Journal of Geophysical Research: Atmospheres

Influence of sea surface temperature on soil moisture and precipitation interactions over the southwest

Authors


Abstract

[1] This paper presents a hypothesis that soil moisture (SM) and precipitation (P) interactions over the Southwest depend on sea surface temperature anomalies (SSTAs). On the basis of moisture transport and geography, the Southwest can be separated into two regions. The western region (32°–36°N, 107.5°–113°W) includes Arizona and western New Mexico and the eastern region (32°–36°N, 103°–107°W) includes eastern New Mexico. For both regions, years from 1900 to 2004 are classified based on the winter to summer P evolution. When winter and the following summer P anomalies have an inverse relationship, SSTAs do not persist. The summer SSTAs have strong influence on P. Soil moisture does not play a major role in modulating P anomalies. For cases that wetness (dryness) occurs in both winter and the following summer, the SSTA forcing associated with the P regime tends to persist. Positive SM-P feedbacks enhance P anomalies. For eastern New Mexico, there is a linear relationship between SM anomalies in spring and P anomalies in summer when SSTAs persist.

1. Introduction

[2] The antecedent land surface conditions on monsoon rainfall over the Southwest have been the focus of many studies, but they often offer contrasting conclusions. Matsui et al. [2003] examined relationships among snow cover, skin temperature and monsoon precipitation using satellite data. An inverse relationship between rainfall and surface temperature over the monsoon region suggests positive soil moisture (SM) and precipitation (P) feedbacks. Findell and Eltahir [2003] examined the convective triggering potential and low level humidity index under different soil conditions. They found positive feedbacks between soil moisture and moist convection in summer. The only area showing possible negative feedback is located in the dryline and monsoon region over the Southwest. They emphasized the importance of atmospheric forcing. Later, Zhu et al. [2005] proposed a theory to explain the land-atmosphere interactions over the Southwest. Wet winter and spring snowmelt increase soil moisture in spring. The increase of soil moisture lowers surface temperature, which leads to a dry summer. The hypothesis is consistent with the inverse relationship between winter and summer rainfall over the Southwest [Higgins et al., 1999] and the negative correlations between spring snowpack and summer monsoon [Gutzler and Preston, 1997]. However, this theory does not explain these years that P anomalies persist from winter to summer. Zhu et al. [2005] did not find strong connections between surface temperature and monsoon rainfall. They concluded that atmospheric circulation anomalies may play a role.

[3] Large scale circulation is often influenced by both sea surface temperature (SST) and soil moisture anomalies. Many authors found that both SSTs in the North Pacific and the tropical eastern Pacific influence monsoon rainfall [Higgins and Shi, 2000;Mo and Paegle, 2000; Castro et al., 2001]. The inverse relationship between winter and summer P is more likely to occur if the SST anomalies (SSTAs) in the eastern Pacific are strong. The persistence of SSTAs in the North Pacific paves the way for winter P anomalies to persist through summer. In addition to SSTAs, soil conditions can also modulate P locally and directly through evaporation (E) from surface or by triggering local deep convection. SM can also influence P indirectly through changes in circulation anomalies and moisture transport [Kanamitsu and Mo, 2003].

[4] It is clear that the combination of SST and SM modulates rainfall over the Southwest. The question is whether SM and P interact differently under different SSTA forcing. The regional experiments performed by Kanamitsu and Mo [2003] indicate that moisture is transported into the Southwest by the low level jet through the Gulf of California (GCLLJ). The transported moisture plays a major role in maintaining soil moisture in the region. If there is little or no moisture transported into the Southwest, then SM will diminish due to evaporation and will not influence rainfall on seasonal timescales. The GCLLJ is often modulated by SSTAs. This suggests that the SM and P relationships may depend on SSTAs.

[5] Recently, the data from the North American Regional Reanalysis (RR) [Mesinger et al., 2006] and the Variable Infiltration Capacity (VIC) Land Data Assimilation (NLDAS) data [Andreadis et al., 2005] from the University of Washington were made available. This provides an opportunity to examine the SM and P relationships. The purpose of this paper is to examine the hypothesis that SM and P over the Southwest interact differently under different SST forcing.

[6] The Southwest is separated by moisture sources and geography into two regions. The west region covers eastern and central Arizona and western New Mexico (w-AZNM, 32°–36°N, 107.5°–113°W). The east region covers eastern New Mexico (e-AZNM, 32°–36°N, 103°–107°W). During winter, moisture transport comes from the North Pacific for both regions. In summer, moisture is transported into w-AZNM by the GCLLJ. The flow from the Gulf of Mexico does not play an important role. On the contrary, the major moisture source for e-AZNM comes from the Gulf of Mexico with some contributions from the North Pacific. Moisture is transported into the area by the Great Plains low level jet (GPLLJ). The GCLLJ does not influence moisture supply to e-AZNM [Mo and Berbery, 2004]. The influence of SSTAs and soil moisture for two regions will be discussed separately. Data sets are described in section 2. The seasonal cycles of SM, E and P are examined in section 3. Years are separated into two categories based on the winter to summer rainfall evolution. The SM and P feedbacks for two regions are discussed in section 4. Conclusions are given in section 5.

2. Data

[7] The study period covers 1900–2004. E, surface temperature and total SM were taken from the VIC model. The VIC is a macroscale hydrologic model [Liang et al., 1994, 1996]. The same data were used by Andreadis et al. [2005] to study droughts. The resolution is 0.5° and it covers the period from 1915–2003. The top soil layer is fixed at 10 cm. The soil depths of the second and third layers vary. The precipitation used to force VIC was derived from the cooperative observer station meteorological daily data. Station data were gridded using the same method described by Maurer et al. [2002]. P was adjusted for elevation using on the Precipitation Regression on Independent Slopes Method (PRISM) [Daly et al., 1994].

[8] The climate division precipitation data were used to select events. There are 344 climate divisions over the United States from 1900–2004 and there is no PRISM adjustment. The VIC P forcing and P from the climate divisions do not differ much over the Southwest. The correlation between monthly mean P anomalies from the two data sets for July–September (JAS) from 1915–2003 is 0.93 for w-AZNM and 0.92 for e-AZNM. P forcing from VIC was used to form composites to assure consistency with composites of E and SM.

[9] The North American Regional Reanalysis (RR) [Mesinger et al., 2006] was used to describe the climatology. The RR has 32 km resolution, but only covers the period from 1979–2006. The RR assimilates P from the Climate Prediction Center operational unified rain gauge analysis over the United States and Mexico [Higgins et al., 2000]. Data from the NCEP-NCAR reanalyses (R1) [Kalnay et al., 1996] were used to form composites of circulation anomalies. The R1 only covers the period from 1949–2004. The sea surface temperature (SST) data are the monthly reconstructed SST from Smith et al. [1996]. The data set covers the period from 1900–2004. Climatological monthly means were removed from each data set to obtain monthly mean anomalies.

3. Seasonal Cycle

[10] Figure 1 shows the climatological monthly means for P, E, runoff and SM averaged over w-AZNM and e-AZNM for the period from 1979–2004. VIC uses observed P as forcing and RR assimilates P. Therefore monthly means from two data sets are similar (open circles). Evaporations depicted by the VIC (crosses) and RR (solid line) are also similar (Figures 1a1d). Runoff (Figures 1c and 1f) means are small in comparison to P and E. The two systems have different soil structure and soil layers. VIC (crosses) has lower total water storage than the RR (solid line) (Figures 1b and 1e). The important point is that both SM climatologies show similar monthly evolution. If the climatology for VIC from 1915–2003 is used, conclusions remain the same.

Figure 1.

(a) Climatological monthly mean P (open circles) and E from RR (solid line) and E from VIC (crosses) averaged for w-AZNM (32–36°N, 107.5–113°W) for the period from 1979–2004. The unit is mmd−1, (b) same as Figure 1a, but for total soil water storage (SM) from the RR (solid line) and the VIC (crosses). The unit is mm, (c) same as Figure 1b, but for runoff. The unit is mmd−1. (d)–(f) The same as Figure 1a–1c, but for e-AZNM (32–36 °N,103–107°W).

[11] For both regions, the SM maximum occurs in March due to heavy rainfall and low evaporation in autumn and winter. Heavy raining season is defined as P greater than 1 mm d−1. The increase of E in spring and relatively low precipitation cause SM to decrease from March onward. SM reaches a minimum in the beginning of summer. For summer, monsoon rainfall is balanced by large E so SM does not increase much.

[12] There are significant differences between the two regions. For w-AZNM, there are 2 P maxima in February and August respectively. Runoff also has 2 maxima in March and August, but magnitudes are small. Spring is a dry season. The increase of E leads to the decrease of SM in spring. SM reaches a minimum just before the monsoon onset in early July. SM does not increase substantially because P and E have similar magnitudes during summer. SM starts to increase after the winter rainfall season.

[13] For e-AZNM, SM has a very weak seasonal cycle in comparison to w-AZNM. The monsoon season lasts from May to October with a maximum in August. Runoff is very small. Because the differences between E and P are small after June, the seasonal cycle is weak (Figure 1e).

[14] Figure 2 plots the summer (July–September JAS) (dark circles) and winter (January–March JFM) (open circles) mean rainfall anomalies normalized by its own standard deviation for two regions respectively. The correlation between summer and winter rainfall is −0.13 (0.14) for w-AZNM (e-AZNM). Assuming 105 degrees of freedom, correlations greater than 0.19 (less than −0.19) are statistically significant at the 5% level. As noticed by Zhu et al. [2005], and Mo and Paegle [2000], correlations change with time. The strongest correlation occurred during the period from 1963–1990 for w-AZNM (−0.48) and e-AZNM (−0.42). These correlations are statistically significant at the 5% level. The correlation between w-AZNM and e-AZNM is 0.21 which is barely statistically significant.

Figure 2.

Normalized winter (January–March JFM) (open circles) and summer (July–September JAS) (dark circles) P anomalies averaged over (a) w-AZNM, and (b) e-AZNM.

4. SM-P Relationships for Different SST Modes

[15] According to the relationships between winter P (JFM) and summer P(JAS) anomalies (Figure 2), years are classified into two cases. Wet (dry) winters followed by dry (wet) summers are classified as positive (negative) case 1 events. Another requirement is that the magnitudes of rainfall anomalies for both seasons have to be above 0.5 standard deviations (std). This requirement is added to exclude years with normal rainfall. Small changes of this threshold will not change conclusions. For case 1 events, the criteria for positive events are P(JFM) > 0.5 std and P(JAS) < −0.5 std. There were total 10 positive events and 7 events occurred after 1948 for w-AZNM. There were total 7 positive events and 6 of them occurred after 1948 for e-AZNM. The criteria for negative events are P(JFM) < −0.5 std and P(JAS) > 0.5 std. There were total 14 negative case 1 events and 9 events occurred after 1948 for w-AZNM. There were total 10 negative case 1 events and 4 events occurred after 1948 for e-AZNM. Wet (dry) winters followed by wet (dry) summers are classified as positive (negative) case 2 events. There were total 8 positive and 16 negative case 2 events for w-AZNM, but only 2 positive and 6 negative events occurred after 1948. There were total 12 positive and 16 negative case 2 events for e-AZNM. Among them, 6 positive and 7 negative events occurred after 1948. Table 1 lists all events for both cases.

Table 1. List of Case 1 and Case 2 Events for w-AZNM and e-AZNM
w-AZNMe-AZNM
Case 1Case 2Case 1Case 2
PositiveNegativePositiveNegativePositiveNegativePositiveNegative
19051921191119001948190419111902
19201923191619021973190619151910
19371925193019031978191419191917
19731929193119101983192519311922
19781946193519241987192719321928
19791955194119281992194219411934
19801963198319342001195019491943
1991196419921942 196719581945
19931967 1947 197119751947
19951971 1948 197219901953
 1984 1950  19971954
 1988 1953  20041956
 1996 1956   1959
 1999 1972   1964
   2000   1970
   2002   2003

[16] For each case, each season and each region, composites of SST, P, E, SM, and surface temperature (Tsurf) anomalies for positive and negative events were obtained based on selected events (Table 1). Overall, composites based on positive and negative events are similar with a sign reversal so the differences between positive and negative composites are presented. The VIC data from 1915 were used to form SM, E and P composites. The R1 data from 1949–2004 were used to form composites of 200 hPa eddy stream function anomalies with the zonal means removed and 850 hPa height anomalies. For the w-AZNM case 2, there are not enough events to obtain meaningful statistics and they are not shown. The statistical significance is based on a normal distribution and one degree of freedom per year. Areas that differences are statistically significant at the 5% level are shaded.

4.1. Case 1

4.1.1. SST Composites

[17] For w-AZNM, the SST composite (Figure 3a) for case 1 indicates that wet (dry) winter in w-AZNM is associated with warm (cold) SSTAs extending from the tropical central Pacific to the eastern Pacific and along the west coast of North America. Negative (positive) SSTAs are located along 30°N in the North Pacific. The SSTAs persist through AMJ (Figure 3b). In summer, SSTAs in the central Pacific along the equator diminish but SSTAs in the eastern Pacific north of the Equator remain (Figure 3c). There is also a dipole of SSTAs in the North Atlantic. As suggested by Higgins et al. [1999], warm (cold) SSTAs in the eastern Pacific are responsible for dry (wet) monsoon rainfall in summer. Results are also consistent with Mo and Paegle [2000] even though their domain for w-AZNM is larger.

Figure 3.

Composite difference of SST anomalies between 10 positive and 14 negative w-AZNM case 1 events for (a) winter (JFM), (b) spring (April–June AMJ), and (c) summer (JAS). Contour interval is 0.3°C. Zero contours are omitted. Contours −0.1 and 0.1°C are added for Figure 3c. Positive (negative) values that are statistically significant at the 5% levels are shaded dark (light), (d)–(f) same as Figures 3a–3c, but for 7 positive and 10 negative e-AZNM case 1.

[18] For e-AZNM case 1 (Figures 3d3f), wet (dry) winter is associated with warm (cold) SSTAs along the equator extending from the central Pacific to the eastern Pacific. Warm (cold) SSTAs are also located in the Atlantic along the African coast and in the Indian Ocean. SSTAs in the central Pacific diminish with time. The composite difference for summer shows warming (cooling) in the tropical North Atlantic, the South Atlantic, the Indian Ocean, and the western Pacific. The SSTA pattern resembles the decadal SSTA mode that shows warming trends after 1990. That mode is responsible for the decreasing trends of rainfall over eastern New Mexico [Mo and Schemm, 2008]. That is also consistent with the uneven distribution of case 1 events. All negative events occurred before 1972 and there were more positive events after 1973 (Table 1).

4.1.2. Circulation Features

[19] For both case 1 events, SSTAs do not persist from winter to summer. Circulation anomalies and P anomalies respond to the contemporaneous SSTA forcing. For w-AZNM, the winter response to positive SSTAs in the tropical Pacific is a dipole straddling the equator in the central and eastern Pacific (Figure 4a). The north-south dipole located over the west coast of the United States signals the eastward extension of the North American subtropical jet and that is consistent with wet winter over w-AZNM. The summer composite (Figure 4b) shows a dipole in the eastern Pacific with a wave train with positive anomalies centered at (50°N, 150°W), negative anomalies over the w-AZNM and positive anomalies to the east. Warm SSTAs in the tropical eastern Pacific are conducive to rising motion and convection. The negative anomalies over the w-AZNM area are associated with the downward branch of the rising motion [Mo and Paegle, 2000].

Figure 4.

Composite difference of 200 hPa eddy stream function anomalies between 7 positive and 9 negative (a) w-AZNM case 1 events for JFM. Contour interval is 2 × 105 m2 s−1. Zero contours are omitted. Positive (negative) values that are statistically significant at the 5% levels are shaded dark (light), (b) same as (Figure 4a), but for JAS. Contour interval is 1 × 105 m2 s−1. (c)–(d) same as Figures 4a– 4b, but for 6 positive and 4 negative e-AZNM case 1.

[20] The winter composite for e-AZNM shows a dipole straddling the Equator in the eastern Pacific. A Pacific North American (PNA) like wave train extends from the Gulf of Alaska to North America with negative anomalies located over the southern United States (Figure 4c). For JAS, the phase of the wave train (Figure 4d) located in the PNA region is in quadrature with the wave train associated with the w-AZNM case (Figure 4b).

4.1.3. P and SM

[21] The monthly mean composites of P and SM anomalies for w-AZNM averaged over case 1 events are presented in Figure 5. Confidence bands based on standard deviations are represented by thin lines which indicate the uncertainties of composites. For both regions, positive (negative) P anomalies in winter are followed by large positive (negative) SM anomalies in spring. After March, SM anomalies start to decrease. The following description is based on positive cases. For negative events, the situation reverses.

Figure 5.

(a) Composite difference of monthly mean P averaged over w-AZNM domain between 9 positive and 14 negative w-AZNM case 1 events (open circles). Thin lines are standard deviations. The unit is mm d−1, (b) same as (Figure 5a), but for SM anomalies. The unit is mm. (c)–(d) same as Figures 5a–5b, but for 7 positive and 12 negative w-AZNM case 2, (e)–(f) same as Figures 5a–5b, but for 7 positive and 7 negative case 1 events for e-AZNM, (g)-(h) same as Figures 5e–5f, but for 11 positive and 14 negative case 2 events for e-AZNM.

[22] Summer rainy season begins in July (Figure 1) in w-AZNM. For positive case 1 (Figures 5a and 5b), P anomalies decrease sharply from weak positive in June to a negative July minimum. P anomalies stay negative for the entire monsoon season from July to September. SM anomalies are positive in winter and early spring due to positive P anomalies. They start to decrease from March onward, but they are still positive in July when P anomalies are already strong negative. The SM anomalies turn to small negative in August and reach a minimum in October near the end of monsoon season. SM anomalies lag P anomalies by at least one month. This suggests that SM anomalies respond to P anomalies.

[23] The situation for e-AZNM case 1 is similar (Figures 5e and 5f). The rainy season starts in May–June (Figure 1). P anomalies turn negative in June but SM anomalies remain positive. SM anomalies turn to small negative in July. The one month lag between SM and P anomalies suggests that SM anomalies respond to P. P anomalies reach a minimum in June and stay negative during summer. SM anomalies turn to weak negative in July and reach a minimum in September while P anomalies start to increase.

[24] For positive case 1, wet winters increase SM anomalies. Positive soil moisture anomalies persist into July for w-AZNM and June for e-AZNM. Wet SM anomalies lead to higher evaporation. Positive E anomalies should contribute to positive P anomalies. However, circulation anomalies associated with SSTAs in summer are in favor of a dry summer. The impact of SM cancels out some of the influence of SSTAs, but is not strong enough to overcome the influence of SSTAs. P starts to decrease. Less P leads to less SM in the area. SM anomalies reach a minimum in September or October. By then the monsoon season comes to the end. For case 1 events, SSTAs modulate P and SM anomalies respond to P.

4.2. Case 2

4.2.1. SST Composites

[25] The SSTA composites for case 2 evolve very differently from case 1. For w-AZNM case 2 (Figure 6a6c), wet (dry) winter corresponds to negative (positive) SSTAs in the North Pacific centered at 35°–40°N and warm (cold) SSTAs along the west coast of the United States. There is no statistically significant signal in the tropical Pacific. As the season progresses, the SSTA pattern strengthens. For JAS, the North Pacific SSTAs remain strong but the center shifts slightly northward. Negative (positive) SSTAs in the North Pacific are associated with positive (negative) P anomalies over w-AZNM. Results are consistent with Higgins and Shi [2000]. The SST composites resemble the first SSTA EOF mode in the North Pacific [Lau and Nath, 1990]. The northward shift of the center (Figures 6a and 6c) is associated with the seasonal pattern change from winter to summer [Zhang et al., 1998]. In addition to negative anomalies in the North Pacific, there are also positive SSTAs along the coast of California (Figure 6c). The influence of these SSTAs along the coast of California on monsoon rainfall was studied by Carleton et al. [1990]. They found negative correlations for wet monsoon years, but no correlation for dry seasons. This suggests that the relationship is not robust.

Figure 6.

Same as Figure 3, but for case 2 events.

[26] For e-AZNM case 2, wet (dry) winters are associated with warm (cold) SSTAs extending from the central tropical Pacific to the west coast of North America (Figure 6d). As season progresses, the SSTA pattern remains the same but anomalies are getting stronger (Figure 6e). In JAS, the composite difference shows positive anomalies extending from the tropical Pacific to the west coast of California and positive SSTAs in the tropical North Atlantic. The correlation between winter and summer SSTAs over the oceans north of 20°S is 0.35. That is statistically significant at the 5% level.

4.2.2. Circulation Features

[27] For both w-AZNM and e-AZNM case 2 events, SSTAs persist from winter to summer. There were only two positive case 2 events for the w-AZNM after 1949. The composites of circulation anomalies will not be meaningful. Therefore they are not shown. As illustrated by Namias et al. [1988], the winter response to SSTAs in the North Pacific is a Pacific North American wave train with negative anomalies over the Gulf of Alaska, positive anomalies over Canada and negative anomalies over the southeastern United States. The circulation pattern is associated with more rainfall over California and Arizona. For JAS, the influence of the north Pacific SSTAs on monsoon rainfall has been established by Higgins and Shi [2000] and Castro et al. [2001].

[28] For e-AZNM, the composite of SSTAs shows anomalies in the Pacific and the Atlantic (Figure 6f). To obtain circulation anomalies associated with this SST pattern (Figure 6f), positive and negative events were identified from the pattern correlation between the JAS mean SST anomalies and Figure 6f over the ocean domain (20°S–60°N, 120°E–360°E) (Figure 7a). Positive (negative) events were selected when the pattern correlation is above (below) 0.4 (−0.4). Reasonable change of criterion will not change conclusions. The composites from these events indicate the influence of SSTAs.

Figure 7.

(a) Pattern correlation between JAS mean SSTA and Figure 6f over ocean points (20°S–60°N, 120°E –0°W), (b) composite 850 hPa height anomaly difference between 6 positive and 7 negative e-AZNM case 2 events for JFM. Contour interval is 5 m. Zero contours are omitted. Areas that positive (negative) anomalies are statistically significant at the 5% level are shaded dark (light), (c) same as (Figure 7b), but composite difference based on pattern correlation (Figure 8a), and (d) same as (Figure 7c but for P (JFM). Contour interval is 0.3 mmd−1, (e)–(g) same as Figures 7b–7d, but for JAS. Contour interval is 3 m for Figures 7e and 7f and 0.3 mmd−1 for (Figure 7g).

[29] For JFM, the composite based on the e-AZNM events (Figure 7b) and the composite based on the pattern correlation (Figure 7c) are similar. Both show negative anomalies over the coast of California and positive anomalies over the northeastern United States and Canada. The dipole pattern in the Atlantic is the response to the Atlantic SSTAs. The rainfall composite based on the pattern correlation shows positive anomalies over the California, the Southwest, the Great Plains and Florida with negative anomalies over the Ohio Valley (Figure 7d).

[30] While composites for JFM are similar, the JAS composites differ significantly over North America (Figures 7e and 7f). The composite difference based on the e-AZNM case 2 events indicates negative anomalies located over the Southwest. That is consistent with heavy rainfall over e-AZNM. The composite difference based on pattern correlation time series (Figure 7f) shows positive anomalies over the central United States and negative anomalies located in the North Pacific along the west coast of North America. Notice that negative anomalies do not extend to the Southwest. The rainfall composite (Figure 7g) shows positive rainfall anomalies extending from the interior west region to New Mexico with negative anomalies over the eastern Untied States. Anomalies over e-AZNM are statistically significant, but they are less organized and magnitudes are weak.

4.2.3. Influence of SM Over w-AZNM

[31] The composite differences of SM, E and P for w-AZNM and e-AZNM case 2 are presented in Figures 8 and 9to examine contributions from SM. For the w-AZNM case 2 events, positive P anomalies extend from California to Arizona (Figure 8a) during winter. Positive P increases SM because E is small in winter (not shown). Spring is not a raining season (Figure 1a), and the composite difference does not show statistically significant P anomalies (Figure 8b). In this case, P anomalies do not persist from winter to summer. There are no significant P anomalies in spring to enhance SM anomalies, but SM anomalies are able to persist from winter to spring (Figures 5d and 8c). SM anomalies decrease in summer because of the increase of E. The composite difference for JAS shows positive SM anomalies in Arizona, but values averaged over the w-AZNM are less than 20 mm (Figures 8d and 5d). Positive SM anomalies lead to positive E anomalies in JAS (Figure 8e) which contribute directly to P anomalies (Figure 8f).

Figure 8.

Composite difference of P anomalies between 7 positive and 11 negative w-AZNM case 2 events for (a) winter (JFM). Contour interval is 0.4 mmd−1. Zero contours are omitted. Contours −0.2 and 0.2 mm d−1 are added. Positive (negative) values that are statistically significant at the 5% levels are shaded dark (light), (b) same as (Figure 8a), but for spring (AMJ), (c) same as (Figure 8b), but for soil moisture difference. Contour interval 20 mm, (d) same as (Figure 8c), but for JAS, (e) same as (Figure 8d) but for E difference. Contour interval 0.3 mm d−1 and (f) same as (Figure 8d) but for summer P (JAS), (g) same as (Figure 8d) but for surface temperature anomalies. Contour interval is 0.5°C, and (h) same as (Figure 8d), but for 850 hPa anomaly difference between 2 positive and 6 negative events. Contour interval is 2 m.

Figure 9.

Composite difference of P anomalies between 11 positive and 16 negative e-AZNM case 2 events for (a) winter (JFM). Contour interval is 0.4 mm d−1. Zero contours are omitted. Contours −0.2 and 0.2 mm d−1 are added. Positive (negative) values that are statistically significant at the 5% levels are shaded dark (light), (b) same as (Figure 9a), but for spring (AMJ), (c) same as (Figure 9b), but for soil moisture difference. Contour interval 20 mm, (d) same as (Figure 9c), but for JAS, (e) same as (Figure 9d) but for E difference. Contour interval 0.3 mm d−1 and (f) same as (Figure 9d) but for summer P (JAS), (g) same as (Figure 9d) but for surface temperature anomalies. Contour interval is 0.5°C, and (h) same as (Figure 9d), but for 850 hPa anomaly difference between 6 positive and 7 negative events. Contour interval is 2 m.

[32] The indirect contributions from SM are small. The differences in radiation anomalies are small, so E anomalies are balanced by sensible heat anomalies. Less sensible heat implies the decrease of surface temperature over the w-AZNM (Figure 8g). There are some negative Tsurf anomalies over w-AZNM, but they only cover a small area and are less than 1°C. The composite of 850 hPa between positive and negative events is given in Figure 8h for reference. There are only two positive events so the plot is suggestive. The indirect influence of SM to P through the changes of circulation anomalies is small.

[33] The scatter diagram (Figure 10b) indicates that there is a linear relationship between spring and summer anomalies with correlation 0.81 (Figure 10a), which is statistically significant at the 5% level. However, there is no relationship between spring SM (AMJ) and summer P (JAS). The correlation of 0.39 is not statistically significant at the 5% level.

Figure 10.

(a) Scatter diagram between SM anomalies averaged for spring (AMJ) and summer SM (JAS) for w-AZNM case 2, (b) same as (Figure 10a), but between SM anomalies in spring (AMJ) and P anomalies for summer (JAS), (c)–(d) same as Figures 10a–10b, but for e-AZNM case 2.

[34] For w-AZNM case 2, the relationship between SSTAs and rainfall is contemporaneous. Negative SSTAs in the North Pacific are in favor of rainfall for both winter and summer. Figure 8 shows that the feedback from SM to P is largely through E, but there is no linear or consistent relationship between spring SM and summer P. The reason may be that soil moisture is just one of many factors modulating rainfall over w-AZNM. The modeling study by Kanamitsu and Mo [2003] indicates that without moisture transport from the Gulf of California, positive SM anomalies in Arizona diminish due to evaporation and are unable to persist. For w-AZNM, synoptic events like moisture surges from the Gulf of California play a role in modulating monsoon rainfall [Higgins et al., 2004; Higgins and Shi, 2006; Johnson et al., 2007]. Surge events are not influenced by soil moisture. On contrary, the moisture transported into the region by the GCLLJ associated with surge can help to maintain soil moisture in the region.

4.2.4. Influence of SM Over e-AZNM

[35] For the e-AZNM case 2 positive events, warm (cold) SSTAs in the tropical Pacific are associated with wet (dry) winter and spring (Figures 9a and 9b). The e-AZNM case 2 differs from the w-AZNM case. The SM is able to modulate P directly through E and indirectly through circulation changes. Wet winter enhances positive SM anomalies and they persist from spring to summer (Figure 5h) and contribute to P directly through E. In comparison to the summer composite for w-AZNM (Figure 8d), the composite for e-AZNM shows stronger and wide spread SM anomalies over entire New Mexico and Texas (Figure 9d). The SM over e-AZNM has a weak seasonal cycle. With large P anomalies, SM anomalies increase in August and September (Figure 5h).

[36] SM anomalies also influence P through circulation changes. E anomalies are balanced by sensible heat anomalies. More E implies less sensible heat and cooler surface temperatures over the e-AZNM (Figure 9g). Over e-AZNM, temperature anomalies are more organized than anomalies in the similar composite for w-AZNM (Figure 8g). The circulation anomalies represented by the 850 hPa height difference show negative anomalies over the Southwest (Figure 8h). The comparison to the 850 hPa composite based on pattern correlation (Figure 7f) suggests the influence of soil moisture.

[37] The scatter diagram (Figures 10c and 10d) shows a linear relationship between spring and summer SM anomalies with correlation 0.9. Figure 10d also shows linear correlation between spring SM and summer P anomalies with correlation 0.73 which is statistically significant at the 5% level. The P and SM relationship is more coherent and strong over e-AZNM. Results suggest that SSTAs and soil moisture work together to enhance summer rainfall over e-AZNM.

5. Conclusions

[38] This paper presents a hypothesis that the soil moisture and precipitation interactions over the Southwest depend on the SSTA forcing. If SSTAs persist from winter to summer then SM has a chance to provide positive feedbacks to P.

[39] For these years that the inverse relationship between winter and summer P is strong, SSTAs do not persist from winter to summer. The JAS SSTAs have strong influence on summer precipitation. For wet winter, SSTAs evolve to the summer mode that is in favor of dry summer. The positive P-SM feedback is not able to overcome the SSTA forcing. After P anomalies turn to negative, SM responds to P and starts to diminish. SM continues to decrease and reaches a minimum in late summer near the end of the monsoon season. For dry winter cases, the situation reverses. For case 1 events, SM does not play a major role in modulating P.

[40] For cases that wet (dry) winter is followed by wet (dry) summer, the associated SSTA forcing persists. Persistent SST forcing allows SM to strengthen or at least to maintain strength. In JAS, positive SM and P feedbacks enhance P anomalies. For e-AZNM, there is a linear relationship between SM anomalies in spring and P anomalies in JAS. That signals that SM is strong enough to modify circulation anomalies to enhance rainfall. For e-AZNM case 2, the SSTA forcing and soil moisture reinforce each other to enhance monsoon rainfall.

[41] For w-AZNM, the SSTAs in the North Pacific persist from winter to summer. Warm (cold) SSTAs are in favor of less (more) rainfall over w-AZNM for both winter and summer. The spring is not a rainy season but soil moisture is able to maintain strength from spring to summer. There are positive feedbacks from soil moisture to P in summer, but there is no linear relationship between SM in spring and P in summer. For w-AZNM, surge events from the Gulf of California can trigger rainfall when the midlatitude conditions are favorable [Higgins et al., 2004; Higgins and Shi, 2006]. The moisture brought by the GCLLJ from the Gulf of California maintains soil moisture in Arizona. The surge events are not controlled by local SM but can be influenced by SSTAs indirectly. For w-AZNM, SM is just one of many factors influencing monsoon rainfall.

Acknowledgments

[42] We wish to thank Andy Wood for providing VIC data. Thanks to Dennis Lettenmaier for helpful discussions. This work was supported by NCPO/CPPA project GC04-072.

Ancillary