Analysis of United States west coast upwelling winds and climatology from the 20 th Century Reanalysis project

Transcription

1 ÉCOLE POLYTECHNIQUE PROMOTION X 27 BORELY Olivier olivier.borely@polytechnique.edu RESEARCH INTERNSHIP REPORT RAPPORT DE STAGE DE RECHERCHE Analysis of United States west coast upwelling winds and climatology from the 2 th Century Reanalysis project RAPPORT NON CONFIDENTIEL Mécannique et Physique pour l Environnement Directeurs de l option : Hervé Le Treut, Jean-François Roussel et Thomas Dubos Directeur de stage : Nathan J. Mantua 12 Avril Juin 21 Climate Impacts Group 3737 Brooklyn Ave. NE PO Box Seattle, WA USA

2 Contents Abstract 1 1 Climatic context of west coastal United States during summer and future evolutions Climatic context of west coastal United States Pressure Gradient Temperature inversion Upwelling Consequences of global warming Bakun s hypothesis Consequences of enhanced upwelling Summer mean analysis Datasets Severaldatasets Accuracy of the 2 th Century Reanalysis Dataset Summer trends Wind Land Temperature Sea temperature Pressure Thermal Inversion Discussion of Bakun s theory Interaction between wind and Temperature In the west coastal United States Other upwelling regions Analysis of daily summer variability Sea-land contrast Interaction between wind and temperature Discussion 31 Bibliography 32 List of Figures 35 2

3 Abstract An increase in alongshore wind off Californian coasts during summer months has been observed and documented during the 2 th century, and has been connected to global warming. This report focuses on the linear trends in several climatic quantities, including temperature, sea level pressure, meridional wind and inversion strength, using the 2 th Century Reanalysis Dataset. It appears that over land, coastal locations have warmed faster than inland locations, leading to a decrease in temperature gradient between coast and land. It comes with an increase in sea surface temperature, and a decrease in sea surface pressure and in inversion strength off central and southern California. A model has been proposed: the North Pacific High seems to have gradually retracted during the 2 th century. The subsidence of warm air over the marine boundary layer is thus less strong, and the inversion strength decreases. The marine boundary layer is thicker, and when the inversion is higher than coastal topography, marine air can drip over the coastal ranges, carrying with it moisture and cloudiness. Clouds over land reduce the incoming solar radiation amount, producing cool conditions inland, while warmer and dryer conditions predominate in the coastal area. Several issues can be pointed out, such as the documented increase in upwelling-favorable winds, which does not match this scenario, and the poor accuracy of the 2 th Century Reanalysis compared to observational measures. Une augmentation pendant les mois d été du vent parallèle aux côtes de California centrale et méridionale a été observée et documentée sur l ensemble du XX e siècle. Cette augmentation a été reliée au réchauffement climatique. Ce rapport examine différentes grandeurs climatiques, parmi lesquelles la temperature, la pression au niveau de la mer, le vent méridionnal et l intensité de l inversion, et tente de déterminer leurs tendances au premier ordre, en se basant sur la base de données de la 2 th Century Reanalysis. Il apparaît que sur terre, la zone cotiêre s est réchauffée plus rapidement que l intérieur des terres, conduisant à une réduction du gradient de température entre la côte et l intérieur du continent. Cette décroissance est accompagnée d une élévation de la température à la surface de la mer, et d une diminution de la pression au niveau de la mer, ainsi que d une réduction de l intensité de l inversion thermique au large des côtes de Californie centrale et méridionale. Un modèle simplifié permet d expliquer ces tendances : une rétraction de l anticyclone Nord Pacifique au cours du XX e siècle pourrait etre à l origine de ces évolutions. Elle aurait conduit à une réduction de l intensité de la subsidence d air chaud au-dessus de la couche limite marine, et donc à une inversion thermique moins intense. La couche limite marine gagne en épaisseur et, lorsque la hauteur de l inversion dépasse la topographie côtière, l air marin peut franchir les massifs côtiers, emportant dans son flot nuages et humidité. Le rayonnement solaire incidant est écranté par la présence de nuages, avec pour conséquences, des températures fraîches a l intérieur des terres et des températures plus chaudes et un air plus sec au niveau des côtes. L étude compte plusieurs imperfections. L augmentation constatée de l intensité de l upwelling ne correspond pas au scénario proposé, et la qualité de la base de données de la 2 th Century Reanalysis est très discutable au regard de mesures expérimentales directes.

4 Acknowledgments I would like to thank Dr. Nathan J. Mantua for his direction, assistance, and guidance. I also wish to thank John M. Wallace, Todd P. Mitchell, James A. Johnstone, Guillaume S. Mauger and Karin A. Bumbaco, for their support and for their recommendations and suggestions which have been invaluable for this project. I am also thankful towards the Joint Institute for the Study of the Atmosphere and Ocean and towards the Climate Impacts Group, where I have spent three months working in nice conditions. I would like to acknowledge the many people who have made it possible for me to have this internship, especially my teachers at the Ecole Polytechnique, Hervé Le Treut, Jean-François Roussel and Thomas Dubos, as well as Francis Codron. 2

5 Chapter 1 Climatic context of west coastal United States during summer and future evolutions 1.1 Climatic context of west coastal United States Pressure Gradient Eastern ocean boundary regions are among the most productive areas in the world, characterized by strong wind-driven upwelling of cold, nutrient-rich waters. From late spring to early fall, the west coast of the United States and adjacent Pacific Ocean are under the influence of persistent southward winds, inducing strong upwelling alongshore. The near-surface air flow over the eastern North Pacific is the result of a complex sea-atmosphere-land pattern. The North Pacific is dominated by the North Pacific High, a subtropical anticyclone lying approximately at 38 N, 15 W, whereas relatively low pressure prevails over the southwestern United States deserts (c.f. Figure 1.1.1). It creates a Sea Level Pressure gradient between land and ocean, which drives persistent southward winds over the coastal ocean Temperature inversion The temperature structure over the Northeast Pacific Ocean is characterized by a marked inversion during the warm season, of the order of 1 C in potential temperature. The subsidence associated with the North Pacific High, coupled with a turbulent mixed marine layer, helps to maintain and strengthen this inversion [Neiburger, 196]. The inversion acts as a lid on the upward transport of water vapor. Above the inversion, the air is very dry and warm. It may be reinforced by westward advection of hot dry air from the continent. Below the inversion, in the 3

6 Figure 1.1: Mean summer sea level pressure between 198 and 26 (data from 2 th Century Reanalysis dataset) Figure 1.2: Average summer cross section from San Francisco to Honolulu (from [Neiburger et al., 1961]) turbulent, well-mixed marine layer, the air is cool and quite moist. Baroclinicity from the horizontal temperature gradient between relatively cool ocean surface and warm heated continent causes the inversion to slope downward toward the coast, with enhanced wind speed adjacent to the coast. The inversion has a mean height of the order of 4 m along the central and northern California coast. It rises quickly in the westward direction for the first 2 km, then more slowly to reach an elevation of the order of 2 m around Hawaii. (cf Figure 1.1.2) According to the conceptual and descriptive model of [Beardsley et al., 1987], we can divide the coastal ocean in 3 zones. In the offshore zone, the marine inversion tilts slowly down towards the coast. This eastward tilts reflects the effect of the subsidence of the subtropical high in the eastern North Pacific in summer. The surface southward winds are of the order of 5 to 7.5 m/s. In the intermediate zone, from 2 km to 2 km, the marine boundary layer tilts more steeply toward the east. This sharper sloped inversion base comes with an enhanced baroclinicity, and a faster surface flow, with a maximum off northern California coast, where average winds reach 1 m/s. 4

7 Figure 1.3: Conceptual model of average lower atmosphere over eastern North Pacific during periods of persistent south and southeastward winds in summer (from [Beardsley et al., 1987]) Figure 1.4: Conceptual model of average lower atmosphere over the nearshore zone during (top) night and (bottom) day. (from [Beardsley et al., 1987]) In the nearshore zone, the surface flow is characterized by a strong land-sea diurnal variability. At night, the wind is strong southward at 2 km offshore, and decreases towards shore, where the lower atmosphere has been cooled during the night. During the day, the sun heats the land surface, changing the stable air over the coast to unstable. The air below the inversion penetrates onshore, which results in an acceleration of the winds over the beach, and in a depression of the nearshore inversion. During the day the inversion tilts downward to a minimum elevation (of the order of 5-1 m) just over the beach. Inland, the inversion lifts very slowly and becomes very diffuse. The climatology of this region is further complicated by a wide variety of mesoscale circulations and coastal trapped events, particularly during the summer. Meso-scale climate zones and circulation are determined by the interaction between the cool synoptic-scale flow from the marine boundary layer and the various mountain ranges, valleys and passes. The Pacific mountain system form an eastern obstacle to the expansion of the marine boundary layer. This system are the series of mountain ranges that stretch along the West Coast of North America from Alaska south to Northern and Central Mexico. Occasionally, coastally trapped events occur that cause the inshore inversion to be lifted. If the inversion remains lower than the prevailing topography of 5

8 the Coast Ranges, inland penetration of marine air is thus limited to a lowelevation coastal zone, with an interior boundary defined by the inversion height and the local relief. Clouds within the moist, cool marine boundary layer are thus sequestered in coastal area, often evolving into fog. Areas located at the west of the Coastal Ranges enjoys an exotic wet climate, while eastern area are extremely dry and warm. The inland penetration of the marine boundary layer induces a strong contrast in daily temperature, which differs by nearly 2 C between coastal and inland locations Upwelling Those southward alongshore winds produce strong upwelling in the coastal area. Near-surface water is transported farther offshore through Ekman pumping, and is replaced with cold and salty water full of nutrients from the depths below. Along with dissolved CO 2 and light energy from the sun, those nutrients are used by phytoplankton to produce organic compounds, through the process of photosynthesis. Very high levels of primary production (the amount of carbon fixed by phytoplankton) result from these specific conditions. High primary production propagates up the food chain following the course of [Mann and Lazier, 26]: P hytoplankton Zooplankton P redatoryzooplankton F ilterfeeders Predatoryfish Other evidences of the intense upwelling are the bending of isotherms south- Figure 1.5: A conceptual diagram of the coastal upwelling process [from ward along the coast, producing sea surface temperatures 4 C lower than at comparable latitudes in the open Pacific, and lowering of alongshore sea-level, of about.1-.2 m during upwelling events. 6

9 1.2 Consequences of global warming Bakun s hypothesis The large scale of these pressure systems makes them sensitive to climate variability. The long-term increase in upwelling has been observed and well documented in coastal California since the mid 2 th century [Bakun, 199, Garcia-Reyes and Largier, 21]. It has been suggested that global warming would be responsible for this increase. [Bakun, 199] proposed that global greenhouse warming should lead to intensification of the continental thermal lows adjacent to upwelling regions. This intensification would be reflected in increased onshore-offshore atmospheric pressure gradients, enhanced alongshore winds, and accelerated coastal upwelling circulation. As the earth warms, the land warms faster than the ocean. The land-sea temperature difference is thus expected to increase, with the associated land-sea pressure gradient, leading to stronger alongshore winds and thus enhanced coastal upwelling. Analysis of wind time series along the coast of California have indeed confirmed the increase in alongshore wind speed [Bakun, 199, Garcia-Reyes and Largier, 21, Schwing and Mendelssohn, 1997, Mendelssohn and Schwing, 22] Numerical models have confirmed that increased global temperature would lead to stronger winds along the coast [Snyder et al., 23, Auad et al., 26]. The consequences of such intensification in the upwelling-favorable winds have received considerable attention in the past decade Consequences of enhanced upwelling Enhanced upwelling is expected to cool waters over the shelf, while further offshore, an increase in surface heating can be expected to exceed the cooling effects of increased upwelling. The effects are further complicated by other change in oceanographical conditions, such as the deepening of the level of the thermocline and changes in the thermal stratification [Palacios et al., 24]. The strength of the upwelling affects the amount of primary production available, and the amount delivered to coastal ecosystems rather than offshore ecosystems. Since 1951, the biomass of macrozooplankton in waters off southern California has decreased by 8 percent [Roemmich and McGowan, 1995]. Enhanced upwelling would push surface water away from the coast more rapidly, resulting in less phytoplankton availability in coastal waters and a greater but more diffuse supply of phytoplankton to waters over the outer shelf and slope, while enhanced surface heating reduces phytoplankton availability further offshore[botsford et al., 23, Botsford et al., 26]. Upwelling intensification could result in dramatic consequences on the biological cycles in the coastal zone. Furthermore, it would have a human impact, by rising the sea level, threatening Californian beaches, with an important economic impact [Pendleton et al., 26]. Tourism and fisheries would also suffer from a change in climatic conditions, as spotted by [Cooley and Doney, 29]. Expanding job losses and indirect economic costs will follow harvest decreases as ocean acidification broadly damages marine habitats and alters marine resource availability. Losses will harm many regions already possessing little economic resilience. 7

10 Thus, climate change would have a dramatic economic impact. 8

11 Chapter 2 Summer mean analysis 2.1 Datasets Several datasets The Twentieth Century Reanalysis (V2) data are provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at This is a reanalysis dataset spanning 1871 to present, assimilating only surface observations of synoptic pressure, monthly sea surface temperature and sea ice distribution. It provides the first estimates of global tropospheric and stratospheric variability at six-hourly resolution. It has a two degree longitudelatitude horizontal resolution, which corresponds approximately to a 15km- 22km zonal-meridional resolution in the North Pacific. Daily direct observations measurements are also available at stations all over the United States. Those data are gathered in the United States Historical Climatology Network (USHCN). The USHCN is a data set of daily and monthly records of basic meteorological variables from observing stations across the United States. To conduct further analyses of the climate of the western Unites States area, only the 114 stations located in California and in the west part of Oregon and Washington states have been used. The USHCN stations are located inland. To study climatic features over the ocean, data from 15 buoys of the National Data Buoy Center (NDBC) have been used. Available measures are more scattered than for the USHCN stations. There are a lot of missing days, and it is difficult to find continuous time series of more than 2 years. We selected the buoys for which the available data covered the longest time frame Accuracy of the 2 th Century Reanalysis Dataset To discuss the reliability of the 2 th Century Reanalysis Dataset, correlations coefficient are computed between mean summer temperature for each of the 114 west coast USHCN stations between 198 and 26 and the mean summer temperature at sigma level.995 for the same locations and same time frame In all this section, the term summer refers to the period between June 1 st and September 3 th 9

12 given by the 2 th Century Reanalysis Dataset. The correlation coefficient for each station is plotted on Figure 2.1: Correlation of mean summer (JJAS) temperature between the 2 th Century Reanalysis Data and the USHCN Data, from 198 to 26 It appears clearly that the correlation between the observations at the USHCN stations and the reconstructed temperature of the 2 th Century Reanalysis are not very conclusive. The mean correlation is.4, and the value are widely distributed around the mean (standard deviation =.12). One can notice that the stations located along the south coast of the area are less correlated with the 2 th Century Reanalysis data than those located further inland. It is probably due to the fact that summer climate in this area is modulated by local topography. The resolution of the 2 th Century Reanalysis is too coarse to detect those very local patterns. Especially along the coast, cooler temperature is observed from the beach to the Coast Ranges, within an area of a few dozens of kilometers. The 2 th Century Dataset has a spatial resolution of about 16 km and can t detect this local pattern. The mean summer meridional wind value from the 15 NDBC buoys are compared with the value of the meridional wind at the same locations from the 2 th Century Reanalysis dataset, between 1983 and 26. The correlation are calculated for each of the buoys, and displayed on 2.2 The meridional wind from the 2 th Century Reanalysis over the ocean is better correlated with NDBC data. The mean correlation is.57, but the value are more widely distributed (standard deviation =.24). Data are not very well correlated along the Californian coast, certainly because the buoys are located over the continental shelf over this area, where the wind is affected by coastal topography on a small scale. Further offshore, wind from 2 th Century Reanalysis is rather well correlated with observations from the buoys. 1

13 Figure 2.2: Correlation of mean summer (JJAS) meridional wind between the 2 th Century Reanalysis Data and the NDBC Data, from 198 to Summer trends Wind Literature review [Bakun, 199] published a controversial paper, documenting an intensification of the alongshore wind stress off western United States, from 3 Nto48 N during the spring-summer season. It established a link between global warming and this escalation of alongshore winds. He calculated 4 years of upwelling index derived from atmospheric pressure records at several locations along the west coast of North America from the 194 s to the 199 s. [Schwing and Mendelssohn, 1997] found also an increase in alongshore wind stress between 32 Nand4 N during April-July from 1946 to 199. They used quality-controlled reports of ocean surface conditions, mostly taken by ships-of-opportunity. The key to identify the strengthen wind stress was the ability to separate the long-term nonlinear trend, using the state-space models [Shumway, 1988, Harvey, 1989], which mask the signal of increased upwelling in the observations. Those studies have focused on spatial scales larger than that of the coastal upwelling process, which typically occurs over the continental shelf within 1 km off the shore in intense and localized upwelling zones [Capet et al., 24]. [Bakun, 1973] used a 3-degree mesh length geographical grid, which means 23 km width squares at 45 N, and [Schwing and Mendelssohn, 1997] used spatial regions of approximately 2 latitude by 4 longitude, which means 15 km wide at 45 N. The most recent publication addressing this phenomenon is by [Garcia-Reyes and Largier, 21]. They observed an increase in the upwelling-favorable winds in Central California (35 N-39 N) during the upwelling season (March-July). They use 27 years ( ) of hourly data on winds over the shelf, which are available from buoys deployed by the National Data Buoy Center (NDBC). 11

14 Using the datasets We used the available datasets to compute the trends of the meridional wind offshore in the area..12????????? Figure 2.3: Mean linear trend of mean NDBC summer meridional wind, from 1983 to 29. The? represent the stations for which no linear trend can be observed with a confidence level higher than 95% (P<.5 1 ) NDBC buoys The linear trend in mean summer southward meridional wind at each of the NBDC buoys have been computed for 27 years, from 1983 to 29. The buoys for which the sign of the linear trend is sure with a 95% confidence level (P<.5 1 ) are displayed in Figure 2.3. Consistant with the work of [Garcia-Reyes and Largier, 21], only the buoys located in a nearshore area along central California coasts show a significant increase trend, of about 2 m.s 1 /century. For the other locations, no significant trend can be observed, neither positive nor negative. The increase of wind speed alongshore off central California is consistent with the theory of increase upwelling caused by global warming. However, we do not have time series long enough to make a conclusion, all the more so even within the 27 years of the analysis, there are a lot of missing values. 2 th Century Reanalysis Data The linear trend in the 2 th Century Reanalysis wind are computed for summer between 198 and 26. The 95% confidence level (P<.5 1 ) values are displayed in Figure for the mean summer southward meridional wind at sigma level.995. For the meridional wind, we notice an increase in the speed only for 3 grid points, off central California, consistent with an increase of upwelling-favorable winds in this area. The mean linear slope is.7 m.s 1 /century. Even if the 2 th Century Reanalysis Data are not as accurate as observational data, it corroborates the observations of NBDC from 1983 to 29, in which an increase in temperature was only observable off central California. It seems that this increase is not just an effect of wind fluctuation, but a real trend that we find all over the century. 12

15 x Figure 2.4: Mean linear trend of mean 2 th Century Reanalysis summer southward meridional wind at sigma level.995, from 198 to 26. Only are represented the grid points where the sign of the linear trend is sure at the 95% confidence level (P<.5 1 ) Land Temperature Figure 2.5: Mean linear trend of mean USHCN summer average temperature, from 198 to 26. Only are plotted the stations for which the sign of the linear trend is sure at the 95% confidence level (P<.5 1 ) In Figure 2.5, the slope in long term summer mean temperature for each of the 114 USHCN stations have been computed. Only are displayed the stations for which the sign of the linear trend is sure at the 95% confidence level (P<.5 1 ). The global trend is an increase in mean summer temperature for each of the stations except one, the mean slope being of 1 C/ century. This increase is consistent with the well-known phenomenon of global warming. No specialpatternintheincreaseinmeansummertemperaturecanbereadonthis figure. 13

16 Coast - Inland contrast Following previous studies [Johnstone and Dawson, 21], a principal component analysis was performed with the correlation matrix of the 114 USHCN stations of the area, for the summer mean T max records from 198 to 26. The western boundary of the spatial domain was defined by the U.S. Pacific coastline, and the eastern margin was delimited by the eastern boundary of California and by the 12 W longitude line in Oregon and Washington, which continues the California boundary northward. This area covers approximately the first 3 km of the U.S. Pacific coastal region. All stations within this area were used in the analysis. PC1 PC2 PC3 r r r (a) The correlation between the temperature time series of each station and the time series of the corresponding principal component are represented Tmax PC Tmax PC2 Tmax PC (b) Figure 2.6: The three first principal component of the mean summer air temperature from the 114 western U.S. USHCN stations The maps of the correlation between T MAX and its three first PC are represented in Figure 2.6(a). T MAX PC 1 captures 42% of total interannual variance and reflects coherent variations over nearly the entire region, with all stations but one sharing PC loadings of the same sign. Santa Cruz is the sole exception, with a loading slightly below zero. The PC 1 time series is virtually identical to the unweighted mean of all 114 stations records (r>.99). 14

17 T MAX PC 2 accounts for 12% of variance and captures a north-south contrast. T MAX PC 3 explains 1% and primarily reflects a coast-interior contrast. It appears that T MAX PC 3 has a pronounced increase trend. Further analyses have been performed to discuss the physical relevance of this mode. The regional PC 3 series can be exemplified (r=.72) by the T MAX difference between time series from San Luis Obispo, CA, located in coastal southern California, in an area fully exposed to marine air penetration, and Orland, CA, located in the Northern California Coast Range interior, (cf Figure 2.2.2). Figure 2.7: Map of United States West Coast. The stations of Orland and of San Luis Obispo are represented Thus, this third mode of the Principal Component Analysis of mean summer T MAX accounts for the temperature contrast between coastal locations and inland locations. An analysis of the time series of San Luis Obispo (coastal locations) and of Orland (inland locations) makes it clear that, when the T MAX PC3 increases, the temperature contrast decreases, and inversely, when T MAX PC3 decreases, the temperature contrast increases. As during the 2 th century, T MAX PC3 has increased, it means that the temperature gradient has decreased. Indeed, coastal stations temperature (EOF3>) have increased by 1.36 C/century, whereas inland stations temperature (EOF3<) have only increased by.94 C/century Sea temperature At last, the linear trend in the 2 th Century Reanalysis Data are computed for 198 to 26. The 95% confidence level (P<.5 1 ) values are displayed in Figure for the mean summer Temperature at sigma level.995. An increase in temperature over ocean is apparent in this Figure The average increase is of.7 C/century, with an extreme value of 1.45 C/century off Southern California. There is no discernible increase in temperature over land in this dataset. As we have seen above (2.1.2, page 1), data from the reconstructed 2 th Century Reanalysis data are not very well correlated with 15

18 Figure 2.8: Mean linear trend of mean 2 th Century Reanalysis summer temperature at sigma level.995, from 198 to 26. The grid points where the sign of the linear trend is not sure at the 95% confidence level (P<.5 1 ) are represented in green USHCN data. We can assume that this reconstructed dataset does not capture the real evolution of land temperature. It would account for the lack of noticeable increase in temperature over land. However, as the dataset is built using observations of monthly sea surface temperature (c.f , page 9), the value of sea surface temperature should be relevant. Thus, an increase in temperature is observable all over the ocean, especially over the waters off central and southern California, where it goes with an increase in southward alongshore wind speed. The T MAX PC 3 time series is also correlated (r=.69) with Sea Surface Temperature offshore Southern California, as we can see in Figure The increase in sea surface temperature seems to be connected with the decrease in temperature contrast between coast and land Multiple Decay Rates T at sigma level Tmax PC Year 1 Figure 2.9: Mean sea surface temperature at (32 N,122 W) and the PC3 of T MAX 16

19 2.2.4 Pressure The sea level pressure from the 2 th Century Reanalysis have also been analyzed. We used mean summer pressure at sigma level.995 to calculate the time series of pressure from 198 to 26. The linear slope is shown on Figure for locations where the confidence level is superior to 95%. The mean slope is -228 Pa/century. The data should be accurate over ocean, as the dataset is built with synoptic pressure records. Thus, there is a decrease trend off central and southern Californian coasts. As we can see on Figure 2.2.4, the low pressure system located over south-western United States has extended during the 2 th Century, leading to a retraction of the North Pacific High further offshore Figure 2.1: Mean linear trend of mean 2 th Century Reanalysis summer pressure at sigma level.995, from 198 to 26. The grid points where the sign of the linear trend is not sure at the 95% confidence level (P<.5 1 ) are represented in green. The units of the colorbar are Pa/an The correlation between PC3 of T MAX and Sea level Pressure from the 2 th Century Reanalysis have also been computed for the mean summer value, from 198 to 26. The results are shown in Figure 2.12(a). The Sea level Pressure is negatively correlated with the PC3 of T MAX in the whole area. The correlation is most negative offshore California. The most negative correlation is of -.59, and is reached at (32 N,118 W). Both PC3 of T MAX andmeansummersealevel pressure are plotted on Figure 2.12(b). The sea level pressure at this location has a negative liner trend of -346 Pa/century. Thus, the decrease in sea level pressure seems to be also connected with the reduced temperature contrast between coastal and inland locations. 17

20 (a) (b) (c) Figure 2.11: Mean summer (JJAS) Sea Level Pressure, from 198 to 26 The pressure difference between 2 isobars is 1 Pa 18

21 (a) Correlation between the T MAX PC3 and sea level pressure, for mean summer time series between 198 and 26 Sea Level Pression x Multiple Decay Rates PC3 of SLP Year 1 (b) PC3 of T MAX ans mean summer sea level pressure at (32 N,118 W) 19

22 2.2.5 Thermal Inversion This PC 3 time series displays a good correlation (.68, cf 2.2.5) with the Temperature difference between pressure level 85 mbar and 95 mbar (Δ T) over Southern California (32 N,116 W). This temperature difference is computed using mean summer value of atmospheric temperature from the 2 th Century Reanalysis dataset, and is an index of the strength of the inversion. Multiple Decay Rates 1 Temperature difference between 85 and 95 mb Year Tmax PC3 Figure 2.12: Temperature difference between 85 mbar and 95 mbar at (32 N,116 W), and T MAX PC3 (inverted) Discussion of Bakun s theory From these analyses, we can conclude that contrary to [Bakun, 199] s expectations, the temperature gradient between coast and land has not increased during the 2 th century, but decreased. The temperature at the stations located in the coastal area has increased by 1.36 C/century, whereas for those located inland, the temperature has only increased of.94 C/century. It seems to be connected with a gradual retraction of the North Pacific High, indicated by the decrease of sea level pressure off central and southern California. The strong subsiding air is thus located further offshore, which result in a more diffuse marine boundary layer alongshore. As it is suggested by the decrease in temperature difference between pressure level 85 mbar and 95 mbar, the inversion has weakened. The sea level temperature, as well as the coastal air temperature has risen. It can be explained by the decrease of cloud cover over the coastal area, inducing warmer condition because of more incoming solar radiation, while further inland, those clouds and moist results in cooler conditions. 2

23 2.3 Interaction between wind and Temperature In the west coastal United States The simple scenario proposed above in 2.2.6, page 2, is not consistent with the observed increase of alongshore southward upwelling-favorable winds during the 2 th century. The thickening of the marine boundary layer should result in weaker winds alongshore within the marine boundary layer. The reverse phenomenon is observed, as the alongshore winds increase. [Bakun, 199] had proposed that the increase in temperature contrast should induce stronger winds. To address this hypothesis, we have focused on the interaction between wind and temperature contrast. The correlation between T MAX PC3 and southward alongshore meridional wind have been computed for each of the 2 -length squares of the 2 th Century Reanalysis. The results are displayed in However, no significant correlation can be observed. The maximum correlation is reached offshore southern California, but remains very low (r=.28). The lack of correlation might be connected to the multiple time scales which interfere in wind variability Figure 2.13: Correlation between mean summer southward meridional wind and the PC3 of T MAX In order to capture a better picture of the interactions between wind variability and temperature variability in the area, an Singular Value Decomposition (SVD) of both fields in the area has been performed, with the cross-covariance matrix of the mean summer T AV E value for each of the 114 USHCN stations located along the U.S. West Coast from 198 to 26 on the one hand, and of the mean summer southward meridional wind between meridians 132 Wand 116 W, and parallels 32 Nand48 Non the other hand. We focus on the second mode of this analysis, which explains 22% of total squared variance. Relevant figures are displayed in It appears that the second mode of the SVD analysis for T AV E exhibits a contrast between coastal and inland locations. There is a slightly linear increase trend, but its confidence level is lower than 8%. This second mode is 21

24 rather well correlated (r=.7) with the PC3 of T MAX of the PC analysis. The correlation between the second mode for temperature and for wind is.47 (cf Figure 2.14(c)). There is a light positive correlation (r=.2) between wind and T AV E second mode over central California, but the principal area where the wind is strongly correlated with the T AV E second mode is off Oregon coasts, with a maximum correlation at (44 N,126 W) (r=-.55). The second mode of the meridional wind shows no significant linear trend. The area where the variability in upwelling-favorable alongshore winds is best explained by the variability in the temperature gradient between coast and land is thus located off the Oregon coasts. However, even if the variability mode for the temperature contrast shows a linear slope, it is not combined with an increase for the associated mode in wind. The correlation might only been linked to the local maximum of southward wind off Oregon shore, and not to a physical interaction of both physical data. 22

25 (a) Correlation between Temperature and its Second Mode (b) Correlation between southward meridional wind and temperature second mode USHCN Tave Meridional V Wind (c) Wind and Temperature second modes: The second mode accounts for 22% of total squared variance. Correlation between the two modes: r=.47 Figure 2.14: Results of the Singular Value Decomposition of the mean summer (JJAS) temperature from the 114 USHCN stations and of the southward meridional wind, between 198 and 26 23

26 2.3.2 Other upwelling regions [Bakun, 199] also pointed out that upwelling-favorable wind stress had not increased only off Californian coastline, but also in other the other major subtropical eastern ocean boundary regions, such as the Canary current system off the Iberian Peninsula and northwestern Africa, the Benguela current system off southwestern Africa, and the Peru current system off western South America. The upwelling in those regions is induced by the same type of synoptic pattern than in the coastal ocean off the western United States: a strong atmospheric pressure gradient between warm heated land and an the higher pressure over the ocean that drives vigorous alongshore equatorward wind. As the synoptic patterns are the same in those region than in the eastern United States, an increase in land temperature should also lead to an increase upwelling. Bakun computed the linear trend of alongshore wind stress in those regions, from the 194 s to the late 198 s. We have conducted the same kind of analysis with the 6-hour daily meridional wind values from the 2 th Century Reanalysis, for the same frame of time but also for a longer period, to determine if a trend can be observed during the 2 th century. The wind stress is computed using 6-hour-daily meridional wind data, using the same classical square-law formula than [Bakun, 1973] τ = ρ a C d v v where τ is the stress vector, ρ a is the density of air(1.22kg/m 3 ), C d is an empirical drag coefficient, v is the estimated wind vector near the sea surface with magnitude v. A relatively high value,.26, of the drag coefficient was used to partially offset the effect of using mean data. The linear slope was determined by the method of least square. In each area, an increase trend can be noticed. However, as we have pointed out above, the increase trend off California is not significant. In the other area, the trend is significant at 95% for the period between 1946 and 1988, and significant at 99% for Spain, Morocco and summer in Peru for the period between 198 and 26. Bakun s theory seems to be more relevant in those area, even if further investigations should be conducted, first of all to confirm the accuracy of the wind data from the 2 th Century Reanalysis. California is the upwelling region where the alongshore wind has less increased during the 2 th century. This is thus the region where Bakun s hypothesis is the less likely to be verified. That might be the reason why we concluded that Bakun s theory was not accurate for this region. 24

27 (a) From [Bakun, 199], slope=.26** dyne/cm (b) From 2 th C.R., slope=.21 dyne/cm (c) From 2 th C.R., slope=.8 Figure 2.15: Alongshore April to September wind stress off California, (a) from 1946 to 1988, (39 N,125 W), (b) from 1946 to 1988, (38 N,124 W), (c) from 198 to 26, (38 N,124 W) [significance: * = P <.5; ** = P <.1 1 ], wind stress is in dyne/cm 2 (a) From [Bakun, 199], slope=.33** dyne/cm (b) From 2 th C.R., slope=.56* dyne/cm (c) From 2 th C.R., slope=.34** Figure 2.16: Alongshore April to September wind stress off Spain, (a) from 1946 to 1981, (43 N), (b) from 1946 to 1988, (42 N,1 W), (c) from 198 to 26, (42 N,1 W) [significance: * = P <.5; ** = P <.1 1 ], wind stress is in dyne/cm 2 (a) From [Bakun, 199], slope=.174** dyne/cm (b) From 2 th C.R., slope=.31* dyne/cm (c) From 2 th C.R., slope=.19** Figure 2.17: Alongshore annual wind stress off Morocco, (a) from 1946 to 1981, (28 N), (b) from 1946 to 1988, (28 N,22 W), (c) from 198 to 26, (28 N,22 W) [significance: * = P <.5; ** = P <.1 1 ], wind stress is in dyne/cm 2 25

28 (a) From [Bakun, 199], slope=.35** dyne/cm (b) From 2 th C.R., slope=.59** dyne/cm (c) From 2 th C.R., slope=.41 Figure 2.18: Alongshore October to March wind stress off Peru, (a) from 1953 to 1984, (between 4.5 Sand 14.5 S), (b) from 1946 to 1988, (between 4 Sand 14 S, 82 W), (c) from 199 to 26, (between 4 Sand 14 S, 82 W) [significance: * = P <.5; ** = P <.1 1 ], wind stress is in dyne/cm 2, each mean value for October to March is assigned to the year in which the January to March portion falls (a) From [Bakun, 199], slope=.38** dyne/cm Year (b) From 2 th C.R., slope=.77** dyne/cm Year (c) From 2 th C.R., slope=.24** Figure 2.19: Alongshore April to September wind stress off Peru, (a) from 1953 to 1984, (between 4.5 Sand 14.5 S), (b) from 1946 to 1988, (between 4 Sand 14 S, 82 W), (c) from 198 to 26, (between 4 Sand 14 S, 82 W) [significance: * =P<.5;**=P<.1 1, wind stress is in dyne/cm 2 26

29 Chapter 3 Analysis of daily summer variability 3.1 Sea-land contrast r r r (a) PC1 47.4% of total variance (b) PC2 1.3% of total variance (c) PC3 4.75% of total variance Figure 3.1: First three modes of principal component analysis of daily summer temperature at the USHCN stations, between 198 and 26 Similarly to what has been done in 2.2.2, page 14, a principal component analysis was performed using the correlation matrix of the daily summer maximal temperature of the 114 USHCN stations, from June to September, from 198 to 26. The results are shown in Figure 3.1. TMAX d PC 1 captures 47% of total daily variance and reflects coherent variations over the entire region, with all stations sharing PC loadings of the same 27

30 sign. The PC 1 time series is virtually identical to the unweighted mean of all 114 stations records (r>.99). TMAX d PC 2 accounts for 1% of variance and captures a north-south contrast. TMAX d PC 3 explains 5% and primarily reflects a coast-interior contrast. The Principal Component Analysis of daily summer TMAX d is very similar to the one performed with summer mean values in Figure 3.1 shows the mean correlation between the PC3 of TMAX d and the daily temperature at sigma level.995 from the 2 th Century Reanalysis for each year between 198 and 26. It displays a marked difference between the ocean and the land, with the coastline as a boundary. PC3 of TMAX d is positively correlated with air temperature over the ocean, and negatively correlated with air temperature over the land. It supports the hypothesis that this principal component has a physical meaning. The maximum absolute correlation is reached off Oregon coast (44 N, 128 W) Figure 3.2: Mean correlation of the PC3 of summer daily TMAX d temperature, for each summer between 198 and 26 with daily 3.2 Interaction between wind and temperature ThetimeseriesofPC3ofTMAX d is compared with the time series of southward meridional wind at sigma level.995 from the 2 th Century Reanalysis for each year between 198 and 26. The mean correlations for each gridded point are represented on Figure 3.2. The correlation is rather weak. The maximum correlation is reached at 4 N, 126 W, off Northern California. An area of negative correlation can be noticed off the coasts of Oregon and Northern California, while an area of positive correlation covers the central and southern California. The area of minimum correlation corresponds roughly to the area of maximum southward wind. To go further, for each year, we compare the time series of the PC3 of TMAX d with the time series of the southward meridional wind at 4 N, 126 W. We introduce a lag between the two time series: the time series of the PC3 of TMAX d goes from June 1st to September the 3 th, and the wind time series start 28

31 Figure 3.3: Mean correlation of the PC3 of summer daily TMAX d with daily southward meridional wind, for each summer between 198 and 26 t days later, t going from -15 to 15. We compute the correlation coefficient for each value of t, and for each year. In Figure 3.2, the mean value for each lag is represented, along with the first and last quartile. It appears that the correlation distribution is non-symmetric. The PC3 of TMAX d is more correlated with the wind until a few days before, and less correlated with the wind a few days after. ThetimeseriesofthePC3ofTMAX d is thus linked to the time series of the wind a few day before. However, the correlation is so low (R MAX <.25) that any definitive assumption is difficult. According to this last Figure 3.2, it seems difficult to assume that the temperature gradient between coastal and land locations drives the upwelling-favorable wind alongshore. Rather than a response of the temperature to the wind alongshore, wind evolutions seems to antedate temperature gradient fluctuations. 29

32 Figure 3.4: Correlation between PC3 of TMAX d and meridional wind t days later. For each summer, the correlation between the PC3 of TMAX d time series from June 1 st and September 3þand the time series of southward meridional wind, with a lag of t days is computed. Are plotted the mean value for each lag, and the first and last quartile of the distribution of correlation coefficient. 3

33 Chapter 4 Discussion Several trend have been exhibited in the atmospheric fields over the U.S. West coast area. The sea surface temperature as well as the land temperature have increased during the 2 th Century. The inversion has lost strength off central and southern California, whereas the mean pressure has decreased. Over land, the temperature at the coastal locations has increased faster than at inland locations, leading to a decrease of the temperature gradient between coast and land. We can try to account for these trends with a simple model. The North Pacific High seems to have gradually retracted, leading to a decrease of pressure off Californian coasts. The subsidence of warm air over the marine boundary layer is thus weaker, and the inversion strength decreases. The marine boundary is thicker, and when the inversion is higher than coastal topography, marine air can drip over the coastal ranges, carrying with it moisture and cloudiness. Clouds over land reduces the incoming solar radiation amount, producing cool conditions inland, while warmer and dryer conditions predominates in the coastal area. It accounts for the stronger increase in temperature for coastal locations than for inland locations, and thus for the decrease in temperature gradient. The sea surface temperature increase can be explained both by this phenomenon of increased coastal temperature, as well as by global warming. Even if the upwelling has increased, which is not clear, its contribution in cold water from the depths is not enough to oppose the global increase trend in temperature. Such a model have been recognized by [Fosberg and Schroeder, 1966] on daily time scales, and by [Johnstone and Dawson, 21] to explain the interannual variability. This scenario is not perfect. As the marine boundary layer is thicker and thicker, the alongshore wind should decrease. However, in several studies ([Bakun, 199, Schwing and Mendelssohn, 1997, Garcia-Reyes and Largier, 21]) an increase of the alongshore wind stress has been observed. Increase in alongshore wind stress is part of an opposite scenario, which has been documented on daily [Fosberg and Schroeder, 1966] and interrannual [Johnstone and Dawson, 21] time scales. When the Pacific high is extended and penetrates into Oregon and Washington, the subsidence is strong over the marine boundary layer, which is thus shallow and can t drip over the coastal topography. Cool marine air and clouds tend to be sequestered at the coastline, evolving into fog, while clear skies and compressional adiabatic warming prevail over the interior. The alongshore 31

34 wind is strong, due to the thinner marine boundary layer. Such a behavior can be observed for August, 15 th 1962 [Fosberg and Schroeder, 1966], or for the whole year 1951 [Johnstone and Dawson, 21]. Notes 1 Indicated significance level are taken directly from standard tables of the Student s t distribution,without adjustment for residual autocorrelation. 32