shelf in winter and spring

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 99, NO. C8, PAGES 16,001-16,017, AUGUST 15, 1994 Heat and salt balances ver the nrthern shelf in winter and spring Califrnia E. P. Dever and S. J. Lentz Department f Physical Oceangraphy, Wds Hle Oceangraphic Institutin, Wds Hle, Massachusetts Abstract. Heat and salt balances are estimated ver the nrthern Califrnia shelf frm early December 1988 thrugh late February 1989 (winter) and frm early March thrugh early May 1989 (spring) frm mred meterlgical and ceangraphic time series taken in 93 m f water 6.3 km frm the cast. We find a winter mean ffshre heat flux f 8.7 x 105 W m -, abut a factr f 5 smaller than earlier estimates f the mean summer (upwelling seasn) ffshre heat flux n the nrthern Califrnia shelf. The mean ffshre heat flux is predminantly in the surface bundary layer and is balanced by an alng-shelf heat flux divergence (as represented by an eddy alng-shelf temperature gradient flux) and a cling trend making the mean winter heat balance fundamentally th_ree dimensinal. In cntrast t winter, the spring mean ffshre heat flux f 6.4 x l0 s W m -1 is balanced by a psitive air-sea heat flux f 8.3 x 105 W m - which is abut 80% f the mean air-sea heat flux in summer. This makes the spring mean heat budget primarily tw dimensinal, like the summer mean heat budget ff nrthern Califrnia. On timescales f days the dminant terms in the fluctuating heat budget in bth winter and spring are the crss-shelf heat flux and lcal changes in heat cntent. These are well crrelated with each ther and with the lcal alng-shelf wind stress. The alng-shelf temperature gradient flux, uncrrelated with the alngshelf wind stress, is usually weak n timescales f days. Occurrences when it is strng are interpreted as effects f messcale features. Mean and fluctuating crss-shelf salt fluxes prvide essentially the same infrmatin as crss-shelf heat fluxes. This is nt surprising in light f the strng temperature-salinity relatinship n the nrthern Califrnia shelf. 1. Intrductin The cast f nrthern Califrnia exhibits tw distinct seasns, a summer upwelling seasn and a winter-spring strm seasn [Strub et al., 1987a; Lentz and Chapman, 1989]. These seasns are typically separated by a rapid spring transitin [Lentz, 1987a] and a mre gradual fall transitin [Strub and James, 1988]. Mean meterlgical cnditins during the summer upwelling seasn are distinguished by strng psitive (frm the atmsphere t the cean) sea surface heat flux [Nelsn and Husby, 1983; Lentz, 1987b] and strng persistent equatrward winds [Nelsn, 1977; Strub et al., 1987b]. In respnse t the equatrward wind stress and a negative alng-shelf pressure gradient [Hickey and Pla, 1983], alng-shelf mean currents are equatrward near the surface and exhibit vertical shear, becming weaker and smetimes pleward near the bttm, and mean crss-shelf currents are ffshre near the surface and near zer in the interir [Winant et al., 1987]. Studies f the summer heat budget ver the nrthern Califrnia shelf shw that the equatrward wind stress and resulting circulatin prduce a large mean ffshre flux f heat in the surface bundary layer [Lentz, 1987b; Rudnick and Davis, 1988; Send, 1989] characteristic f castal upwelling. The magnitude f the mean crss-shelf heat flux is similar t that fund during upwelling cnditins in ther Cpyright 1994 by the American Gephysical Unin. Paper number 94JC /94/94JC regins such as Oregn and nrthwest Africa [Bryden et al., 1980; Richman and Badan-Dangn, 1983] and, as in these ther upwelling regins, a psitive mean sea surface heat flux balances the mean ffshre heat flux. Off nrthern Califrnia, alngshre heat transprt and in situ cling play secndary rles in the mean heat balance fr summer, and the rle f nshre eddy heat flux is still unclear. In a vlume heat budget, Lentz [1987b] finds nly a small cntributin due t crss-shelf eddy heat flux. In a study cnsidering a single mring at midshelf, Send [1989] finds a much strnger nshre eddy heat flux which he attributes t wind frcing. Thugh there have been several studies f the heat budget fr the upwelling seasn in castal regins, there have been few studies f the heat budget under nnupwelling cnditins which exist ver the nrthern Califrnia shelf during the winter and early spring mnths. Frm December until the spring transitin in April r May, mnthly mean alngshelf wind stresses n the nrthern Califrnia shelf are typically weak and may be pleward [Nelsn, 1977]. And frm December t February, mnthly mean surface heat fluxes are weak r even negative, becming persistently psitive and strnger in March [Nelsn and Husby, 1983]. Mean alng-shelf currents are pleward and less vertically sheared than in summer, and mean crss-shelf currents are much weaker [Lentz and Chapman, 1989]. All this suggests the heat budget n the nrthern Califrnia shelf during winter and spring may be quite different frm that during the summer upwelling seasn. 16,001

2 16,002 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 39' 38' Gualala Pt. Park 236' 237' &Clverdale Figure 1. Map f SMILE regin. The central mring lcatin was chsen t be near the CODE C3 mring lcatin. This study will primarily use data frm the central C3 mring and the alng-shelf G3 and M3 mrings. The mean and fluctuating heat and salt balances during winter and spring are studied using data cllected at a midshelf mring site ff nrthern Califrnia frm December 1988 t May We intrduce these data in sectin 2. In sectin 3 we develp the heat (and salt) balance equatins per unit alng-shelf distance and the methds used t estimate the terms in these balances frm the data. The heat and salt balances are presented in sectin 4. We cmpare ur results t previus studies f summer upwelling heat balances and discuss their general applicability in light f climatlgical data in sectin 5. Lab 39' 38' U.S. Gelgical Survey (USGS). The STRESS mring had five vectr-averaging current meters (VACMs) with temperature sensrs between 30 and 6 m abve the bttm (Figure 2). Three VACMs were als equipped with SeaBird cnductivity cells (VACM/C). Tgether, the C3 SMILE and STRESS mrings prvided excellent vertical reslutin in temperature frm 5.5 m belw the surface t 6 m abve the bttm. Thugh current bservatins initially had the same reslutin, several VMCMs develped bearing crrsin prblems which caused them t fail befre the end f the 6-mnth deplyment, and ne VACM did nt prvide usable velcity data during its secnd deplyment (Figure 3). In additin t the central SMILE and STRESS C3 mr- ings, sme use will be made f the peripheral SMILE mrings (Figure 1). The G3 and M3 mrings, apprximately 15 km nrth and suth f C3 and with similar temperature reslutin, will be used t characterize alngshelf temperature gradients in the upper 49.5 m. They resembled the C4 mring shwn in Figure 2. VMCM current data exist at 10 m; hwever, upward lking acustic Dppler current prfilers (ADCPs) at G3 and M3 failed shrtly after deplyment. Als, temperature bservatins at C2 (nt present in winter) will be used t check estimates f heat cntent change in spring. Further infrmatin cncerning the mring lcatins, deplyment times, cnfiguratins, and data return frm the SMILE and STRESS experiments are given by Alessi et al. [1991]. Fr reasns related t data cverage (Figure 3) and the lcal air-sea heat flux (Figure 4), the data were divided int C4 C3 C2 2. Observatins 2.1. Field Prgram This study uses bservatins frm the Shelf Mixed Layer Experiment (SMILE) and the Sediment Transprt Events ver the Shelf and Slpe (STRESS) studies. These field prgrams tk place ver the nrthern Califrnia shelf frm mid-nvember 1988 t mid-may 1989 in the same regin as the previus Castal Ocean Dynamics Experiment (CODE) [Beardsley and Lentz, 1987]. The central element f the SMILE array was a surface mring, dented C3, lcated at 38ø38.71'N, 123ø129.56'W in 93 m f water (Figures 1 and 2). C3 was deplyed Nvember 12, 1988, and recvered May 19, This mring supprted vectr-measuring current meters (VMCMs) with temperature sensrs at 12 depths frm 5.5 t 49.5 m (Figure 2) and a surface meterlgical package. Cnductivity was measured with SeaBird SeaCats at six depths (VMCM/C) frm 8.5 m t 49.5 m. A STRESS subsurface mring, within 0.5 km f the SMILE surface mring and als dented C3, was deplyed December 6, 1988, recvered February 27, 1989, and turned arund and redeplyed frm March 3 t May 5, 1989, by B. Butman, 6O loo lo Offshre Distanc (km) Figure 2. Crss sectin f shelf shwing mrings and instrument lcatins. The central C3 site includes a surface mring with instruments spanning the upper 49.5 m and a subsurface mring with instruments spanning the lwer 30 m deplyed as part f STRESS. Alng-shelf mrings at G3 and M3 (Figure 1) had cnfiguratins similar t the C4 mring.

3 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,003 VMCM 5.5 m NOV DEC JAN FEB MAR APR MAY I,,,,.,,,,,,,,,,,,,,;,.,,,,,h,,,,,,,,,,,,...,,,,,,,h,,,,,...,,,,,,,,I,,...,...,,,,,,,,,I,,,,,,...,,,...,,I,,,,,,...,... h,,,,,...,...,,,,, VMCM,/C 8.5 m.. VMCM 11,5 m VMCM/C 14,5 m... VMCM 17,5 rn VMCM/C 20.5 rn,, VMCM 24,0 rn VMCM/C 29.0 m... VMCM 34,0 rn VMCM/C 39.0 rn... VMCM 44,5 rn VMCM/C 49.5 rn... VACM/C 67.0 (65.0) rn VACM 73.0 (71.0) rn VACM/C 79.0 (77.0) rn VACM 85.0 (83.0) m VACM/C 91.0 (89.0) rn Figure 3. Data return fr the SMILE and STRESS C3 mrings. Thugh many VMCM (and ne VACM) velcity sensrs (slid lines) failed early, almst all temperature (lng-dashed lines) and temperature and cnductivity sensrs (shrt-dashed lines) returned cmplete recrds. tw time perids: 0400 UT n December 6, 1988, t 2300 UT n February 20, 1989 (winter), and 2200 UT n March 3 t 0400 UT n May 2, 1989 (spring). The names winter and spring were chsen nt n the basis f large-scale meterlgical patterns, such as the psitins and strengths f the Nrth Pacific subtrpical high and Aleutian lw (C. E. Drman et al., Structure f the lwer atmsphere ver the nrthern Califrnia cast during winter, submitted t Mnthly Weather Review, 1994, hereinafter referred t as C. E. Drman et al., submitted manuscript, 1994) [Halliwell andallen, 1987; Strub et al., 1987b; Lentz, 1987a], but rather the effect f lcal air-sea heat flux n the mean heat budget. The winter perid cvers the meterlgical cnditins which cntrast mst strngly frm summer upwelling cnditins in that mean winds and sea surface heat flux (Figure 4) are bth weak. During spring, mean winds were again weak but surface heating increased, distinguishing this time perid frm winter. Thugh current meter failures began t NOV DEC JAN FEB MAR APR MAY i I,, i,, i i i 7- I -.m 5- - / 5- / i i i i 0, ---, : ', ', ' < -winfer, spring - 1,,... i,,, 1... i... i i... i,,i... i... i... i I... Figure 4. Ttal energy added/subtracted by net ai -sea heat flux f m Nvembe ]988 t May F m Nvembe until the be innin f arch the su ace heat flux is small, variable, and ften f m the cean t the atmsphere. Be innin in March the su ace heat flux becmes persistently psitive. chan e f ]0 W s m - c espnds t a chan e in the depth-averaged temperature C3 (93 m depth) 0. 6øC.

4 16,004 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA affect severely near-surface velcity reslutin during spring, we believe this des nbt change the qualitative results Mean Winter and Spring Cnditins Befre estimating the heat budgets in winter and spring, mean bservatins frm these time perids are presented t give an verview and cmparisn t histrical CODE data taken during the summers f 1981 and 1982 and the intervening winter. Velcity data are presented in a reference frame apprximately parallel t lcal isbaths at C3 and identical t that f the CODE prgram [Winant et al., 1987]. In this crdinate system the alng-shelf directin y is defined such that y = 0 at C3 and increases tward 317øT. The crss-shelf directin x is defined such that x = 0 at the cast and decreases ffshre. Crss-shelf and alng-shelf velcities are dented u and v, respectively. The mean crss-shelf transprt in this crdinate frame was nearly zer during bth winter and spring. Rtatins f 3.7 ø clckwise in winter and less than 0.3 ø cunterclckwise in spring gave zer mean crss-shelf transprt. These rtatins did nt qualitatively affect any f the results discussed in this paper; therefre nly results in the 317 ø reference frame will be presented. In winter during bth SMILE and CODE, mean alngshelf wind stress was weak (0.03 N m -2 in bth years) and equatrward. Thugh upwelling favrable, it was a factr f Figure less than the wind stress averaged ver similar recrd lengths during summer [Winant et al., 1987]. The mean SMILE sea surface heat flux in winter was 2 W m -2 as.() i i i I lo 20 F 30 _c 40 c)_ 50 ) 60 r F 30 c- 40 c)_ 50 ) 60 r I I I ' U cm/s 0 I I I v cm/s 8 () (d) I I I I I. S Psu Mean winter (a) crss-shelf and (b) alng-shelf velcity prfiles at C3. The slid line indicates means between 0400 UT n December 6, 1988, and 2300 UT n February 20, 1989 (SMILE), and the dashed line indicates means between 1300 UT n December 12, 1981, and 1200 UT n March 22, 1982 (CODE). Mean (c) temperature and (d) salinity prfiles are als shwn. N salinity time series were cllected during the time perid. Means are nly shwn fr recrds lasting the entire time perid. Fr SMILE/STRESS these depths are indicated by circles. cmpared t climatlgical mnthly mean values f nearly 200 W m -2 in June, July, and August [Nelsn and Husby, 1983]. Figure 5 shws mean bservfitins frm the winter (SMILE) and frm a similar perid during winter CODE [Lentz and Chapman, 1989]. Mean winter crss-shelf currents were similar fr CODE and SMILE, 1-3 cm s -1 ffshre in the upper 30 m, nshre in the interir with a maximum f abut 2 cm s-1 and ffshre frm 75 m t near bttm during SMILE (Figure 5a). In cntrast during (Figure 6a). Mean alng-shelf currents (Figure 6b) were the summers f 1981 (CODE 1) and 1982 (CODE 2), mean mre Vertically sheared than in winter. Near the surface, ffshre currents at C3 (nt shwn) exceeded 6 cm s- 1 in the flw was equatrward with a speed ver 11 cm s -1. It upper 30 m, and mean crss-shelf currents were weak (-1 became weaker with depth and was pleward belw 70 m. cm s -1) belw 30 m [Winant et al., 1987]. During SMILE, While this sheared equatrward flw is similar to the mean winter mean alng-shelf currents were pleward with a flw in summer, the weak equatrward mean wind stress and maximum speed f ver 6 cm s-1 frm abut 50 t 75 m and high temperature and l0 TM salinity at C3 suggest it was caused weaker flw near the surface and bttm. During CODE, by a messcale feature ver the uter shelf and slpe in winter mean alng-shelf currents were pleward with speeds March and April 1989 (see als Largier et al. [ 1993]), nt by f abut 5 cm s -1 but were nt as vertically sheared. By wind-driven upwelling. Spring mean temperature and salincntrast, summer 1982 [Winant et al., 1987] mean alngity prfiles (Figures 6c and 6d) were bilinear in character. shelf currents were equatrward and vertically sheared with Abve 49.5 m, vertical gradients were 0.04øC m -1 in temspeeds f ver 7 cm s -1 near the surface and weaker flw perature and psu m -1 in salinity. These gradients were abut 3 times thse bserved in winter and abut twice thse belw 40 m at C3. Mean vertical temperature and salinity gradients were 0.01øC m -1 and practical salinity units bserved in upwelling seasn [Winant et al., 1987]. Belw (psu) m -1, respectively, fr winter (see Figures 49.5 m, mean vertical gradients were 0.01øC m -1 and c and 5d). These gradients are cmparable t, but slightly psu m -1 similar t thse bserved in winter. smaller than, the mean temperature and salinity gradients reprted fr summer by Winant et al. [1987] and Huyer [1984]. In spring the equatrward mean alng-shelf wind stress Methds Develping the Heat Budget Equatins was less than 0.02 N m -1 and the mean sea surface heat flux Surface t bttm vertical cverage existed nly at the during this time perid was 133 W m -2, appraching typical central C3 mring. This cnstrained us t lk at the heat summer values. Mean near-surface crss-shelf currents were budget ver a tw-dimensinal crss-shelf area bunded even weaker than in winter and were ffshre nly at 8.5 m vertically by the bttm, z = -H(x), and the surface, z

5 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,005 0, and in the crss-shelf directin by the C3 mring, x = -L, and the cast, x = 0. The heat budget fr this area is pcp dx dz = utix =-œ dz L H Ot. H - dx T--+ v dz L H Oy + f œ Q dx (1) where p% is the heat capacity per unit vlume (assumed cnstant and equal t 4.1 x 106 W s m -3 øc-i), T is the water temperature, and Q is the net air-sea heat flux. Equatin (1) states that changes in heat cntent integrated ver the crss-shelf area are caused by crss-shelf advectin thrugh the ffshre side at x = -L, alng-shelf heat flux divergence due t vlume flux divergence and temperature gradient flux, and the net surface heat flux. In integrating t frm (1), the vertical velcity is taken t be zer at the surface (w = 0 at z = 0) and given by the free slip kinematic bundary cnditin at the bttm (w = -u OH/Ox at z = -H), the crss-shelf heat flux is assumed t be zer at the cast (the vertical integral f u T = 0 at x = 0), and the heat flux thrugh the seaflr is taken t be zer. Equatin (1) is useful because the crss-shelf integral f the crss-shelf heat flux is easily evaluated. Hwever, as described by Lentz [1987b] and thers, the additin f cntinuity t cnservatin f heat in (1) leads t a crss-shelf () 0 i i i E 30.c 40 Q. 50 m 60 n I I I I U cm/s (b) 0 I I lo I I V cm/s (c) i i. I I I 8 10 (d) TøC I I I S psu Figure 6. Mean (a) crss-shelf and (b) alng-shelf velcity prfiles at C3 between 2200 UT n March 3, 1989, and 0400 UT n May 2, Mean (c) temperature and (d) salinity prfiles fr the same time perid are als shwn. Means are nly shwn fr depths (indicated by circles) with recrds lasting the entire time perid. VMCM failures in the upper 49.5 m reduced near-surface velcity reslutin during this time perid. flux f temperature and alng-shelf flux divergence f temperature which cancel when added tgether but depend n the abslute temperature scale when cnsidered separately. In rder t cnsider these terms separately, the velcity cmpnents u and v and temperature T in the crss-shelf and alng-shelf vlume flux divergence terms f (1) are decmpsed int spatial average and perturbatin quantities, fr example, and u - (u) = ti. The angle brackets represent depthaveraged quantities at x = -L fr u and T and the crss-shelf area averaged quantity fr v. Tildes represent departures frm these spatial averages. Hence the quantities (u), (T), and (v) are functins f time nly while ti, T, and 7 depend n bth space and time. Using this decmpsitin, (1) can be rewritten p% dx -- dz = p% alx =-œ dz H H t. H - dx T- dz - dx v -- dz L H Oy L H Oy +/ L Q dx (3) where cntinuity has been used t eliminate the terms H(u )(T) = (T) (2) dx dz (4) L H Oy Separating T, u, v, and Q in (3) int time mean (verbar) and fluctuating (prime) cmpnents and time averaging gives the mean heat budget equatin. Using this decmpsitin, the mean crss-shelf heat flux is pce L tit dz = pce L tit dz + ti' T' dz The crss-shelf eddy heat flux in (5), ti't', is handled differently than in several previus papers [e.g., Bryden et al., 1980; Richman and Badan-Dangn, 1983; Send, 1989]. Depth-averaged eddy heat flux terms are ften nt separated frm the depth dependent flux terms, and the crss-shelf eddy heat flux is ften written as f H u't' dz= f H ti' ' dz + f H (u)'(t)' dz (6) Hwever, by (4) the crss-shelf, depth-averaged, eddy heat flux cancels the alng-shelf, depth-averaged, eddy heat flux divergence in the same manner as the mean crss-shelf depth-averaged flux and alng-shelf depth-averaged flux divergence cancel. Therefre we cnsider nly the depth dependent crss-shelf eddy heat flux in the mean heat budget. The vertically integrated depth dependent fluxes in (3) are nt a functin f the abslute temperature scale. The vertical structures f the depth dependent fluxes d depend L (5)

6 16,006 DEVER AND LENTZ' HEAT BALANCES OFF NORTHERN CALIFORNIA n the temperature scale, but they remain similar and have sensible interpretatins prvided temperature is measured relative t an average such as the depth r crss-shelf area average. The apprach f separating the depth dependent cmpnents f the heat budget is the tw-dimensinal analg t the methd used t cmpute heat fluxes ver a vlume by Lentz [ 1987b] Estimatin f Terms in the Mean and Lw-Passed Heat Budgets Terms in a time mean and a 38-hur lw-passed (using the PL64 filter described by Limeburner [1985]) versin f (3) were estimated using time series f u, T, and v at C3 and T at G3 and M3. At deplyment time, temperature and velcity bservatins existed at cmmn depths at C3; hwever, current meter failures (see Figure 3) necessitated interplatin f velcity cmpnents at abut half the temperature sensr depths fr at least a prtin f the time series. It was als necessary t extraplate abve and belw the shallwest and deepest bservatins t the surface and bttm. Because vertical length scales f variability were much greater than the instrument separatins, interplated values were insensitive t the particular interplatin scheme, and we chse linear interplatin. Extraplatin f current and temperature infrmatin frm the shallwest VMCM (at 8.5 m r 5.5 m) t the surface was mre prblematic. We chse t extraplate abve and belw the shallwest and deepest functining sensr depths assuming a vertically unifrm prfile. In winter, unifrm extraplatin f u between 5.5 m and the surface and between 91 m and the bttm yielded similar results as mre cmplicated extraplatin schemes, despite winter near-surface and near-bttm current data during SMILE and STRESS [Santala, 1991; Grss et al., 1992] which suggest this is an versimplificatin. Fr example, linear extraplatin abve the shallwest and belw the deepest bservatins changed mean winter values f the crss-shelf heat flux by nly 10%. Unifrm extraplatin prbably did well because the surface bundary layer, as indicated by mean and median surface mixed layer (SML) depths f 16 and 14.5 m (estimated fllwing Lentz [1992]), was reslved by several current meters; hence it was nt slely dependent n transprt abve 5.5 m. This is supprted by a cmparisn f the lw-passed SML t surface Ekman transprt, Vy/pf, which shws they are crrelated at the 99.9% level (crrelatin cefficient 0.71) with a regressin cefficient f SML transprt n Ekman transprt f In spring, unifrm extraplatin f current and temperature abve the shallwest bservatins almst certainly resulted in sme underestimate f the crss-shelf heat flux. The mean SML depth in spring was 6 m, and ver half the SML depth estimates were 0 m, suggesting the surface bundary layer was nt well reslved by functining current meters. This is als indicated by a cmparisn f lw-passed SML transprt t the lw-passed surface Ekman transprt which shws that thugh they are still crrelated at the 99.9% level (crrelatin cefficient 0.78), the regressin cefficient f SML transprt n Ekman transprt is 1.45 in spring. Inclusin f a transitin layer belw the SML, as suggested by Lentz [1992], did nt accunt fr this discrepancy. Thugh linear extraplatin abve 8.5 m increased the spring mean ffshre heat flux by almst a factr f 2, nly unifrm extraplatin results are presented because unifrm extraplatin was cnsistent with the winter analysis, and because the absence f current measurements abve 8.5 m during mst f spring made all extraplatin schemes untestable. After interplating and extraplating current bservatins t the C3 temperature depths, cmpnents f the mean and fluctuating heat budgets were calculated. The depth dependent crss-shelf velcity t was calculated by subtracting the depth-averaged crss-shelf velcity (u) (fund by trapezidally integrating u and dividing by the depth, 93 m) frm u at each bservatin depth. Depth dependent temperature time series were estimated in the same way. The lw-passed time series f crss-shelf heat flux was calculated by multiplying the hurly time series f the t7 and T tgether, vertically integrating them, and lw-pass filtering the result. The mean crss-shelf heat flux was estimated frm the unfiltered crss-shelf heat flux time series, and the crssshelf eddy heat flux was calculated by vertically integrating the cvariance f t and T. T estimate the change in heat cntent ver the crss-shelf area, time series f 0 T/Ot were cmputed at each C3 depth using centered differences f hurly bservatins. Because there were n temperature measurements between C3 and the cast in winter, heat cntent was assumed t be unifrm frm C3 t the cast. T estimate heat gain r lss ver the crss-shelf area, the time derivatives f temperature at each depth were multiplied by the estimated crss-shelf distance frm C3 t the isbath equal t each measurement depth and vertically integrated. Time-averaged and lw-pass-filtered changes in heat cntent were derived frm the hurly time series. The assumptin that heat cntent changes were unifrm in the crss-shelf directin was checked fr the upper 49.5 m in spring using the C2 mring. C2 was deplyed just prir t the beginning f the spring time perid and was halfway between the C3 mring and the cast (Figure 2). Because f near-shre warming, inclusin f the C2 mring increased the estimate f mean heat cntent change in spring by abut ne third but did nt alter the fundamental balances f the spring mean and fluctuating heat budgets discussed in sectin 5. Fr this reasn, and because crss-shelf temperature infrmatin was unavailable belw 49.5 m in spring, we chse t present estimates fr crssshelf integrated heat cntent change based n the C3 mr- ing alne fr bth winter and spring. The net surface heat flux at C3 was estimated frm bulk frmulas using hurly meterlgical measurements at C3 and nearby lcatins. It was assumed spatially unifrm and was multiplied by the distance t the cast, 6.3 km, t get the crss-shelf integrated sea surface flux. The hurly time series was then averaged and lw-pass filtered t give mean and fluctuating air-sea heat fluxes. The prcedure fr estimating the sea surface flux and a discussin f uncertainties are given in Appendix A. Lack f infrmatin abut spatial variability f alng-shelf velcity and temperature made estimates f alng-shelf heat flux divergence subject t strng assumptins. The alngshelf heat flux divergence in (3) is given by -pcp dx v dz L H Oy the heat flux divergence caused by alng-shelf advectin f an alng-shelf temperature gradient, and (7)

7 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,007 -pc, dx T dz L H Oy the heat flux divergence caused by the alng-shelf vlume transprt divergence. Fllwing Richman and Badan- Dangn [1983], (7) is called the alng-shelf temperature gradient flux. It was estimated frm temperature bservatins at G3 and M3 and velcity measurements at C3. The alng-shelf temperature gradient in the upper 49.5 m at C3 was estimated by differencing G3 and M3 temperatures and dividing by the alng-shelf distance, 29.2 km. G3 and M3 temperatures were linearly interplated t the same depths as the C3 bservatins where necessary. T estimate the alng-shelf temperature gradient flux, the alng-shelf temperature gradient was multiplied by the alng-shelf velcity at C3. Gaps in alng-shelf velcity bservatins were filled fllwing the same interplatin/extraplatin prcedure used n the crss-shelf velcity. In the absence f ther infrmatin the alng-shelf temperature gradient flux was assumed t be unifrm in x s that the integral f v 0 T/Oy in the crss-shelf directin was estimated by multiplying by the distance t the cast at each depth. This was vertically integrated frm 49.5 m t the surface. Mean, eddy, and fluctuating cmpnents f the alng-shelf temperature gradient flux were estimated as fr the crss-shelf heat flux. Observatins f alng-shelf velcity at C2 in spring and f the crss-shelf structure f temperature, alng-shelf velcity, and alng-shelf heat flux in summer [Winant et al., 1987; Lentz, 1987b] indicate the assumptin f unifrmity in x is an versimplificatin. Hwever, the sign and magnitude f the alng-shelf temperature gradient flux as estimated abve are cnsistent with ther infrmatin in bth the time mean and the fluctuating heat budgets. Unfrtunately, the depth dependent alng-shelf heat flux divergence in (8) cannt be even crudely estimated because Ov/Oy is unknwn. We feel it is unlikely that (8) dminates the winter heat budget given its near clsure. Hwever, (8) may play a significant rle in spring when the heat balance des nt clse as well and circumstantial evidence indicates the presence f messcale features Estimating the Mean and Lw-Passed Salt Budgets Cnductivity time series measured at the central SMILE and STRESS mrings allwed sme terms in the salt budget t be estimated and cmpared t the heat budget. The salt balance used was analgus t (3), and terms in the salt budget were estimated in the same manner as thse in the heat budget. T gauge qualitatively the effects f the reduced vertical reslutin f salinity measurements relative t temperature measurements, the heat budget was recalculated at the same reduced reslutin as the salt budget. Vertically integrated values fr crss-shelf fluxes and heat cntent changed by rughly 10%, but the basic balances did nt. The lack f alng-shelf salinity time series measurements meant alng-shelf salinity fluxes culd nt be estimated. The mean net freshwater surface flux was estimated frm evapratin based n the latent heat flux at C3 and precipitatin frm a castal rain gauge at Stewart's Pint (see Figure 1). Other rainfall measurements taken during SMILE shw this rain gauge is representative f rainfall extending nrthward apprximately 90 km t Cabrill Pint [see Alessi et al., 1991, Figure 32]. (8) 4. Results 4.1. Mean Winter and Spring Heat Budgets Using the methds develped in sectin 3, we summarize the vertically integrated winter and spring heat balances per unit alng-shelf distance in Tables 1 and 2. In bth winter and spring the mean crss-shelf heat flux, thugh abut a factr f 5 smaller than in summer, is a dminant term in the mean heat budget. Tables 1 and 2 shw it is largely the result f the mean advectin f the mean temperature field (Figures 5 and 6). Mean crss-shelf velcities are highest in the surface and bttm bundary layers. This leads t the vertical structures in Figures 7 and 8 which shw a nearsurface ffshre maximum caused by the ffshre flw f the warmest water n the shelf and a near-bttm nshre maximum caused by the ffshre flw f the cldest water n the shelf. Thugh the general character f the vertical structures f the winter and spring crss-shelf heat flux is similar, they differ in detail. In winter the near-surface ffshre heat flux extends t 25 m. In spring, near-surface ffshre heat flux values are larger than in winter but are cnfined t the tp 10 m. This is due t a strnger spring near-surface stratificatin and near-surface ffshre mean flw nly abve 10 m in spring. The highly surfaceintensified structure f the spring mean crss-shelf flux is a further suggestin that it may be underestimated, since the tp velcity sensr (at 8.5 m during mst f this time) is nly barely within the near-surface regin. The near-bttm ffshre heat transprt is cnsistent with Ekman bttm bundary layer dynamics in that bserved interir pleward flw is assciated with ffshre near-bttm flw (the crrelatin between v at 67 m and u at 91 m, -0.45, is significant at the 99% level). The ther advective cntributin t the mean heat bud- gets, the alng-shelf temperature gradient flux, is the secnd largest cmpnent f the winter mean heat budget and the third largest cmpnent f the spring mean heat budget. Tables 1 and 2 shw this term (especially in winter) is largely the result f an eddy alng-shelf temperature gradient flux, nt f a mean advective flux. Thugh the magnitude f the alng-shelf temperature gradient flux shuld be viewed skeptically, since its estimatin is subject t strng assumptins, Table 1. Estimated Terms in Winter Heat and Salt Balances Mean Eddy Cntributin Standard Standard t Mean Errr Deviatin Heat cntent change* Crss-shelf heat flux* Alng-shelf temperature gradient flux? Air-sea heat flux* Heat balance residual? Salt cntent change* Crss-shelf salt flux* Heat budget units are in 105 W m -1, and salt budget units are in 10-3 m 2 psu s -1. Standard errr estimates are calculated frm lw-passed data with the number f independent bservatins fund frm the recrd hurs divided by an integral time scale. *Standard errr assumes integral timescale f 40 hurs.?standard errr assumes integral timescale f 60 hurs.

8 16,008 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA Table 2. Estimated Terms in Spring 1989 Heat and Salt Balances Mean Eddy ] 0 - Cntributin Standard Standard t Mean Errr Deviatin 20 - Crss-shelf heat flux* Alng-shelf r0 - temperature gradient flux? Air-sea heat flux* Heat balance residual? t 0 - Salt cntent change* Crss-shelf salt flux* Heat budget units are in 105 W m -i, and salt budget units are in 10-3 m 2 psu s -1. Standard errr estimates are calculated as in Table 1. *Standard errr assumes integral timescale f 40 hurs.?standard errr assumes integral timescale f 60 hurs. the negative crrelatin cefficients assciated with the eddy alng-shelf temperature gradient flux in winter were greater than (the 95% significance level), suggesting that strnger than average pleward flw was assciated with equatrward temperature gradients and vice versa at C3. During spring this crrelatin cefficient is weaker, and the mean advectin f the mean temperature gradient became abut half as large as the eddy alng-shelf temperature gradient flux. The mean alng-shelf temperature gradient s tel W/m 2 Figure 7. Vertical prfiles f the three largest terms in the mean winter heat budget. The mean crss-shelf heat flux is indicated by the slid line, the lcal heat cntent change by the shrt-dashed line, and the alng-shelf temperature gradient flux by the lng-dashed line. The mean crss-shelf heat flux is almst entirely the result f the mean advectin f the mean temperature field. In cntrast, the mean alng-shelf temperature gradient flux is almst entirely due t eddy alng-shelf temperature gradient flux pc Heat cntent change* p%f T/Ot d / I I W/m Figure 8. Prfiles f terms in the mean spring heat budget. The mean crss-shelf heat flux is indicated by the slid line, the lcal heat cntent change by the shrt-dashed line, and the alng-shelf temperature gradient flux by the lng-dashed line. The largest single cmpnent f the mean spring heat flux is the mean surface heat flux (nt shwn). flux is greatest near the surface and decreases mntnically with depth. This decrease is primarily due t larger nearsurface values f alng-shelf temperature gradient flux rather than t the smaller crss-shelf area represented by deeper bservatins. It is in the air-sea heat flux that winter and spring shw the greatest difference. In winter its cntributin t the mean heat budget is an rder f magnitude smaller than the largest estimated terms (Table 1), but in spring it becmes the single largest term in the mean heat budget (Table 2). This was due t an increase in incming shrtwave radiatin frm an average f 104 W m -2 in winter t 190 W m -2 in spring and a decrease in the latent heat flux frm an average f -39 W m -2 in winter t -18 W m -2 in spring. As a result f these prcesses, the lcal heat cntent decreases in winter and increases in spring. Winter in situ cling crrespnded t a depth-averaged temperature drp at C3 f abut 1.0øC ver a 77-day perid. The winter change in mean heat cntent is largest near the surface and decreases with depth. Abve 60 m this decrease was due primarily t larger near-surface mean values f in situ cling, and belw 60 m it was due t the smaller crss-shelf area represented by near-bttm bservatins. Spring in situ heating wuld crrespnd t a depth-averaged temperature increase at C3 f abut 1. IøC ver a 59-day perid, but this increase ccurred primarily in the upper 20 m with little change in heat cntent belw this depth. Our analysis has shwn that the mean winter and spring balances are qualitatively different. In winter the absence f surface heating leads t a three-dimensinal balance between the negative crss-shelf heat flux and the psitive alng-shelf temperature gradient flux. In spring, surface heating is a majr term in the mean heat balance, which is

9 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,009 primarily between surface heating and a negative crss-shelf heat flux. In bth winter and spring the mean crss-shelf heat flux is due t the mean advectin f the mean temperature field and has a vertical structure cnsistent with a windfrced surface bundary layer and an alng-shelf velcityfrced bttm bundary layer. In cntrast, the alng-shelf temperature gradient flux is due t an eddy gradient flux, rather than t mean alng-shelf advectin Fluctuating Heat Budgets Fluctuatins in the lw-pass-filtered heat budget (Figures 9 and 10) are 1-2 rders f magnitude larger than means (in Tables 1 and 2). In cntrast t the mean heat budgets, the dminant balance in the fluctuating heat budgets at perids f days t weeks was MAR APR lllllllllllllillllllllllllllllllllllllllllllllllllllllllllll.,ia ø" dx dz = t7 dz (9) L H H The alng-shelf temperature gradient flux was nly ccasinally imprtant at these time scales, and air-sea heat flux nly DEC JAN FEB ;,,, ß ii., 0.0 v - -x =,,.E -.: ,..., 0 [-"v -V-" "'XJ- - - V IIllllllllllllllllllllllllll I IIlllllllllllllll Illl Illllll[l MAR APR Figure 10. Lw-passed time series f alng-shelf wind stress and terms in the depth-integrated spring heat budget. The dashed line in the third plt frm the tp is the sum f the crss-shelf heat flux, alng-shelf temperature gradient flux, and net air-sea heat flux. The crrelatin between this sum and the estimated heat cntent change is Other zer-lag crrelatins are given in Table 4. IIIIIIIIIIIIIIIIIIIIIIII DEC JAN FEB Figure 9. Lw-passed time series f alng-shelf wind stress and terms in the depth-integrated winter heat budget. Zer-lag crrelatins f terms are given in Table 3. The dashed line in the third plt frm the tp is the sum f the crss-shelf heat flux, alng-shelf temperature gradient flux, and net air-sea heat flux. The crrelatin between this sum and the estimated heat cntent change is became appreciable at time scales f 1 mnth r lnger in spring (Figures 11 and 12). The crrelatins f terms in (9) with each ther and with the lw-passed alng-shelf wind stress at C3 (Tables 3 and 4) indicate that the likely physical prcess accunting fr the balance represented in (9) was the respnse f the crss-shelf heat transprt and temperature t lcal alng-shelf wind frcing. The lwer spring crrelatin cefficients may be caused by a lack f velcity infrmatin abve 8.5 m, weak wind frcing in April, and/r the presence f a messcale feature ver the nrthern Califrnia shelf during this time [Largier et al., 1993]. On an event basis, examinatin f Figures 9-12 shws episdic remval f heat frm the shelf ccurring n five ccasins (December 7, December 24, January 11, January 24, and April 10) fllwing equatrward wind stresses f 0.2 N m -2 r greater, several shrt-lived increases in heat cntent assciated with pleward winds immediately preceding winter upwelling events, and a lnger increase in heat cntent during pleward wind stresses fr several weeks in March. This last increase in heat appears similar in character t the remval f heat caused by upwelling events in winter. It des nt

10 16,010 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA F DEC JAN FEB MAR APR iiiiiiiiiiiiiiiiiiiiiiiiiiiijlllllllllllllllllllllllllllllll 4- t 2.--,,-, / ' ( _. f ft/ot dz iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii IIIIIIIIIIIIIIIIIIIIlllllll[ lllllllllllll[I Figure 11. Time-integrated change in heat cntent at C3 Figure 12. Time-integrated change in heat cntent at C3 thrugh the winter. The heat cntent change is indicated by thrugh the spring. The heat cntent change is indicated by the shrt-dashed line, the change due t crss-shelf advec- the shrt-dashed line, the change due t crss-shelf advectin by the slid line, that due t alng-shelf advectin by the tin by the slid line, that due t alng-shelf advectin by the lng-dashed line, and that due t surface heating by the lng-dashed line, and that due t surface heating by the alternately lng-dashed, shrt-dashed line. A change f t0 i2 alternately lng- and shrt-dashed line. A change f t012 W W s m -1 crrespnds t a change in the depth-averaged s m -1 crrespnds t a change in the depth-averaged temtemperature f 0.54øC. perature f 0.54øC. shw that upwelling and dwnwelling are symmetric, thugh it des indicate that the shrt-lived upwelling/dwnwelling events prir t the transitin t upwelling seasn may have similar effects n the fluctuating heat budget. T examine the vertical structure f the lw-passed heat balance, cvariance empirical rthgnal functins (EOFs) were frmed frm the lw-passed time series f heat cntent change, alng-shelf temperature gradient flux, and crssshelf heat flux. The vertical structure f the lwest EOF f the fluctuating crss-shelf heat flux (Figures 13 and 14) has near-surface and near-bttm maxima in the same directin, unlike the mean crss-shelf heat flux (Figures 7 and 8) which has ffshre flw f cl water in the bttm bundary layer caused by the mean pleward flw. Figures 13 and 14 suggest a tw-dimensinal cnceptual mdel with ffshre flw f relatively warm water in a wind-driven surface bundary layer and nshre return flw f cler water in the interir and in the bttm bundary layer. Like the depth-integrated time series, the lwest EOF amplitude time series f heat cntent change are crrelated with the crss- shelf heat fluxes abve the 99% level (crrelatin cefficient f 0.56 (0.49) in winter (spring)), and bth are crrelated with the wind stress (0.64 (0.61) fr heat cntent change and 0.63 (0.69) fr crss-shelf heat flux). Alng-shelf temperature gradient flux usually made a secndary cntributin t the fluctuating heat budgets. It was nt crrelated with the alng-shelf wind stress and became evident during several events lasting 2-3 weeks in December, early February, March, and late April (Figures 9 and t0), factrs which suggest it may be due t messcale features. The lwest EOF (Figures 13 and 14) f the alngshelf temperature gradient flux is surface intensified, and its amplitude time series shws that all instances f large alng-shelf temperature gradient flux are surface intensified. In at least tw events (February and April) the psitive cntributin f the alng-shelf temperature gradient flux acts t balance a simultaneus negative crss-shelf heat flux. Hwever, in the spring event f April 20-26, estimates f heat cntent change d nt agree well with the sum f crss-shelf heat flux, alng-shelf temperature gradient flux, Table 3. Lw-Passed Crrelatin Cefficients fr Winter Heat and Salt Balances Alng- Alng-Shelf Shelf Wind Heat Crss-Shelf Temperature Salt Crss-Shelf Stress Cntent Heat Flux Gradient Flux Cntent Salt Flux Alng-shelf wind stress Heat cntent Crss-shelf heat flux Alng-shelf temperature gradient flux Salt cntent Crss-shelf salt flux 1 Zer-lag crrelatin cefficients were very near maximum lagged crrelatins. Fr a 40-hur integral timescale (used fr wind stress, heat (salt) cntent changes, and crss-shelf heat (salt) fluxes), crrelatins f 0.20, 0.30, and 0.39 are significant at the 80%, 95%, and 99% levels, respectively. Fr a 60-hur integral timescale (used fr crrelatins invlving alng-shelf temperature gradient fluxes), crrelatins f 0.24, 0.36, and 0.46 are significant at the 80%, 95%, and 99% levels, respectively.

11 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,011 Table 4. Lw-Passed Crrelatin Cetficients fr Spring 1989 Heat and Salt Balances Alng-Shelf Alng-Shelf Heat Crss-Shelf Temperature Salt Crss-Shelf Wind Stress Cntent Heat Flux Gradient Flux Cntent Salt Flux Alng-shelf wind stress Heat cntent Crss-shelf heat flux Alng-shelf temperature gradient flux Salt cntent Crss-shelf salt flux 1 Integral timescales are as in Table 3. Crrelatins f 0.23, 0.35, and 0.45 are significant at the 80%, 95%, and 99% levels, respectively, fr crrelatins between wind stress, heat (salt) cntent change, and crss-shelf heat (sa10 flux. Crrelatins f 0.28, 0.41, and 0.53 invlving alng-shelf temperature gradient flux are significant at the 80%, 95%, and 99% levels, respectively. and air-sea heat flux, which may indicate that the alng-shelf heat flux divergence cntained in (8) is imprtant during this time. The fluctuating balances in bth winter and spring are predminantly between heat cntent change and crss-shelf heat fluxes. These are well crrelated with the wind, and the fluctuating crss-shelf heat flux has a vertical structure determined by wind-driven surface and bttm bundary layers. The alng-shelf temperature gradient flux is nly secndarily imprtant. It is nt crrelated t the wind, may result frm pleward advectin f an equatrward temperature gradient r vice versa, and is prbably due t messcale features. The air-sea heat flux is nt imprtant t the fluctuating winter heat budget and becmes imprtant n time scales f 1 mnth t the spring heat balance Salt Budget In general, the salt balance was similar t the heat balance. The largest term in the mean salt balances (Tables 1 and 2) was the nshre salt flux. Its vertical structure (shwn nly fr winter in Figure 15) was essentially a mirrr image f the vertical structure f the mean crss-shelf heat flux (Figures 7 and 8), suggesting that it was caused by a mean ffshre advectin f lw-salinity water near the surface. Other estimated terms in the mean salt balance were much smaller than the mean crss-shelf salt flux which, in the absence f ther prcesses, wuld increase salinity abut psu in winter and psu in spring. In winter the bserved mean salinity change was much less, abut psu, and in spring the mean salinity change was actually psu. The net surface freshwater flux culd have ffset the crss-shelf flux by nly 0.04 (0.11) psu in winter (spring). This imbalance suggests that an alng-shelf salt flux divergence clsed the mean salt budgets. It is pssible that this alng-shelf salt divergence was a result f river runff frm the Russian River suth f C3 (Figure 1), but even in the absence f river runff the mean temperature-salinity (T-S) relatinship (Figure 16) suggests that the mean alng-shelf temperature W/m 2 Figure 13. Lwest EOFs f terms in the lw-passed winter heat budget. These accunt fr 79%, 77%, and 86% f the crss-shelf heat flux (slid line), lcal heat cntent (shrtdashed line), and alng-shelf temperature gradient flux (lngdashed line) variance, respectively W/m 2 Figure 14. Lwest EOFs f terms in the lw-passed spring heat budget. These accunt fr 77%, 62%, and 81% f the crss-shelf heat flux (slid line), lcal heat cntent (shrtdashed line), and alng-shelf temperature gradient flux (lngdashed line) variance, respectively.

12 16,012 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 0 I 10! 20 1= 5 40 /s/at m I lo psu.m/s Figure 15. Salt flux prfiles f terms in the mean winter salt budget. The mean crss-shelf salt flux is indicated by the slid line, and the lcal salt cntent change by the dashed line. gradient flux at C3 wuld be assciated with a negative cntributin t the salt balance. T gauge qualitatively whether alng-shelf salinity gradients were assciated with alng-shelf temperature gradients, bservatins frm three SMILE cnductivity-temperaturedepth (CTD) cruises [Limeburner and Beardsley, 1989a, b, c] in Nvember 1988, February thrugh March 1989, and May 1989 were cmpared with simultaneus C3 mring temperatures and salinities. The cruises included repeated surveys and alng-shelf sectins with a ttal f 92 statins alng the 93-m isbath with rughly 15 km reslutin. Alng-shelf temperature and salinity differences were negatively crrelated with a crrelatin cefficient f Assuming each CTD survey r alng-shelf sectin represented an independent bservatin (time evlutin was rapid enugh that individual surveys lasting 1 r 2 days culd nt be cnsidered synptic), this crrelatin cefficient is significant at the 95% level. The negative crrelatin cefficient indicates that a pleward transprt f lw-salinity water wuld be assciated with a pleward transprt f heat and vice versa. Like the mean salt balance, the lw-passed fluctuating salt balance prvided essentially the same infrmatin as the lw-passed heat balance. Crss-shelf heat and salt fluxes are negatively crrelated (Tables 3 and 4), as are changes in heat and salt cntent. Changes in salinity and the crss-shelf salt flux are well crrelated t each ther and t the alng-shelf wind stress. The lwest EOF mdes f vertical structure fr salt cntent and crss-shelf salt flux are very similar t, and negatively crrelated with, thse f heat cntent (crrelatin cefficients and in winter and spring, respectively) and crss-shelf heat flux (crrelatin cefficients and in winter and spring, respectively). The perids when crss-shelf salt flux and changes in salt cntent agree least crrespnd t the perids when the alng-shelf temperature gradient flux was imprtant, a further indicatin that the alng-shelf salt flux divergence is related t the alng-shelf temperature gradient flux. 5. Discussin and Summary Mean and fluctuating heat and salt budgets per unit alng-shelf distance have been estimated using measurements frm the SMILE/STRESS field prgrams fr the nrthern Califrnia shelf. On the basis f data cverage and the absence r presence f persistent surface heating, these time series have been analyzed as tw distinct perids: a winter perid characterized by weak surface heating and a spring perid during which surface heating is imprtant. Bth winter and spring were subject t weak mean winds, making meterlgical cnditins different frm summer upwelling seasn, which began immediately after the end f the spring time perid in The gals f this discussin are t cntrast the winter and spring heat budgets with thse bserved in the summer upwelling seasn n the nrthern Califrnia shelf and elsewhere and t put ur results in the cntext f climatlgical frcing cnditins. One f several pints which stand ut in cmparing the mean winter, spring, and summer upwelling heat balances is the relative magnitudes f terms in the balances. T cmpare the winter and spring heat budgets per unit alng-shelf distance t the vlume heat budget f Lentz [ 1987b], the heat cntent change and alng-shelf heat flux divergence values reprted by Lentz were divided by his study vlume f 79.1 km 3 and multiplied by the crss-shelf area between CODE C3 and the cast, 0.39 km 2. Similarly, the crss-shelf heat flux values were divided by the alng-shelf distance, 56 km; the net air-sea heat flux was divided by the surface area f 936 km 2 and multiplied by the distance between CODE C3 and the cast, 5.8 km. The mean crss-shelf heat flux and net air-sea heat flux are directly dependent n the persistence f upwelling favrable winds and seasnal variatin in slar Slinify psu Figure 16. Mean T-S relatin at C3 in winter (slid curve) and spring (dashed curve). Surface mring bservatin depths were at 8.5, 14.5, 20.5, 29.0, 39.0, and 49.5 m. Subsurface mring bservatin depths were at 67, 79, and 91 m in winter and 65, 77, and 89 m in spring.

13 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,013 inslatin. Therefre they shw the greatest variatins in magnitude between winter, spring, and summer. The mean crss-shelf heat fluxes per unit alng-shelf distance in winter and spring (Tables 1 and 2) are abut a factr f 5 smaller than the estimated summer value n the nrthern Califrnia shelf f x 105 W m -. The mean net air-sea heat flux als shws large variability; it increases frm winter and spring values (Tables 1 and 2) t 10.0 x 105 W m - in summer. By cntrast, the summer mean alng-shelf temperature gradient flux (a prtin f the ttal heat flux divergence) and heat cntent change (-2.7 x 105 W m - and -2.4 x 105 W m - respectively) have magnitudes similar t thse in winter and spring (Tables 1 and 2). These changes in relative magnitude lead t differences in the character f winter, spring, and summer mean heat balances. The winter heat balance is between the negative mean crss-shelf heat flux and psitive alng-shelf temperature gradient flux, making it three dimensinal t the lwest rder. The mean spring and summer heat balances are between the negative crss-shelf heat flux and the psitive net air-sea heat flux, making them mre tw dimensinal in character. Hwever, the crss-shelf heat flux is much weaker in spring than in summer, s that a mean increase in heat cntent ccurs in spring, rather than the slight decrease bserved in summer upwelling seasn. Regardless f the seasn, the mean crss-shelf heat flux is f imprtance t the mean heat balance. The mst prminent characteristic f its vertical structure in winter and spring, a near-surface ffshre heat flux, is similar t that bserved in upwelling systems ff Oregn [Bryden et al., 1980], nrthwest Africa [Richman and Badan-Dangn, 1983], and nrthern Califrnia [Lentz, 1987b]. Belw the surface the winter and spring mean crss-shelf heat fluxes decrease in magnitude and have a sign which varies with depth. Near the bttm the winter and spring crss-shelf heat fluxes are strnger and nshre because f the ffshre flw f the cldest water n the shelf. Richman and Badan-Dangn [1983] als bserved a near-bttm increase in crss-shelf heat transprt magnitude n the nrthwest African shelf, thugh it was ffshre because f nshre transprt f cld water. The directins f near-bttm crss-shelf heat trans- prt fund by Richman and Badan-Dangn [1983] and this study can be explained by a near-bttm Ekman transprt driven by a mean interir alng-shelf flw. Bryden et al. [1980] and Lentz [1987b] find n near-bttm increase in crss-shelf heat transprt, pssibly because the bttm bundary layer, expected t be thin during active upwelling events [Weatherly and Martin, 1978; Trwbridge and Lentz, 1991], was nt reslved with available measurements. The mean winter and spring crss-shelf heat flux vertical structures are cnsistent with the ntin that mean winds, thugh weak, drive an ffshre flw f relatively warm water in the surface bundary layer and pleward alng-shelf currents set up an ffshre flw f the cldest water in the bttm bundary layer. Thugh mean heat budgets bserved in winter, spring, and summer are quite different, fluctuating heat budgets are similar. The magnitude f the fluctuatins in winter and spring are within a factr f 2 f fluctuatins in the summer upwelling budget [Lentz, 1987b]. In all three cases the dminant balance is between the crss-shelf heat flux and lcal changes in heat cntent. These terms are highly crrelated with the lcal alng-shelf wind stress and each ther (this is true t a lesser degree in spring). The vertical structure f the fluctuating crss-shelf heat flux, as represented by the lwest EOF mde, suggests wind-driven surface and bttm bundary layers. Thus it appears that shrt-timescale variability is dminated by a wind-driven crss-shelf heat flux with a resulting change in heat cntent in winter, spring, and summer [Lentz, 1987b; Lentz and Chapman, 1989]. It is interesting t nte that pleward winds assciated with dwnwelling can result in a shreward flux f heat and an increase in heat cntent. In this respect at least, dwnwelling events appear t be similar t upwelling events in winter and spring. This tw-dimensinal picture f the fluctuating heat budget is upset in February, March, and late April when the alng-shelf temperature gradient flux becmes imprtant n time scales f weeks. During these times the alng-shelf temperature gradient flux tends t be ffset by the crssshelf heat flux, reducing the net change in heat cntent. Lentz [1987b] fund a similar event in the 1982 summer upwelling heat budget which he attributed t the presence f an ffshre messcale feature seen in satellite infrared images. The events in February, March, and April 1989 d nt appear t be lcally wind driven and may als be due t messcale features. Largier et al. [1993] als bserved messcale features ver the nrthern Califrnia shelf and slpe in March and April 1989 during the Nrthern Califrnia Castal Circulatin Study (NCCCS), a field prgram designed t study large-scale circulatin. The mean heat balances in winter and spring 1989 differ t varying degrees frm summer upwelling balances which are thught t be relatively rbust. Hwever, even the largest terms in the mean balances (Tables 1 and 2), with the exceptin f the spring air-sea heat flux, are scarcely larger than their standard errrs. One questin which then arises is hw general are the mean balances? T help answer this questin, we cnstructed a heat balance frm mre limited data taken in the winter f (Appendix B). This analysis yielded mean (and fluctuating) balances very similar t thse in the winter f ; hence it prvided sme cnfidence in the generality f results. Hwever, the alng-shelf wind stress, the net air-sea heat flux, and the alng-shelf velcity, all imprtant factrs in determining the winter and spring heat balances, are likely t vary significantly ver the curse f a single seasn as well as between seasns and gegraphically. Because the dminant terms in the mean winter and spring heat balances have small magnitudes cmpared t their fluctuatins, variatin in these frcing factrs may alter the magnitude r even the sign f individual terms in the mean winter and spring heat balances. The mst imprtant factr in determining the crss-shelf heat flux, the alng-shelf wind stress, is less unifrm in time and space than in summer when winds are generally upwelling favrable. Nelsn [1977] and Strub et al. [1987a] shw seasnal alng-shelf winds t be equatrward all alng the Washingtn, Oregn, and Califrnia casts during summer mnths but spatially variable in winter mnths, becming pleward nrth f 39øN, remaining equatrward suth f this latitude, and becming weak and variable ff nrthern Califrnia. The lcatins f bth the SMILE and the CODE field prgrams are near this latitude where the sign and magnitude f a mean winter r spring alng-shelf wind stress may be expected t vary. Figure 17, the 30-day lw-pass-

14 16,014 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA O. lo DEC JAN FEB MAR APR MAY Figure 17. Alng-shelf (317øT) wind stress cmpnent fr the mnths f December thrugh May. The mnthly climatlgical winds frm Nelsn [1977] are indicated by the squares. As an indicatin f the variability in bservatins used by Nelsn, apprximate standard deviatins were fund by taking the square rt f the sum f the squares f standard deviatins fr east and nrth wind stress cmpnents (calculated frm standard errr and bservatin numbers presented by Nelsn [1977]). The day lw-pass-filtered winds are indicated by the slid line, and the day lw-pass-filtered winds by the dashed line. filtered alng-shelf (in the 317øT reference frame) cmpnent f wind stress bserved frm December thrugh May during CODE and SMILE at Natinal Data Buy Center (NDBC) 13 and the mnthly mean climatlgical wind stress calculated by Nelsn [1977] at a 1 ø square centered at 38øN, 123øW, shws year t year variability in winter and spring wind fields near the SMILE and CODE lcatins. Frm December thrugh February, climatlgical mnthly mean winds are near zer but equatrward. Winds during this time in are equatrward (within apprximately 1 standard deviatin) f this climatlgical mean, and winds during are als equatrward within a standard deviatin f the climatlgical mean. C. E. Drman et al. (submitted manuscript, 1994) fund that the number f equatrward, pleward, and weak wind events during winter f was fairly typical based n an examinatin f 10 years f wind recrds at NDBC 13. Mnthly mean winds during spring, prir t the spring transitin t upwelling seasn, are prbably als generally equatrward and weak as in Figure 17. All this indicates that a weak mean upwelling may generally be present near 39øN in winter and spring, suggesting a resulting mean ffshre heat flux. The net air-sea heat flux was nt a dminant factr in the winter mean heat balance but became s in the spring heat balance. The separatin int winter and spring in is dependent n the mean winter air-sea heat flux being small and n a relatively rapid change t mean spring surface heating prir t the transitin t upwelling. Figure 18 shws the 30-day lw-passed air-sea heat flux at C3 in with the climatlgical heat flux at a 1 ø square centered at 38øN, 123øW frm Nelsn and Husby [1983]. This shws mnthly mean air-sea heat fluxes are indeed small relative t summer values, and may be negative during winter mnths, and that they increase rapidly in March and April. The spring transitin in ccurred in early May 1989, which allwed fr 2 mnths f surface heating prir t the spring transitin t upwelling. Hwever, the spring transitin is generally thught t ccur earlier, in March r April [Strub and James, 1988]. This wuld cut shrt the spring perid f surface heating prir t strng upwelling, s may be anmalus in this respect. Figure 17 shws the wind stress pltted with that in 1982 when the spring transitin ccurred in mid-april [Lentz, 1987a]. Anther factr which may influence the heat balance is the alng-shelf velcity. Thrugh bttm bundary layer dynamics, this may affect the near-bttm crss-shelf heat flux as well as the alng-shelf heat flux divergence. Year t year variability is evident in a cmparisn f SMILE mean alng-shelf currents with thse during a similar time f year in at the same lcatin (see Figure 4 and Lentz and Chapman [ 1989]). Hwever, mean interir alngshelf currents are pleward in bth years. If pleward interir currents are a regular part f the seasnal cycle, as $trub et al. [1987a] find, they wuld set up a mean ffshre Ekman flw in the bttm bundary layer. The near-bttm nshre heat flux assciated with this Ekman flw culd be a persistent feature in winter and spring. Alng-shelf velcity is als imprtant in determining the alng-shelf temperature gradient flux. It is imprtant t nte that the alng-shelf temperature gradient flux bserved in winter and spring and winter was nt the result f a mean advectin f a mean temperature gradient, but rather was an eddy flux. The physical mechanisms behind the alng-shelf temperature gradient flux deserve further study. In winter and spring it is nt significantly crrelated t the lcal alng-shelf wind, but appears t be assciated with messcale features. Winter re- sults at CODE C3 and R3 suggest that the cntributin f this term t the heat balance may vary ver alng-shelf separatin scales f 30 km. In summer, Lentz [ 1987b] fund an alng-shelf heat flux divergence f apprximately the same magnitude n the nrthern Califrnia shelf; hwever, the mean summer alng-shelf heat flux divergence was primarily the result f the mean alng-shelf advectin f a mean temperature field, rather than an eddy heat flux divergence. Fluctuatins in the summer alng-shelf heat flux divergence were attributed t wind-related ccurrences such as relaxatin frm upwelling [Send et al., 1987] as well as an ffshre messcale feature. Because the alng-shelf temperature gradient flux seems t be assciated with messcale features and because nly a few alng-shelf temperature O. DEC JAN FEB MAR APR MAY Figure 18. Net surface heat flux fr the mnths f December thrugh May. The mnthly climatlgical air-sea heat flux frm Nelsn and Husby [1983] is indicated by the squares with standard deviatins calculated frm standard errrs and bservatin numbers presented by Nelsn and Husby [1983]. The day lw-pass-filtered surface heat flux is indicated by the slid line.

15 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA 16,015 gradient fluxes were bserved during this study, their effects n a climatlgical winter heat budget are uncertain. Appendix A: Estimatin f the Net Sea Surface Heat Flux The net surface heat flux was estimated at C3 as Q = Qi q- Qb q- QI q- Qs (10) where Q i is the net slar radiatin, Qb is the net lngwave radiatin, Q I is the latent heat flux, and Q s is the sensible heat flux. These fur terms were estimated using frmulas very similar t thse emplyed by Lentz [1987b]. The explanatin belw is taken frm that paper with mdificatins as necessary: Qi - I(1 - Ab) where I is the measured inslatin and Ab is the cean albed given by Payne [1972]. The inslatin was measured at C3 using tw types f Eppley pyranmeters, a mdel 8-48 and a mdel PSP. Because the 8-48 returned a nearly cmplete recrd (the PSP failed in early February), we used this instrument fr ur estimate f net surface heat flux. Hwever, a cmparisn f the 8-48 data with ther available inslatin data at C3 and n the cast shwed the 8-48 suffered a gain prblem, cnsistently reading lw by abut 20%. T accunt fr this, we regressed the 8-48 n available C3 PSP data and used this regressin cefficient t btain a mre accurate estimate f inslatin. Q b = Q cs - R, where the Efimva frmula f Simpsn and Paulsn [1979] was used t cmpute the clear sky upward lngwave radiatin, Qcs, frm the water temperature at 5.5 m, and R is the dwnward measured lngwave radiatin reflected frm the atmsphere. Qs = pacp - ChUw(Ta - Ts), where Pa is the density f air, C, is the heat capacity f air, Ca is the sensible heat flux cefficient given by Friehe and Schmitt [1976], u w is the wind speed, T a is the air temperature 52 km suth f C3 at NDBC 13, and Ts is the water temperature at 5.5 m. Ql = LCeuw(q - qs), where L is the heat f evapratin, C e is the latent heat flux cefficient given by Friehe and Schmitt [1976], q is the abslute humidity given by multiplying the measured relative humidity by the saturatin humidity at Ta, and qs is the abslute humidity at the cean surface taken t be 0.98qsat at the temperature T s, where qsat is the saturatin humidity. The necessity f using the air temperature at NDBC 13 and the water temperature at 5.5 m causes sme uncertainty in lngwave, latent, and sensible heat flux estimates. The air temperature at NDBC 13 was used because air temperature recrds at C3 were lst because f a leak in the vectr- averaging wind recrder (VAWR). Cmparisns f NDBC 13 t NDBC 14 (lcated 76 km nrth f C3) air temperature bservatins suggest the temperature difference is at mst 3øC ver 127 km, the distance between NDBC 13 and 14, s that C3 air temperatures are apprximately thse at NDBC 13. The water temperature at 5.5 m was used because water temperature recrds at 1 m belw the surface were lst wing t leakage in the VAWR at C3. The net surface heat flux at C3 was assumed t be unifrm frm C3 t the cast. Frm shre measurements and the C2 buy (present in spring) sme cmpnents f the net sea surface heat flux can be checked fr crss-shelf variatin. All meterlgical variables measured at C3 were als measured n the cast at the Stewart's Pint Beach lcatin shwn in Figure 1. Stewart's Pint and C3 data presented by Alessi et al. [1991] shw that averages f relative humidity and incming lngwave radiatin were within 5% and their standard deviatins were within 15%. Wind speed averages and standard deviatin d decrease by abut a factr f 2 between C3 and the cast. Spring measurements f wind speed at C2 indicate that mst f this decrease ccurs between C2 and the cast. Neglecting the crss-shelf variatin in wind speed prbably resulted in an verestimate f the crss-shelf integrated latent heat flux and sensible heat flux. Hwever, sensible and latent heat fluxes were nt the largest cmpnents f the mean winter and spring net air-sea heat fluxes. These were the mean incming shrtwave radiatin and utging lngwave radiatin. Appendix B: The Winter Heat Budget in Fr cmparisn with the results the heat balance frm 1300 UT n December 12, 1981, t 1200 UT n March 22, 1982, is examined using tw mrings deplyed n the nrthern Califrnia shelf as part f the Castal Ocean Dynamics Experiment (CODE). One mring, dented CODE C3, was 6 km sutheast f the SMILE C3 lcatin. The secnd mring was 34 km sutheast f the CODE C3 lcatin and was dented R3. Bth mrings were in 90 m f water and had temperature and (VACM) velcity sensrs at 9, 35, 55, and 75 m. The C3 mring had an additinal VACM sensr at 15 m. Wind stress data were als acquired at NDBC 13 (lcated 43 km frm C3 and 16 km frm R3). The winter CODE data are presented and analyzed by Lentz and Chapman [1989]. Thugh the vertical reslutin f the winter CODE data is prer than that f the SMILE data and the air-sea heat flux cannt be estimated, the winter CODE data appear t be sufficient t reslve the simple vertical structures fund in the winter SMILE heat balance (Figure 7). Additinally, histrical data [Nelsn and Husby, 1983] and the winter SMILE results suggest that the surface heat flux is negligible during winter. The heat budgets at CODE C3 and R3 were estimated using essentially the same methds presented in sectin 3 except fr the alng-shelf temperature gradient flux. Because nly tw mrings were present, the alng-shelf temperature gradient at bth C3 and R3 was represented by the temperature difference divided by the alng-shelf distance between C3 and R3; hence differences in the alngshelf temperature gradient flux between C3 and R3 were entirely due t spatial variatins in v. The CODE reference frames [Winant et al., 1987] f 317øT and 329øT at C3 and R3, respectively, were used. Rtatin t the reference frame required t make the mean crss-shelf transprt at C3 zer (325øT) affected mean heat flux results by nly abut 10%. The actual lcal isbath rientatin at R3 is nt clear, and rtatin t ther plausible reference frames shwed that mean crss-shelf heat flux values at R3 varied frm 50% t 200% f the values in the 329øT crdinate frame, thugh the fluctuating balance was less sensitive. Bth mean and fluctuating heat balances at C3 fr are qualitatively similar t thse in bth in the magnitude f the vertically integrated budget (Table B 1) and in vertical structure. The mean ffshre heat flux in the upper 30 m is apprximately balanced by mean cling and an eddy alng-shelf temperature gradient flux. The lwpassed fluctuating heat balance in is again an

16 16,016 DEVER AND LENTZ: HEAT BALANCES OFF NORTHERN CALIFORNIA Table BI. Estimated Terms in Winter Heat Balances at C3 and R3 Eddy Cntributin Standard Mean t Mean Errr Standard Deviatin Heat cntent change at C3' Crss-shelf heat flux at C3' Alng-shelf temperature gradient flux at C3t Heat balance residual at C3t Heat cntent change at R3* Crss-shelf heat flux at R3* Alng-shelf temperature gradient flux at R3? Heat balance residual at R3? -4.1 ß ß ß ß ß ß ß ß ß ' ' Heat budget units are in 105 W m -1. Standard errr estimates are calculated as in Table 1. *Standard errr assumes integral timescale f 40 hurs. 'Standard errr assumes integral timescale f 60 hurs. Table B2. Lw-Passed Crrelatin Cefficients fr Winter Balances at C3 and R3 Alng-Shelf Alng-Shelf Heat Crss-Shelf Temperature Heat Wind Stress Cntent Heat Flux Gradient Flux Cntent C3 R3 Alng-Shelf Crss-Shelf Temperature Heat Flux Gradient Flux Alng-shelf wind stress C3 heat cntent C3 crss-shelf heat flux C3 alng-shelf temperature gradient flux R3 heat cntent R3 crss-shelf heat flux R3 alng-shelf temperature gradient flux Integral timescales are as in Table 3. Crrelatins f 0.17, 0.25, and 0.33 are significant at the 80%, 95%, and 99% levels, respectively, fr wind stress, heat cntent change, and crss-shelf heat flux. Crrelatins f 0.20, 0.30, and 0.39 invlving alng-shelf temperature gradient flux are significant at the 80%, 95%, and 99% levels, respectively. apprximate balance, (9), between changes in heat cntent and crss-shelf heat flux which are well crrelated (Table B2) with the wind stress. The alng-shelf temperature gradient flux is uncrrelated with the wind and becmes imprtant nly during several events. Thugh the heat balance at C3 is similar t that in , spatial differences between C3 and R3 did exist. The ffshre heat flux at C3 is larger than that at R3, and the mean alng-shelf temperature gradient flux, which adds at C3, remves it at R3. These qualitative results are nt sensitive t changes in the R3 crdinate frame. Acknwledgments. Financial supprt was prvided by NSF grant OCE We thank all wh cntributed t the SMILE field prgram, especially Russ Davis and Llyd Regier at Scripps Institutin f Oceangraphy and Rick Trask, Gerge Tupper, Paul Buchard, Je Pirier, Jerry Dean, Dave Simneau, and Bb Beardsley at WHOI. Special thanks are als extended t Brad Butman at USGS Wds Hle fr making available his subsurface mring data. Susan Tarbell's and Carl Alessi's effrts in prcessing the SMILE time series are appreciated, as are the effrts f Dick Limeburner in acquiring and prcessing the SMILE CTD data. We als thank Bb Beardsley, Ken Brink, and tw annymus reviewers fr their critiques f this manuscript and Clive Drman, Amelit Enriquez, and Carl Friehe fr helpful discussins. WHOI cntributin References Alessi, C. A., S. J. Lentz, and R. C. Beardsley, The Shelf Mixed Layer Experiment (SMILE): Prgram verview and mred and castal array data reprt, WHOI Tech. Rep , 211 pp., Wds Hle Oceangr. Inst., Wds Hle, Mass., Beardsley, R. C., and S. J. Lentz, The Castal Ocean Dynamics Experiment cllectin: An intrductin, J. Gephys. Res., 92, , Bryden, H. L., D. Halpern, and R. D. Pillsbury, Imprtance f eddy heat flux in a heat budget fr Oregn castal waters, J. Gephys. Res., 85, , Friehe, C. A., and K. F. Schmitt, Parameterizatin f air-sea heat fluxes f sensible heat and misture by bulk aerdynamical frmulas, J. Phys. Oceangr., 6, , Grss, T. F., A. E. Isley, and C. R. Sherwd, Estimatin f stress and bed rughness during strms n the nrthern Califrnia shelf, Cnt. Shelf Res., 12, , Halliwell, G. R., and J. S. Allen, The large-scale wind field alng the west cast f Nrth America, , J. Gephys. Res., 92, , Hickey, B. M., and N. E. Pla, The seasnal alngshre pressure gradient n the west cast f the United States, J. Gephys. Res., 88, , Huyer, A., Hydrgraphic bservatins alng the CODE central line ff nrthern Califrnia, 1981, J. Phys. Oceangr., 14, , Largier, J. L., B. A. Magnell, and C. D. Winant, Subtidal circulatin