN, SC, and GL, are positively correlated, and the rest, RM, WM, and GM, are negatively correlated. The two star symbols and numbers are the locations and basin IDs of level-5 basins, which are chosen for detailed study (see Table 3)

Fig. 3.

Six ENSO-related core regions for soil temperature anomalies. Three of them, PN, SC, and GL, are positively correlated, and the rest, RM, WM, and GM, are negatively correlated. The two star symbols and numbers are the locations and basin IDs of level-5 basins, which are chosen for detailed study (see Table 3)

Fig. 4.

Extremal correlations between ENSO and (top) near-surface soil water deficit (NSWD) and (bottom) total soil water deficit (TSWD). [Adapted from Chen and Kumar (2002)]

Fig. 4.

Extremal correlations between ENSO and (top) near-surface soil water deficit (NSWD) and (bottom) total soil water deficit (TSWD). [Adapted from Chen and Kumar (2002)]

We also notice that as depth increases the spatial extent of the correlation shrinks. However, the depth of penetration of the ENSO signal is quite high for the core regions going all the way down to the center of the fifth or even the sixth layer (i.e., 2.3–4.7 m, see Fig. 1). To investigate the property of the propagation of ENSO-related soil temperature anomaly down into deep soil obtained from the above analysis, further analyses are given in sections 3b and 3c.

Figure 5 shows the extremal cross-correlation coefficient between ENSO and the ERA-15 2-m air temperature anomalies and the associated time lags. Comparing with Fig. 2 (top left), we see that the first-layer soil temperature anomaly shows a spatial pattern that is consistent with the forcing. The time lags in Fig. 5b show less lag for the negative ENSO-correlated regions than for the positive regions, the reason for which needs further investigation. Also, it is worthy noting that the correlation property of the three ENSO-related regions found by Ropelewski and Halpert (1986) is consistent with that of the ENSO-correlated regions in Fig. 5a. This may strengthen two points, that (i) the quality of the ERA-15 data of air temperature is acceptable, and (ii) the method to determine the ENSO-correlated regions is rational.

Fig. 5.

(a) Extremal correlations between ENSO and ERA-15 2-m air temperature anomalies. (b) The associated monthly lag

Fig. 5.

(a) Extremal correlations between ENSO and ERA-15 2-m air temperature anomalies. (b) The associated monthly lag

Figure 6 shows the relationship between the ENSO and anomaly of the TH [see Eq. (3)] for each of the six soil layers. For the first five layers we see that the pattern and size of the coherent regions are similar to each other, although the area shrinks gradually. However, at soil-layer six, the correlation with the TH anomaly is similar to that obtained from soil temperature anomaly (the bottom right panel of Fig. 2). This can be explained by observing that the TH is a function of both temperature and soil moisture [see Eq. (3)]. In these simulations it was observed that the water table generally lies above the sixth layer (Chen and Kumar 2001), usually in layer five. Consequently, in most of the continent, the sixth model layer usually does not undergo any moisture fluctuation and, therefore, the total enthalpy anomaly is solely a function of the temperature anomaly. However, in soil layers 1–5, both moisture and temperature are influenced by the ENSO signal and, therefore, the TH anomaly displays a more complex pattern. Comparing Figs. 6 and 4 we see that the TH anomaly correlations from layers 1 to 4 have spatial patterns that are similar to that of near-surface soil moisture deficit anomaly (the top panel of Fig. 4), and the TH correlation pattern in layer 5 closely matches that of total soil moisture deficit, suggesting that the soil moisture anomaly may play a significant role in the manifestation of the terrestrial enthalpy anomaly.

Fig. 6.

Same as Fig. 2, but for terrestrial enthalpy anomaly (THA).

Fig. 6.

Same as Fig. 2, but for terrestrial enthalpy anomaly (THA).

From Fig. 4 we see that the ENSO-correlated coherent area around California and Nevada, which exists for the near-surface soil water deficit, disappears with increasing depth. From Fig. 6, we see a similar pattern for the terrestrial enthalpy, which dies out at the fifth layer. It is known that the near-surface soil water generally relates to the fast component of rainfall response, while the total soil water includes the slow component such as baseflow (Chen and Kumar 2002). So, the shrinking of the coherent region with depth reflects that the impact on terrestrial hydrology over this region probably does not have long memory consequences as in other regions. This should be expected because the ENSO-related extra soil moisture can be easily lost to evapotranspiration before it propagates down into deep soil in this arid to semiarid region.

b. Teleconnection lags

To further verify the penetration of the ENSO-related temperature anomalies we analyze the temperature and total enthalpy time series for each layer. We selected two basins for this study (Table 3 and Fig. 3, respectively) which are positively and negatively correlated, to the ENSO signal.

Table 3.

Two level-5 basins, from positively and negatively ENSO-correlated soil temperature anomaly regions (GL and RM, see Fig. 3) over North America are chosen for a detailed study. Level-5 basin area is obtained from the HYDRO1k data (Verdin and Verdin 1999)

Two level-5 basins, from positively and negatively ENSO-correlated soil temperature anomaly regions (GL and RM, see Fig. 3) over North America are chosen for a detailed study. Level-5 basin area is obtained from the HYDRO1k data (Verdin and Verdin 1999)
Two level-5 basins, from positively and negatively ENSO-correlated soil temperature anomaly regions (GL and RM, see Fig. 3) over North America are chosen for a detailed study. Level-5 basin area is obtained from the HYDRO1k data (Verdin and Verdin 1999)

The products of ENSO and the soil temperature and terrestrial enthalpy anomalies for all six soil layers are presented in Figs. 7 and 8, and the statistical summaries are given in Table 4. The soil temperature and terrestrial enthalpy anomalies are lagged by the lag of maximum correlation τmax for each layer; that is, we look at the product E(t) × y′(t + τmax). The vertical coordinates of the products that are the same in both basins are on the same scale. The time periods of El Niño events (the positive phase of ENSO index) and La Niña events (the negative phase of ENSO index) (see the top panel of Figs. 7 and 8) are identified when the absolute value of the ENSO index (Niño-3 SST anomaly) are above 0.5°C for at least six consecutive months (Trenberth 1997).

Fig. 7.

Products time series of (top) ENSO index E(t), (middle) monthly soil temperature anomalies (STA), and (bottom) terrestrial enthalpy anomalies (THA) at six soil layers in level-5 basin ID 88253. The anomaly time series of STA and THA are lagged from the ENSO signal by τ = τmax months corresponding to the lag τmax at which the extremal cross-correlation value occurs. The vertical scale at each of six layers is same for the (middle) STA but not for (bottom) THA

Fig. 7.

Products time series of (top) ENSO index E(t), (middle) monthly soil temperature anomalies (STA), and (bottom) terrestrial enthalpy anomalies (THA) at six soil layers in level-5 basin ID 88253. The anomaly time series of STA and THA are lagged from the ENSO signal by τ = τmax months corresponding to the lag τmax at which the extremal cross-correlation value occurs. The vertical scale at each of six layers is same for the (middle) STA but not for (bottom) THA

Fig. 8.

Same as Fig. 7, but for level-5 basin ID 98980.

Fig. 8.

Same as Fig. 7, but for level-5 basin ID 98980.

Table 4.

The statistical summary of the extremal cross correlations, confidence intervals, and related lags (months) between ENSO and the anomalies of soil temperature and terrestrial enthalpy in each of six soil layers for level-5 basins 88253 and 98980

The statistical summary of the extremal cross correlations, confidence intervals, and related lags (months) between ENSO and the anomalies of soil temperature and terrestrial enthalpy in each of six soil layers for level-5 basins 88253 and 98980
The statistical summary of the extremal cross correlations, confidence intervals, and related lags (months) between ENSO and the anomalies of soil temperature and terrestrial enthalpy in each of six soil layers for level-5 basins 88253 and 98980

In level-5 basin 88253 (Fig. 7 and Table 4), we see that both soil temperature and terrestrial enthalpy anomalies are distinctly correlated with the ENSO signal from layer 1 to 6, except that in deep soil layers, the strength of this relationship becomes weak. This should be expected because the amplitude of temperature anomaly decreases with depth. However, from the time lags of the extremal cross correlations, we see that the lag of the temperature anomaly increases with depth, but not for the terrestrial enthalpy anomaly. This further confirms that the correlations of the TH anomaly from soil layers 1 to 5 are not determined by soil temperature anomaly alone. In contrast, in soil layer 6, the lag and cross correlation of the TH anomaly are the same as those of soil temperature anomaly, and this reflects that the variation of the former is dominated by the soil temperature anomaly.

In level-5 basin 98980 (Fig. 8 and Table 4), the ENSO extremal cross correlation with the soil temperature anomaly is negative and the associated time lag increases with depth. However, in layer 1, the lag is larger than those of deeper layers and the extremal correlation is significantly smaller (Table 4). This suggests that the variability of the land surface temperature anomaly can hinder the identification of the ENSO impacts but the filtering process associated with the penetration of the signal through the soil depth may make it identifiable due to the attenuation of the noise. From Table 4 we also see that in contrast to the temperature, except for the first layer, the terrestrial enthalpy anomaly does not show significant correlation. This further suggests that the TH is largely controlled by moisture variability, which is not coherent with the ENSO signal in this region (see the correlation of the total soil moisture deficit anomaly with ENSO, the bottom panel of Fig. 4). In this basin, even in soil layer 6, the soil moisture varies and the variation of the terrestrial enthalpy anomaly is different from that of the temperature anomaly.

To intuitively view the propagation of the ENSO-related anomaly, Fig. 9 shows the soil temperature anomalies in six soil layers for both level-5 basins: 88253 and 98980. All vertical coordinates for the soil temperature anomalies for the six layers in each basin are on the same scale. We find that the amplitude and variability of the soil temperature anomaly series decrease from layers 1 to 6. Decreasing of variability reflects the smoothing action of the filtering process of depth penetration mentioned above. In Fig. 9, we also see that there are three wave crests corresponding to the three El Niño events and two wave troughs corresponding to the two La Niña events during the 15-yr time period (see the thick dash lines in Fig. 9).

Fig. 9.

The time series of the (top) ENSO index and (bottom) anomalies of soil temperature for six soil layers in level-5 basin IDs (left) 88253 and (right) 98980. The dash lines schematically display the propagation of the signatures of the ENSO signal to different soil layers.

Fig. 9.

The time series of the (top) ENSO index and (bottom) anomalies of soil temperature for six soil layers in level-5 basin IDs (left) 88253 and (right) 98980. The dash lines schematically display the propagation of the signatures of the ENSO signal to different soil layers.

From Figs. 7 to 9 and Table 4, we clearly see that the ENSO-related temperature anomaly can propagate down into deep soil and needs more than half a year to propagate from the land surface to the deepest soil layer. To support this modeling result, a simplified theoretical analysis is presented below.

c. Explanation of propagation of temperature anomaly with depth

To verify that the propagation of ENSO-related temperature anomaly is not a computational artifact, we perform a theoretical analysis using simplified assumptions. Equation (1) can be solved analytically assuming a land surface sinusoidal forcing of the form T(z = 0, t) = T + A0 sinωt, where T is the average temperature, ω is the frequency and is equal to 2π/P, where P is the fluctuation period. We can obtain (after Hillel 1980, p. 168)

 
T(z, t) =
T
+ A0ez/d[sin(ωtz/d)].
(12)

The constant d is damping depth and is related to the thermal properties of the soil and the frequency of the temperature fluctuation (Hillel 1980), that is,

 
d = (2Dh/ω)1/2,
(13)

where Dh is the thermal diffusivity [kT/C, see Eq. (1) for definition of kT and C].

The range of the soil-damping depth is from 3.3 to 15.2 cm (for thermal diffusivity Dh in the range 0.4 × 10−3 to 8.4 × 10−3 cm2 s−1) for the diurnal component (i.e., P = 86400 s and ω = 7.27 × 10−5 s−1) of the soil temperature fluctuation (see Hillel 1980, Table 9.3 on p. 162). The average fluctuation period P of the ENSO signal for our study period 1979–93 (see the top of Fig. 7) is about 4.5 yr. Using Eq. (13) and ω, corresponding to P = 4.5 yr, we find that the damping depth can be from 1.34 to 6.16 m for the various soil thermal properties. Using Eq. (12), we can obtain the time for propagating to depth z as

 
formula

Using Eq. (14), the propagation of the ENSO-related soil temperature anomaly from the land surface to z = 4.7 m (center of the sixth soil layer) is 30 months and is 6 months for d = 1.34 m and d = 6.16 m (which are the damping depths for P = 4.5 yr), respectively. The propagation time for large damping depth is similar to the time lag of soil temperature anomaly obtained in the basins studied in section 3b.

According to Eq. (12), the amplitude of fluctuation at depth z is Az = A0ez/d. Therefore, at the center of the sixth soil layer, the amplitude of temperature fluctuation decreases to 0.03A0 for d = 1.34 m and to 0.47A0 for d = 6.16 m. So, for a small value of damping depth, even though the ENSO-related temperature anomaly can propagate down into deep soil, its amplitude decreases rapidly and can be obliterated easily by noise. Moreover, even for large damping depth, the amplitude decreases more than half in the lowest soil layer. The P = 4.5 yr used in this analysis does not imply that ENSO is a periodic signal, but merely is used to illustrate that a signal with a long period of fluctuation has the capacity to penetrate into deeper layers. The simplified example given here merely highlights the physical aspects needed to develop an intuitive understanding. The model study serves to illustrate how this takes place in all its complexity.

d. Teleconnection between ENSO and terrestrial energy components

1) Soil water enthalpy

Because the soil moisture storage plays a very vital role in maintaining land surface water balance (Chen and Kumar 2001) and the heat capacity of water is higher than that of soil particle (Table 1), the relationship of the ENSO signal and SWH is investigated. Two types of SWH are studied, (i) the soil water enthalpy in the near-surface soil, that is, the first soil layer with thickness 10 cm, and (ii) the water enthalpy in the entire soil column. The external cross correlation between the near-surface SWH anomaly and ENSO (Fig. 10, top) shows that there is a large spatially coherent positively correlated region in the middle-latitude region of the continent, which includes the area located in the center of the continent, and the area around the Gulf of Mexico, including the southeast United States. In addition, there is an area around California and Nevada. The associated time lags of extremal cross correlation for these regions are from zero to several months long. In contrast, there are spatially coherent negatively correlated regions in north of the continent and southern Mexico. These patterns are consistent, although not identical, with the relationships obtained from the near-surface soil water deficit (see the top panel of Fig. 4). Also, these patterns are similar to those obtained from the TH anomaly in the first layer (the top left panel of Fig. 6). Consistency of these results with those in section 3b confirms that the anomaly of the terrestrial enthalpy is dominated by that of the soil water enthalpy.

Fig. 10.

(left) Extremal cross-correlation and (right) related monthly lag of ENSO with the soil water enthalpy anomaly (SWHA) for (top) the near-surface (NS) soil and (bottom) the total soil (TS) column

Fig. 10.

(left) Extremal cross-correlation and (right) related monthly lag of ENSO with the soil water enthalpy anomaly (SWHA) for (top) the near-surface (NS) soil and (bottom) the total soil (TS) column

The bottom panels of Fig. 10 shows the relationship between ENSO and the anomaly of the soil water enthalpy in the total soil column. It shows that the ENSO correlation patterns are similar to those obtained from the total soil water deficit (see the bottom panel of Fig. 4). There are two coherent negatively correlated regions, in Canada and south Mexico, and two positively correlated regions, namely, the center of the continent and the area around the Gulf of Mexico.

Similar to the terrestrial enthalpy anomaly (Fig. 6), it shows that the size of the ENSO-related coherent regions for the total soil column dramatically decreases compared to those for the near-surface soil (see the top and bottom panels of Fig. 10). The signs of correlation in the ENSO-related coherent regions in the total soil column are the same as those in the near-surface soil.

2) Land surface heat fluxes

One of the objectives of studying the relationship between the terrestrial energy components and ENSO is to understand the impacts of climate variabilities on the partitioning of net radiation into various heat fluxes at the land–atmosphere interface. Figure 11 presents the extremal ENSO cross correlations and their associated monthly lags with the anomalies of the sensible and latent heat fluxes over North America. From the correlation between ENSO and the anomaly of sensible heat flux (Fig. 11, top), it is found that there are coherent negatively correlated regions in the American west, as well as several coherent positively correlated regions, such as the south of Mexico, New Mexico, the regions west of the Great Lakes, and the center of Canada. As for the latent heat flux anomaly (Fig. 11, bottom), there are two coherent negatively correlated regions—the center of Canada and the south of Mexico. In contrast, most of the middle-latitude land of the continent is positively correlated with ENSO.

Fig. 11.

(left) Extremal cross correlation and (right) related monthly lag of ENSO with (top) the sensible heat flux anomaly (SHA) and (bottom) the latent heat flux anomaly (LHA).

Fig. 11.

(left) Extremal cross correlation and (right) related monthly lag of ENSO with (top) the sensible heat flux anomaly (SHA) and (bottom) the latent heat flux anomaly (LHA).

Comparing the ENSO correlations with anomalies of the sensible and latent heat fluxes, we find that they generally show similar coherent correlation patterns with ENSO, but with an opposite sign, and the associated time lags of the two correlations are also similar. This should be expected, because an increase in one tends to be balanced by a corresponding decrease in the other. Moreover, the comparison of the ENSO correlations with the near-surface soil water enthalpy anomaly (Fig. 10, top) and with the latent heat flux anomaly (Fig. 11, bottom) illustrates similar ENSO-correlated regions, such as the area around California and Nevada, the Gulf of Mexico, and the south Mexico. Because the soil water variation dominates the soil water enthalpy variation, we may infer that the ENSO-related near-surface soil water anomaly most likely has significant effects on the partition of the net radiation into latent and sensible heat fluxes, and, further, affects the interaction between the land surface and atmosphere.

4. Summary and conclusions

A 15-yr (1979–93) simulation of terrestrial hydrology of North America, using the LABs land surface hydrology model (Chen and Kumar 2001), driven by ERA-15 data (Gibson et al. 1999), is used to study the impacts of the ENSO signal on the terrestrial energy components. The North American continent is divided into 5020 basins using the HYDRO1k data (Verdin and Verdin 1999) with an average basin size of 3255 km2, and the LABs model was implemented for each of these basins with a time step of 30 min. The LABs model and the 15-yr hydrologic simulation were validated in our previous studies (Chen and Kumar 2001, 2002). The model output is aggregated to monthly time scale for identifying the ENSO-related effects on terrestrial energy profiles.

There exist several ENSO-related spatially coherent correlation regions for the soil temperature anomaly over the continent, where the temperature anomaly can propagate down into deep soil. It is shown that this is theoretically plausible due to the long fluctuation period of the signal. Considering the effects of deep soil temperature anomaly on local climate from other studies (e.g., Xue et al. 2002), the results from this study suggest that understanding the deep-layer temperature propagation may have important implication in the prediction studies of regional climate.

The pattern of ENSO-related spatially coherent correlation with the terrestrial enthalpy anomaly is primarily consistent with that obtained from the soil water enthalpy anomaly, and the coherent correlation regions of ENSO and soil water enthalpy for the total soil column are similar to those of soil water deficit. This reflects that the variation of the ENSO-related soil water anomaly dominates the variations of the soil water enthalpy and, consequently, the terrestrial enthalpy anomalies.

Both ENSO-related coherent patterns for the sensible and latent heat fluxes are roughly similar but with the opposite correlation sign. Moreover, these coherent regions are approximately consistent with those from the near-surface soil water enthalpy anomaly. Considering that the ENSO-related near-surface soil water anomaly dominates the near-surface soil water enthalpy anomaly, it can be suggested that the latter has significant influence in the partitioning of net radiation into sensible and latent heat fluxes.

Due to the absence of soil temperature profile measurements we have been unable to independently validate the simulation results. However, the model performance has been validated for water fluxes (Chen and Kumar 2002). Also, we note that there were three major volcanic eruptions during the study period, Mount St. Helens, Washington, in 1980, El Chichón, Chiapas, Mexico, in 1982, and Mount Pinatubo, Luzon, Philippines in 1991. Some researches (e.g., Robock 2000) already showed that the volcanic eruptions have profound effects on the climate system at the global scale. However, in our study, we didn't interpret that anomaly meteorological forcings are only due to ENSO. In fact, we just identified the ENSO-related terrestrial hydrologic anomalies. From Figs. 7 and 8, we see that the values of the cross correlation highlight the ENSO-related temperature anomalies. Certainly, when ENSO extremes and other climate events concur, our current method will introduce certain errors in our results. This is a limitation of our current study and is in need of further research.

Acknowledgments

Support for this project has been provided by NASA Grants NAG5-3361, NAG5-8555, NAG5-7170, and NSF Grant EAR 97-06121. Computational support was also provided by NCSA Grant EAR990004N.

REFERENCES

REFERENCES
Beven
,
K. J.
, and
M. J.
Kirkby
,
1979
:
A physically based variable contributing area model of basin hydrology.
Hydrol. Sci. Bull.
,
24
,
43
69
.
Beven
,
K. J.
,
R.
Lamb
,
P. F.
Quinn
,
R.
Romanowicz
, and
J.
Freer
,
1995
:
TOPMODEL.
Computer Models of Watershed Hydrology, V. P. Singh, Ed., Water Resources Publications, 627–668
.
Bonan
,
G. B.
,
1996
:
A land surface model (LSM version 1.0) for ecological, hydrological, and atmospheric studies: Technical description and user's guide.
NCAR Tech. Note NCAR/TN_417+STR, 150 pp. [Available online at http://www.cgd.ucar.edu/tss/clm/.]
.
Bradley
,
R. S.
,
H. F.
Diaz
,
G. N.
Kiladis
, and
J. K.
Eischeid
,
1987
:
ENSO signal in continental temperature and precipitation records.
Nature
,
327
,
497
501
.
Chen
,
J.
,
2001
:
Influence of climate variability on terrestrial hydrology in North America.
Ph.D. thesis, University of Illinois at Urbana–Champaign, 218 pp
.
Chen
,
J.
, and
P.
Kumar
,
2001
:
Topographic influence on the seasonal and interannual variation of water and energy balance of basins in North America.
J. Climate
,
14
,
1989
2014
.
Chen
,
J.
, and
P.
Kumar
,
2002
:
Role of terrestrial memory in modulating ENSO impacts in North America.
J. Climate
,
15
,
3569
3585
.
Climate Prediction Center
,
cited 2001
:
Monthly atmospheric and SST indices.
.
Dickinson
,
R. E.
,
A.
Henderson-Sellers
, and
P. J.
Kennedy
,
1993
:
Biosphere–Atmosphere Transfer Scheme (BATS) version 1e as coupled to the NCAR Community Climate Model.
NCAR Tech. Note NCAR/TN-387+STR, National Center for Atmospheric Research, 72 pp
.
Dirmeyer
,
P. A.
,
2001
:
An evaluation of the strength of land–atmosphere coupling.
J. Hydrometeor.
,
2
,
329
344
.
Emery
,
W. J.
, and
R. E.
Thomson
,
2001
:
Data Analysis Methods in Physical Oceanography. 2d ed.
Elsevier Press, 638 pp
.
Gesch
,
D. B.
,
K. L.
Verdin
, and
S. K.
Greenlee
,
1999
:
New land surface digital elevation model covers the Earth.
Eos, Trans. Amer. Geophys. Union
,
80
,
69
70
.
Gibson
,
J. K.
,
P.
Kållberg
,
S.
Uppala
,
A.
Hernandez
,
A.
Nomura
, and
E.
Serrano
,
1999
:
ERA-15 description (Version 2).
ECMWF Re-analysis Project Report Series, Part 1, ECMWF, 73 pp
.
Green
,
P. M.
,
D. M.
Legler
,
C. J.
Miranda V
, and
J. J.
O'Brien
,
1997
:
The North American climate patterns associated with the El Niño–Southern Oscillation.
COAPS Project Report Series 97-1, 8 pp. [Available online at http://www.coaps.fsu.edu/lib/booklet/.]
.
Higgins
,
R. W.
,
J. E.
Janowiak
, and
Y. P.
Yao
,
1996
:
A Gridded Hourly Precipitation Database for the United States (1963–1993).
NCEP/Climate Prediction Center Atlas No. 1, NCEP, 47 pp
.
Hillel
,
D.
,
1980
:
Introduction to Soil Physics.
Academic Press, 364 pp
.
Hirsch
,
R. M.
,
D. R.
Helsel
,
T. A.
Cohn
, and
E. J.
Gilroy
,
1993
:
Statistical analysis of hydrologic data.
Handbook of Hydrology, D. R. Maidment, Ed., McGraw-Hill, 17.1–17.55
.
Jenson
,
S.
, and
J.
Domingue
,
1988
:
Extracting topographic structure from digital elevation data for geographic information system analysis.
Photogramm. Eng. Remote Sens.
,
54
,
1593
1600
.
Koster
,
R. D.
, and
M. J.
Suarez
,
2001
:
Soil moisture memory in climate models.
J. Hydrometeor.
,
2
,
558
570
.
Kumar
,
P.
, and
J.
Chen
,
2002
:
Influence of hydrologic memory on terrestrial impacts of ENSO.
Preprints, 16th Conf. on Hydrology, Orlando, FL, Amer. Meteor. Soc., CD-ROM, J4.8
.
Kumar
,
P.
,
K. L.
Verdin
, and
S. K.
Greenlee
,
2000
:
Basin level statistical properties of topographic index for North America.
Adv. Water Resour.
,
23
,
571
578
.
Olson
,
J. S.
,
J. A.
Watts
, and
L. J.
Allison
,
1983
:
Carbon in live vegetation of major world ecosystems.
Oak Ridge National Laboratory Tech. Rep. ORNL-5862, 164 pp
.
Roads
,
J.
, and
A.
Betts
,
2000
:
NCEP–NCAR and ECMWF reanalysis surface water and energy budgets for the Mississippi River basin.
J. Hydrometeor.
,
1
,
88
94
.
Robock
,
A.
,
2000
:
Volcanic eruptions and climate.
Rev. Geophys.
,
38
,
191
219
.
Ropelewski
,
C. F.
, and
M. S.
Halpert
,
1986
:
North American precipitation and temperature patterns associated with the El Niño/Southern Oscillation (ENSO).
Mon. Wea. Rev.
,
114
,
2352
2362
.
Sivapalan
,
M.
,
K. J.
Beven
, and
E. F.
Wood
,
1987
:
On hydrologic similarity, 2, A scaled model of storm runoff production.
Water Resour. Res.
,
23
,
1289
1299
.
Trenberth
,
K. E.
,
1997
:
The definition of El Niño.
Bull. Amer. Meteor. Soc.
,
78
,
2771
2777
.
Verdin
,
K. L.
, and
J. P.
Verdin
,
1999
:
A topological system for delineation and codification of the Earth's river basins.
J. Hydrol.
,
218
,
1
12
.
Webb
,
R. S.
,
C. E.
Rosenzweig
, and
E. R.
Levine
,
1993
:
Specifying land surface characteristics in general circulation models: Soil profile data set and derived water-holding capacities.
Global Biogeochem. Cycles
,
7
,
97
108
.
Xue
,
Y. K.
,
L.
Yi
,
M.
Ruml
, and
R.
Vasic
,
2002
:
Investigation of deep soil temperature–atmosphere interaction in North America.
Preprints, 13th Symp. on Global Change and Climate Variations, Orlando, FL, Amer. Meteor. Soc., CD-ROM, J6.5
.

Footnotes

Corresponding author address: Dr. Praveen Kumar, Environmental Hydrology and Hydraulic Engineering, Department of Civil and Environmental Engineering, University of Illinois at Urbana–Champaign, 205 North Matthews Avenue, Urbana, IL 61801. Email: kumar1@uiuc.edu