The Effects of Orbital Precession on Tropical Precipitation

Transcription

1 University of Miami Scholarly Repository Open Access Theses Electronic Theses and Dissertations The Effects of Orbital Precession on Tropical Precipitation Kimberly Ann Chamales University of Miami, Follow this and additional works at: Recommended Citation Chamales, Kimberly Ann, "The Effects of Orbital Precession on Tropical Precipitation" (2014). Open Access Theses This Open access is brought to you for free and open access by the Electronic Theses and Dissertations at Scholarly Repository. It has been accepted for inclusion in Open Access Theses by an authorized administrator of Scholarly Repository. For more information, please contact

2 UNIVERSITY OF MIAMI THE EFFECTS OF ORBITAL PRECESSION ON TROPICAL PRECIPITATION By Kimberly A. Chamales A THESIS Submitted to the Faculty of the University of Miami in partial fulfillment of the requirements for the degree of Master of Science Coral Gables, Florida May 2014

3 2014 Kimberly A. Chamales All Rights Reserved

4 UNIVERSITY OF MIAMI A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science THE EFFECTS OF ORBITAL PRECESSION ON TROPICAL PRECIPITATION Kimberly A. Chamales Approved: Amy C. Clement, Ph.D. Professor of Meteorology and Physical Oceanography Brian J. Soden, Ph.D. Professor of Meteorology and Physical Oceanography Larry C. Peterson, Ph.D. Professor of Marine Geology and Geophysics M. Brian Blake, Ph.D. Dean of the Graduate School

5 CHAMALES, KIMBERLY A. (M.S., Meteorology and Physical Oceanography) The Effects of Orbital Precession on (May 2014) Tropical Precipitation. Abstract of a thesis at the University of Miami. Thesis supervised by Professor Amy C. Clement. No. of pages in text. (51) The response of tropical precipitation to precessional forcing is investigated using idealized precession experiments from two climate models as well as mid-holocene experiments from ten climate models participating in the Paleoclimate Modeling Intercomparison Project Phase III. The objectives of this thesis are to analyze an energetic mechanism that describes precessional changes in precipitation and determine if this mechanism is sufficient for explaining changes in precipitation over land and ocean. Both mid-holocene experiments and idealized experiments show a seasonal land-sea asymmetry in tropical precipitation that cannot be explained simply by monsoon dynamics. Instead, this shift in precipitation is interpreted using an energetic framework that describes how changes in the net top-of-atmosphere radiation affect the atmosphere and surface energy balances. Over land, changes in precipitation can be explained through changes in moist static energy flux divergence, but over the ocean, the moist static energy flux divergence plays a lesser role because the surface energy balance compensates for the majority of the change in top-of-atmosphere radiation. Because this mechanism does not sufficiently describe why precipitation changes over the ocean, we break down the dry and moist energy responses of the atmosphere in order to further explain this process. We find that the atmosphere s response to precessional forcing over

6 land drives changes in circulation, which influence the thermodynamic structure of the atmosphere over the ocean, ultimately allowing for precipitation changes over the ocean. To conclude, we examine the regional changes in precipitation in the tropical Atlantic and over Africa, and we discuss how paleoclimate records from this region can be used for model verification.

7 TABLE OF CONTENTS Page LIST OF FIGURES... LIST OF TABLES... v viii CHAPTER 1 INTRODUCTION Motivation Precessional Changes in the Mid-Holocene Mechanisms for Precessional Changes in Precipitation Regional Precession Studies Scientific Questions and Research Objectives PRECIPITATION CHANGES DURING THE MID-HOLOCENE Models and Experiments Mid-Holocene Precipitation Signal EFFECTS OF PRECESSION IN IDEALIZED SIMULATIONS Data and Methods Models and Experiments Top-of-Atmosphere Energetic Framework Linking the Energy and Moisture Budgets Results Changes in Insolation Net TOA Radiation, Precipitation, and Surface Temperature MSE Flux Divergence Energy Budget Balance Dry and Moist Energy Summary THE PRECESSIONAL SIGNAL IN THE TROPICAL ATLANTIC Explaining Regional Changes in Surface Temperature Regional Precipitation Changes in the Mid-Holocene Experiments Implications for Paleoclimate Records Summary CONCLUSION Discussion Future Work iii

8 APPENDIX A RESPONSE OF THE HYDROLOCIAL CYCLE A.1 Idealized Experiments A.2 Mid-Holocene Experiments.. 42 REFERENCES iv

9 List of Figures Figure 2.1: Multi-model mean seasonal precipitation change (mm/day) between experiments (MH-PI) for the 10 PMIP3 models (listed in Table 2). Stippling indicates regions where 7 out of 10 models agreed on the sign of the precipitation change Figure 2.2: Multi-model mean seasonal cycle of the precipitation difference (mm/day) as a function of time (months) between experiments (MH-PI) for the 10 PMIP3 models (listed in Table 2) over land (solid) and ocean (dashed). Shading indicates the model spread. Curves are scaled by the percentage of tropical land and ocean Figure 3.1: Latitudinal distribution of the zonal-mean insolation difference (W/m 2 ) as a function of time (months) between experiments (SS-WS) for each model.. 17 Figure 3.2: Seasonal net TOA radiation difference (W/m 2 ) between experiments (SS-WS) for each model Figure 3.3: Seasonal precipitation difference (mm/day) between experiments (SS-WS) for each model. Contours indicate the mean seasonal WS precipitation Figure 3.4: Seasonal cycle of the precipitation difference (mm/day) as a function of time (months) between experiments (SS-WS) for each model over land (solid) and ocean (dashed). Curves are scaled by the percentage of land and ocean in the tropics.. 20 Figure 3.5: Seasonal surface temperature difference (K) between experiments (SS-WS) for each model Figure 3.6: Seasonal MSE flux divergence (W/m 2 ) between experiments (SS-WS) for each model. Contour lines are changes in precipitation minus evaporation (mm/day) 21 Figure. 3.7: Seasonal cycle of the difference (SS-WS) in tropical-mean (30 N to 30 S) precipitation (blue), net TOA radiation (orange), net surface energy fluxes (purple), and MSE flux divergence (red) in W/m 2 as a function of time (months) over land and ocean for each model. Curves are scaled by the percentage of land and ocean in the tropics 23 Figure. 3.8: Vertical profile of the difference (SS-WS) in tropical-mean (30 N to 30 S) dry (red) and moist (blue) energy (W/m 2 ) as a function of pressure (mb) over land (solid) and ocean (dashed) for each model. Curves are scaled by the percentage of land and ocean in the tropics Figure. 3.9: Illustrates the land and ocean response to TOA forcing in DJF (top panel) and JJA (bottom panel). Red and blue arrows denote net energy gain and loss, respectively, at the TOA and surface. Black arrows denote circulation change. Red lines indicate the change in dθ/dp over the ocean and the cloud represents precipitation 27 v

10 Figure 4.1: Seasonal precipitation difference (mm/day) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM Figure 4.2: Seasonal surface temperature difference (K) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM Figure 4.3: Seasonal cloud cover difference (%) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM Figure 4.4: Seasonal net TOA radiation difference (W/m 2 ) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM. Boxes denote the regions in the eastern and western Sahel for Fig Figure 4.5: The vertical JJA difference in cloud cover (percent) between experiments (SS-WS) averaged over the western Sahel, region 1 (red), and the eastern Sahel, region 2 (blue), for GFDL-CM2.1 and CAM3-SOM. The boxes in Figure 4.4 denote regions over which the averages were computed. The vertical axis is pressure (mb) Figure 4.6: Multi-model mean seasonal precipitation change (mm/day) between experiments (MH-PI) for the 10 PMIP3 models (listed in Table 2). Stippling indicates regions where 7 out of 10 models agreed on the sign of the precipitation change Figure A.1: Scatterplot of the percent change in tropical-mean (30 N to 30 S), monthlymean precipitation vs. the tropical-mean, monthly-mean change in surface temperature for both models. The changes are computed as differences between the SS and WS experiments for the entire tropical domain (black), land only (green), ocean only (blue). The solid lines represent the linear fit of ΔP to ΔT; regression coefficients are listed in Table A Figure A.2: Scatterplot of the percent change in global-mean, annual-mean column integrated water vapor vs. the change in global-mean, annual-mean surface temperature for the 10 PMIP3 models and GFDL-CM2.1 and CAM3-SOM from Chapter 3. Differences are computed as MH PI for each PMIP3 model and as SS WS for GFDL- CM2.1 and CAM3-SOM. Colors indicate the total change (black), the change over land only (green), and the change over ocean only (blue). The solid lines indicate the linear fit of Δq to ΔT; regression coefficients are listed in Table A.2. The dotted red line indicates the rate of increase of column-integrated water vapor (7.5% K -1 ) found in Held and Soden (2006) Figure A.3: Same as Figure A.2 but for the percent change in global-mean, annual-mean precipitation vs. the change in global-mean, annual-mean surface temperature.. 44 vi

11 Figure A.4: Scatterplot of the percent change in tropical-mean (30 N to 30 S), monthlymean precipitation vs. the tropical-mean, monthly-mean change in surface temperature for the 10 PMIP3 models and the multi-model mean. The changes are computed as differences between the MH and PI experiments for the entire tropical domain (black), land only (green), and ocean only (blue). The solid lines represent the linear fit of ΔP to ΔT; regression coefficients are listed in Table A.3 45 vii

12 List of Tables Table 2. The PMIP3 models used in this analysis, the model years averaged over for the mid-holocene and pre-industrial climatologies, and the resolutions of the atmosphere and ocean components of each model... 9 Table A.1: The first column contains values for the change in tropical-mean (30 N to 30 S) surface temperature (WS-SS) for each model. The last three columns contain the regression coefficients of the lines in Figure A Table A.2: Regression coefficients for Figures A.2 and A.3 for the total tropical domain, land only, and ocean only Table A.3: The first column contains values for the change in tropical-mean (30 N to 30 S) surface temperature (MH-PI) for each model. The last three columns contain the regression coefficients of the lines in Figure A viii

13 Chapter 1: Introduction 1.1. Motivation Variations in Earth s orbit are known to impact climate over glacial-interglacial timescales, and an understanding of these changes is a key component in understanding the climate of the past and future. One of these variations is orbital precession, which modulates seasonal insolation on cycles of about 20,000 years. Precessional signals have been observed in a variety of precipitation records including speleothems from caves in Asia (Wang et al. 2008) and South America (Cruz et al. 2005; Wang et al. 2006, 2007) and sediment cores from Asia (Gasse et al. 1991; Herzschuh 2006), South America (Bush et al. 2002), and Africa (Street and Grove 1979; demenocal et al. 2000; Gasse 2000; Trauth et al. 2003). The majority of these studies use oxygen isotopes from speleothems or various proxies for lake level fluctuations in order to provide information about regional changes in precipitation. Understanding the mechanisms that drive these changes in precipitation can help explain how precipitation changed in past climates on glacialinterglacial timescales and how it might change in the future Precessional Changes in the Mid-Holocene The mid-holocene is a warm period that occurred 6,000 years ago (6 ka) and is characterized by stronger insolation during the boreal summer than there is at present day. This is a useful period for studying precession because although there are variations from present day in other forcings, such as greenhouse gases and obliquity, precessional change is the main driver of the mid-holocene climate response. A variety of modeling studies show that a land-ocean shift in precipitation exists between the mid-holocene and 1

14 2 preindustrial climates (Braconnot et al. 2007; Braconnot et al. 2008; Hsu et al. 2010; Bosmans et al. 2012), and paleorecords provide evidence that mid-holocene enhanced insolation in the Northern Hemisphere coincides with intensified monsoonal precipitation (Winkler and Wang 1993; Yu and Harrison 1996; Jolly et al. 1998; Yu et al. 1998; Kohfeld and Harrison 2000; Baker et al. 2001; Haug et al. 2001; Marchant et al. 2009). As part of this thesis, we study the changes in tropical precipitation over land and ocean in the Paleoclimate Modeling Intercomparison Project Phase III (PMIP3) mid- Holocene experiments. Examining the previously documented land-ocean asymmetry in a suite of climate models with a standardized experimental design can help determine the robustness of this precipitation signal and motivate the need for a complete mechanism describing precessional changes in precipitation Mechanisms for Precessional Changes in Precipitation Similar to the mid-holocene forcing, previous modeling studies have documented a land-sea shift in precipitation in the tropics as a result of idealized precessional forcing (Clement et al. 2004; Tuenter et al. 2003). These changes in precipitation were originally explained as land and sea breezes due to differences in the heating of land and ocean (Kutzbach and Otto-Bliesner 1982; COHMAP Members 1988; Tuenter et al. 2003; Ruddiman 2008; Bosmans et al. 2012). Bosmans et al. (2012) summarize this mechanism and show that increased summer insolation strengthens the thermal low over the warming continents, which increases the land-sea pressure gradient and enhances monsoon winds. A recent study by Merlis et al. (2012) instead suggests that changes in precipitation cannot simply be explained through differences in the heating of land and ocean, because

15 3 feedbacks within the climate system can lead to surface temperature gradients that may not be consistent with the sea breeze mechanism. They used idealized aquaplanet simulations with a zonally symmetric land surface to describe precessional changes in precipitation in terms of changes in the net top-of-atmosphere radiation. This research aims to extend the framework of Merlis et al. (2012) to climate models with zonal asymmetries, that is, with realistic distributions of land and more complete sets of physical processes. A study by Hsu et al. (2010) discusses the land-ocean asymmetry of tropical precipitation changes in the mid-holocene using an energetic approach similar to the one used by Merlis et al. (2012). They used a balance of moist static energy and radiative fluxes at the top-of-atmosphere and at the surface to explain changes in precipitation over both the land and ocean. These authors found that the different heat capacity of the land and ocean and the associated slow response time of the ocean allows the insolation forcing of one season to be cancelled out by the insolation forcing of the previous season causing changes in precipitation over the ocean to lag the changes in precipitation over the land. This study highlights another approach that uses the atmospheric and surface energy balance to describe the land-ocean contrast in precessional precipitation. Another method for describing a land-sea contrast in precipitation is to consider the constraints placed on tropical precipitation as a result of the precessional forcing. Held and Soden (2006) show that precipitation increases at a rate of about 2% K -1 in climate models as a result of greenhouse gas forcing. Using this approach to determine the rate of change of precipitation as a result of precessional forcing can help constrain changes in precipitation over land and ocean. However, there are some limitations to

16 4 applying this framework in the context of this study. First, Held and Soden (2006) focus on global- and annual-mean changes, but the precipitation response to precessional forcing has a seasonal signal that is primarily constrained to monsoon regions. Also, Held and Soden (2006) examine the equilibrium response of greenhouse gas forcing; the surface temperature is no longer warming as a result of the forcing. This is not the case in this study, and surface warming and cooling of the ocean plays an important role in the mechanism controlling precipitation changes (discussed in chapter 3). For these reasons, an analysis of the global- and annual-mean changes in precipitation and temperature in this study may not be representative of the type of climate response discussed in Held and Soden (2006), however it may provide insight on the response of the hydrological cycle to external forcings in different climate states Regional Precession Studies Understanding the mechanisms that drive changes in precipitation due to precessional forcing can not only help explain precipitation changes in past and future climate, but it can also help explain how the precipitation signal manifests in regional climate records. Moreover, it is of interest to examine regional precipitation changes because over glacial-interglacial timescales, the regional response to solar forcing may differ significantly from the global response. Many modeling studies have focused on precessional changes in monsoon precipitation on the regional scale (Kutzbach 1981; Kutzbach and Otto-Bliesner 1982; Brostrom et al. 1998; Joussaume et al. 1999; Braconnot et al. 2000; Liu et al. 2003; Tuenter et al. 2003; Clement et al. 2004; Zhao et al. 2005; Ohgaito and Abe-Ouchi 2007; Braconnot et al. 2008). Clement et al. (2004)

17 5 show that regionally, the precessional signal of precipitation can be as large as, or even overwhelm the glacial signal. This has important implications for the interpretation of climate records and for regional climate sensitivity. In this thesis, we will focus on changes in tropical Atlantic precipitation and the African monsoon. Precipitation anomalies in these regions have societal and economical impacts in South America and Africa, particularly the Sahel. Understanding the precessional mechanism in this region will also help us to understand the link between precipitation and the atmospheric general circulation, which is especially important for the transport and deposition of Saharan dust. Transport and deposition of dust from the Sahara has been shown to have impacts on the atmospheric radiation budget, biogeochemical cycles in the ocean and on land, and air quality (Swap et al. 1992; Jickells et al. 2005; Muhs et al. 2007, 2009, 2012; Maher et al. 2010). This region is also useful because there are a variety of proxy-records available for model validation from this region (Street and Grove 1979; Tiedemann et al. 1994; Arz et al. 1998; Qin et al. 1998; Lea et al. 2000; demenocal et al. 2000; Gasse 2000; Bush 2002; Trauth et al. 2003; Weldeab et al. 2007) Scientific Questions and Research Objectives Although there is evidence from both paleoclimate records and modeling studies that monsoonal precipitation varies with precessional forcing, many questions remain regarding the spatial distribution of these precipitation changes, the mechanisms forcing these changes, and their regional signals. The purpose of this study is to test a top-ofatmosphere energetic mechanism that describes changes in precipitation as a result of

18 6 precessional forcing using both mid-holocene experiments and idealized model experiments. My project is motivated by the following questions: (1) How does orbital precession affect the distribution of tropical precipitation, and is this a robust signal among many climate models? (2) Can this precessional signal in precipitation over land and ocean be described using an energetic framework? (3) Can constraints on the atmospheric energy budget and precipitation changes over land predict how precipitation changes over the ocean? We address these questions by examining the precipitation signal over land and ocean in the mid-holocene experiments from ten PMIP3 models. We then use idealized precession experiments to employ an energetic framework that describes how changes in the net top-of-atmosphere radiation affect the atmosphere and surface energy balances. We will examine the regional precipitation signal due to precession in the tropical Atlantic and Africa, and an assessment of the hydrological cycle in both the idealized experiments and the mid-holocene experiments will help to further constrain changes in precipitation over land and ocean. The results of this thesis will help elucidate how precipitation changes in response to top-of atmosphere forcing in the tropics, the differences in the mechanisms controlling precipitation changes over land and ocean, what the regional precession signal looks like in the tropical Atlantic, and how paleorecords can be used for model validation. This study is organized as follows: We first examine the seasonal precipitation change over land and ocean in the PMIP3 mid-holocene experiments. We then describe the idealized experiments and discuss the top-of-atmosphere mechanism used to constrain

19 7 precipitation changes. Finally, we examine regional changes due to precession and how these signals might be observable in paleoclimate records. In the last section we provide a summary of this work and possible future work. Appendix A includes a supplemental analysis and discussion of the strength of the hydrological cycle in the precession experiments.

20 Chapter 2: Precipitation Changes During the Mid-Holocene It is of interest to first determine if the changes in orbital precession produce a precipitation signal over land and ocean that is present in many of today s climate models. In this chapter, we examine the precipitation changes due to precessional forcing in the PMIP3 mid-holocene experiments to determine if a common mechanism that is present in these models may be driving the precipitation signal Models and Experiments PMIP3 is a project that aims to coordinate paleoclimate modeling and model evaluation using standardized experiments from various models and research institutions across the globe. For this project, we use climatological monthly data from the PMIP3 mid-holocene and pre-industrial control experiments. Table 2 lists the 10 models used in this study and the model years used to compute each climatology. The mid-holocene was a period that occurred about 6 ka in which precessional forcing drove changes in the seasonal insolation reaching Earth. We use mid-holocene data (MH) from 10 climate models in which the autumnal equinox is set close to perihelion (ω 180=0.87, where ω is the longitude of perihelion), eccentricity is set to , and obliquity is set to These experiments will be compared to the preindustrial control runs (PI) in which the Northern Hemisphere (NH) winter solstice occurs near perihelion (ω 180=102.04), eccentricity is set to , and obliquity is set to The mid-holocene greenhouse gas concentrations are set to 280 ppmv for CO 2, 650 ppbv for CH 4, and 270 ppbv for N 2 O, and the pre-industrial concentrations are set to ppmv for CO 2, ppbv for CH 4, and ppbv for N 2 O (the 8

21 9 concentrations from 1850). Although there are variations in the greenhouse gas concentrations prescribed for the mid-holocene and pre-industrial experiments, the changes that result from this forcing are small compared to the changes that result from the orbital forcing (Bosmans et al. 2012). Model Years for Climatology Resolution Model Name MH PI Atmosphere Ocean CCSM x192 x L26 320x384 x L60 CNRM-CM x128 x L31 362x292 x L42 CSIRO-Mk3L x56 x L18 128x386 x L40 FGOALS-g x60 x L26 360x180 x L30 FGOALS-s x108 x L26 360x180 x L30 GISS-E2-R x90 x L40 288x180 x L32 IPSL-CM5A-LR x95 x L39 182x149 x L31 MIROC-ESM x64 x L80 256x192 x L44 MPI-ESM-P x98 x L47 256x220 x L40 MRI-CGCM x160 x L48 364x368 x L51 Table 2. The PMIP3 models used in this analysis, the model years averaged over for the mid-holocene and pre-industrial climatologies, and the resolutions of the atmosphere and ocean components of each model Mid-Holocene Precipitation Signal We examine the difference (mid-holocene minus pre-industrial) in the multimodel mean of tropical precipitation over land and ocean for the boreal winter (DJF) and boreal summer (JJA) as shown in Fig In JJA when the Earth is receiving increased insolation, precipitation tends to increase over land and decrease over the ocean. The opposite is true during DJF when the Earth is receiving less insolation. The majority of

22 10 models agree on the distribution of the precipitation change, and a distinct shift in precipitation from land to ocean is apparent. Figure 2.1: Multi-model mean seasonal precipitation change (mm/day) between experiments (MH-PI) for the 10 PMIP3 models (listed in Table 2). Stippling indicates regions where 7 out of 10 models agreed on the sign of the precipitation change. This signal is even clearer when examining the multi-model mean seasonal cycle of precipitation change over tropical land and ocean (Fig. 2.2). In boreal summer, the precipitation changed is maximized over land and minimized over ocean. Again, the opposite is true in boreal winter. This result shows that many models respond to precessional forcing in a similar manner, and it reinforces the importance of understanding the mechanism that drives this distinct precipitation signal.

23 11 mm/day month Figure 2.2: Multi-model mean seasonal cycle of the precipitation difference (mm/day) as a function of time (months) between experiments (MH-PI) for the 10 PMIP3 models (listed in Table 2) over land (solid) and ocean (dashed). Shading indicates the model spread. Curves are scaled by the percentage of land and ocean in the tropics.

24 Chapter 3. Effects of Precession in Idealized Simulations Now that we have identified a robust precipitation signal in the PMIP3 mid- Holocene experiments, two climate models are used to examine the mechanism controlling changes in tropical precipitation as a result of idealized precessional forcing. We apply a top-of-atmosphere energetic framework outlined in Merlis et al. (2012) to understand the changes in precipitation over the land and ocean. A detailed analysis of the mechanisms controlling the precipitation response is performed, and we examine changes in the dry and moist energy budgets in order to further constrain changes in precipitation Data and Methods Models and Experiments We use idealized simulations from two climate models to identify the effects of precessional forcing on precipitation. The first model is the Geophysical Fluid Dynamics Laboratory (GFDL) Coupled Model version 2.1 (CM2.1), which has atmosphere, ocean, land, and sea ice components. The atmosphere and land models have a horizontal resolution of 2 latitude x 2.5 longitude, and the atmosphere has 24 vertical levels. The ocean model has a horizontal resolution of 1 latitude x 1 longitude, with a latitudinal resolution that gradually decreases equatorward of 30 so that it is 1/3 at the equator, and 50 vertical levels (Delworth et al. 2005). The second model is the National Center for Atmospheric Research Community Atmosphere Model version 3 (CAM3), which is coupled to a slab ocean model (SOM), land model, and thermodynamic sea ice model. The atmosphere and land models have a horizontal resolution of 2.8 latitude x

25 13 longitude (T42 resolution), with 26 vertical levels in the atmosphere. The SOM simulates thermodynamic effects but does not include full ocean dynamics; it has a spatially varying mixed layer depth and employs a seasonally varying horizontal ocean heat transport, which is calculated from an atmospheric general circulation model run with fixed sea surface temperatures. Additional details of the CAM3-SOM configuration can be found in Murphy et al. (2009). Two experiments were run with each model in which the precessional signals are greater than those in the PMIP3 experiments discussed in the previous chapter. In the first experiment (hereafter referred to as the summer solstice "SS" experiment), perihelion is set at the NH summer solstice (ω 180=270), and in the second experiment (hereafter referred to as the winter solstice "WS" experiment), perihelion is set at the NH winter solstice (ω 180=90). Because eccentricity modulates the strength of precessional forcing (Jackson and Broccoli 2003), eccentricity is increased to in GFDL-CM2.1 and 0.05 in CAM3-SOM in order to maximize the model's response to the precessional forcing. All other parameters, including obliquity, are set to pre-industrial levels. These experiments were designed to study the precessional response and do not correspond to any particular time periods. The GFDL-CM2.1 simulations were run for 400 years. The CAM3-SOM simulations were run for 30 years. Complete descriptions of the experimental designs for GFDL-CM2.1 and CAM3-SOM can be found in Erb et al. (2013) and Murphy (2010), respectively. The simulations from CAM3-SOM were used for a study of the Messinian Salinity Crisis (Murphy, 2010), a time where the Mediterranean Sea was isolated from the Atlantic Ocean. For this reason, the ocean heat flux divergence prescribed to the SOM

26 14 is zero between the Atlantic Ocean and the Mediterranean Sea. The resulting climate is affected by this change (Murphy et al. 2009), but because the purpose of this study is to understand the mechanisms controlling precessional changes in precipitation and not the fine spatial patterns, the changes in the mean climate due to the Messinian Salinity Crisis do not affect our conclusions. Finally, a calendar adjustment has been made to the GFDL-CM2.1 precession simulations in order to account for the difference in lengths of the seasons that arises from changes in the longitude of the perihelion (Pollard and Reusch, 2002). The adjustments in the tropics during the summer and winter seasons are small and do not have a large impact on our results (discussed in section 3.2), however it is important to note that slight artificial differences may be introduced as a result of the calendar problem when looking at results of the seasonal cycle (Joussaume and Braconnot, 1997) Top-of-Atmosphere Energetic Framework We examine the last 100 years of each experiment from GFDL-CM2.1 and the last 10 years of each experiment from CAM3-SOM and compute seasonal mean climate changes between the two experiments (SS-WS). In order to explain precipitation changes, we use the energetic framework developed by Merlis et al. (2012). Those authors constrained the response of the monsoonal Hadley circulation to precessional forcing with an energy budget at the topof-atmosphere. Their framework was applied to idealized aquaplanet model simulations with a zonally symmetric continent. Here, we extend this framework to GFDL-CM2.1 and CAM3-SOM, which have a more realistic land configuration.

27 15 From Merlis et al. (2012), the energy balance of the atmosphere and surface is!! +!!!!"!!!! =!!"#!!"#!h!!, (1) where the time-mean is denoted by and the vertically mass-weighted integral is denoted by.! =!!! +!" +!" is the total atmospheric energy and h =!!! +!" +!" is the moist static energy (MSE). S TOA is the net top-of-atmosphere (TOA) shortwave radiation, L TOA is the net TOA longwave radiation, ρ o is the ocean density, c po is the ocean heat capacity, d is the ocean mixed layer depth, T s is the surface temperature,!h is the MSE flux divergence, and! is the ocean energy flux divergence. If atmospheric and land energy storage are considered negligible, the balance in (1) becomes!!"#!!"#!!!!!"!!!! +!h +!!, (2) where! = 1 over oceans 0 over land. If we consider insolation changes, it follows that!!!"#!!"#!!!!!"!"!!! +!!h +!!!. (3) Over land, the surface storage term and the ocean energy flux divergence are zero, so changes in the net (shortwave minus longwave) TOA radiation are balanced solely by changes in the MSE flux divergence. Over ocean, changes in net TOA radiation

28 16 can be balanced by changes in atmosphere and ocean energy flux divergence, as well as changes in surface energy storage Linking the Energy and Moisture Budgets While Eq. (3) is useful because it describes the energy balance between the atmosphere and the surface, it does not explicitly include changes in precipitation. Therefore, we link the energy and moisture budgets to understand how changes in net TOA radiation lead to changes in precipitation. One way to link the energy and moisture budgets is to consider the atmosphere in radiative-convective equilibrium. The surface warms through radiative heating and convection heats the upper atmosphere through precipitation and latent heat release. Therefore, in monsoon regions where sufficient surface moisture is available, changes in net TOA radiation over land are balanced by changes in precipitation:!!!!!!"#!!"#!!h, (4) where P L is the precipitation over land Results Changes in Insolation Fig. 2.1 shows the changes in insolation for both GFDL-CM2.1 and CAM3-SOM. The Earth receives more solar radiation in the boreal summer and less solar radiation in the boreal winter. The effects of the calendar adjustment discussed in section can be seen in the second half of the year for CAM3-SOM. The difference in the lengths of the

29 17 seasons comes into affect during this time and leads to an asymmetrical response in insolation. This is not present in the GFDL-CM2.1 insolation change because the lengths of the seasons have been accounted for by applying the calendar adjustment to the data. Although the difference in the lengths of the seasons has a large effect on insolation in the mid-latitudes and the poles, the effect on the tropical insolation (30 S to 30 N) is less significant, especially during the summer and winter months. Figure 3.1: Latitudinal distribution of the zonal-mean insolation difference (W/m 2 ) as a function of time (months) between experiments (SS-WS) for each model Net TOA Radiation, Precipitation, and Surface Temperature Fig. 3.2 and 3.3 show the seasonal differences in net radiation at TOA and precipitation between SS and WS in the tropics for both models. In DJF, the change in net TOA radiation is negative and precipitation decreases over land and increases over

30 18 the ocean, while in the JJA, the change in net TOA radiation is positive and precipitation increases over land and decreases over the ocean. This can also be seen in Fig. 3.4, which shows the seasonal cycle in precipitation change over land and ocean for each model. It is clear that during the months where changes in precipitation over land are positive, the changes in precipitation over the ocean are negative. Figure 3.2: Seasonal net TOA radiation difference (W/m 2 ) between experiments (SS-WS) for each model. We find that these changes in precipitation cannot be explained simply as a response to the differences in warming (or cooling) between land and ocean. This can be seen, for example, in the seasonal difference in surface temperature in the tropics shown

31 19 in Fig Throughout much of the domain, the surface cools in DJF and warms in JJA, as expected. However, over land, there are areas of warming in DJF (regions of Africa, South America, and Australia) and cooling in JJA (regions of Africa, South America, and India), which are not explained by the differential heating between ocean and land. Moreover, the largest changes in precipitation occur over these land regions. The regions with maximum decreases in precipitation are in fact regions of warming in DJF, and regions with increases in precipitation are regions of cooling in JJA. These surface temperature signals over land are due to changes in cloud cover and will be discussed in chapter 4. Figure 3.3: Seasonal precipitation difference (mm/day) between experiments (SS-WS) for each model. Contours indicate the mean seasonal WS precipitation.

32 20 Figure 3.4: Seasonal cycle of the precipitation difference (mm/day) as a function of time (months) between experiments (SS-WS) for each model over land (solid) and ocean (dashed). Curves are scaled by the percentage of land and ocean in the tropics. Figure 3.5: Seasonal surface temperature difference (K) between experiments (SS-WS) for each model.

33 MSE Flux Divergence Next, we look at the relationship between changes in precipitation and MSE flux divergence in GFDL-CM2.1 and CAM3-SOM. The MSE flux divergence is calculated as a residual term from the TOA and surface energy fluxes; it is not calculated explicitly because the use of monthly data in this study does not capture the MSE flux divergence on shorter timescales. Figure 3.6: Seasonal MSE flux divergence (W/m 2 ) between experiments (SS-WS) for each model. Contour lines are changes in precipitation minus evaporation (mm/day). Eq. (3) describes changes in net TOA radiation as a balance between changes in MSE flux divergence, temperature tendency, and ocean heat flux divergence. Over land, changes in temperature tendency are very small and ocean heat flux divergence is zero.

34 22 Therefore, when there is an increase in net TOA radiation, the atmosphere responds with an increase in the MSE flux divergence, and it follows from Eq. (4) that precipitation will increase. The response is more complex over the ocean because changes in the TOA energy budget are balanced by changes in surface energy storage, ocean energy flux divergence, and atmosphere energy flux divergence. Fig. 3.6 shows the MSE flux divergence and changes in precipitation minus evaporation (P E) in both models. Over land, regions of increased P E do indeed overlap with regions of MSE flux divergence, and regions of decreased P E overlap with regions of MSE flux convergence in both models (excluding the region in JJA in GFDL-CM2.1 where there are large changes in mid- and low-level clouds, discussed in chapter 4). Over the ocean, the spatial patterns are more complicated and it is difficult to draw conclusions about the relationship between P E and MSE flux divergence. In order to gain a better understanding of the energy balance and resulting change in precipitation over the ocean, we perform a more detailed analysis on the TOA, surface, and atmosphere energy balances in the following section Energy Budget Balance Now that we have seen the spatial patterns of the changes in net TOA radiation, precipitation, and MSE flux divergence, it is useful to examine the seasonal cycle of the energy budget over the land and ocean as well as the corresponding changes in precipitation. Fig. 3.7 shows the tropical-mean changes in net TOA radiation, net surface fluxes, MSE flux divergence, and precipitation over land and ocean for both models. As expected from Eqs. (3) and (4), changes in precipitation have the same sign as the

35 23 changes in net TOA radiation over land surfaces. This is because over land, the MSE flux divergence needs to compensate for all changes in the net TOA radiation. The atmosphere does this through latent heat release of precipitation, hence increases in net TOA radiation correspond with increases in precipitation and decreases in net TOA radiation correspond with decreases in precipitation. Figure. 3.7: Seasonal cycle of the difference (SS-WS) in tropical-mean (30 N to 30 S) precipitation (blue), net TOA radiation (orange), net surface energy fluxes (purple), and MSE flux divergence (red) in W/m 2 as a function of time (months) over land and ocean for each model. Curves are scaled by the percentage of land and ocean in the tropics.

36 24 Over the ocean, the sign of the change in net TOA radiative forcing is the same as it is over land, but the change in precipitation is opposite, as we have seen in Figs. 3.2 and 3.3. This difference arises as a result of the ocean energy storage. Fig. 3.7 indicates that over the ocean, changes in net TOA radiation are almost entirely balanced by changes in the surface energy fluxes. The ocean responds to the increase in net TOA radiation during JJA by absorbing energy and warming the surface, and it responds to the decrease in net TOA radiation during DJF by releasing energy and cooling the surface. As a result, the change in MSE flux divergence over the ocean is very small; the atmosphere over the ocean does much less work to adjust to the TOA forcing than the atmosphere over land. Although this analysis is useful for understanding the atmospheric energy budget over both land and ocean, it does not explicitly explain why precipitation decreases over the ocean when net TOA radiation is increasing, and why precipitation increases over the ocean when net TOA radiation is decreasing. In order to further understand this detail, we examine the vertical profiles of the dry and moist energy of the atmosphere over land and ocean Dry and Moist Energy Fig. 3.8 shows the change in the vertical profiles of the dry and moist energy over land and ocean in DJF and JJA for both models. The dry energy is taken as c p θ, where θ is the potential temperature and c p is the specific heat for dry air, and moist energy is taken as L v q, where q is the specific humidity and L v is the latent heat of vaporization. The moist energy is most important in the lower levels of the troposphere, and it

37 25 decreases during DJF where the surface is cooling and increases in JJA when the surface is warming. The profile of c p θ has small seasonal change over land and a larger change over ocean. In DJF, the atmosphere over the ocean becomes more unstable (dθ/dp > 0) and in JJA the atmosphere over the ocean becomes more stable (dθ/dp < 0). Figure. 3.8: Vertical profile of the difference (SS-WS) in tropical-mean (30 N to 30 S) dry (red) and moist (blue) energy (W/m 2 ) as a function of pressure (mb) over land (solid) and ocean (dashed) for each model. Curves are scaled by the percentage of land and ocean in the tropics.

38 26 It is this reorganization of the thermodynamic structure of the atmosphere over the ocean, which is driven by circulation changes over land, that ultimately allows for a change in precipitation. This mechanism is depicted in Fig In JJA when there is a net gain in the TOA radiation, the atmosphere over land responds to this forcing through an export of moist static energy (Fig. 3.6), which is achieved through anomalous vertical motion and an increase in precipitation. It is this circulation change over the land that drives anomalous descent over the ocean. As we saw in the previous section, the atmosphere over the ocean does not import more energy; it instead adjusts by becoming more stable, which is consistent with less rainfall. In DJF when there is a decrease in the TOA net radiation, the atmosphere over land responds with anomalous descent and a decrease in rainfall. The circulation over the ocean is driven by the land s response and there is anomalous ascent over the ocean. Because the atmospheric energy over the ocean isn t changing, it responds by changing the thermodynamic structure; the atmosphere become more unstable and an increase in precipitation occurs.

39 Figure. 3.9: Illustrates the land and ocean response to TOA forcing in DJF (top panel) and JJA (bottom panel). Red and blue arrows denote net energy gain and loss, respectively, at the TOA and surface. Black arrows denote circulation change. Red lines indicate the change in dθ/dp over the ocean and the cloud represents precipitation. 27

40 Summary In this chapter, we examined the precipitation response to an idealized precessional forcing using GFDL-CM2.1 and CAM3-SOM. Although there are slight inter-model differences resulting from model setup and experimental design, both models agree that the changes in precipitation over land can be explained by changes in MSE flux divergence. We found that while the energetic mechanism outlined in Merlis et al. (2012) explains how precipitation over land responds to precessional forcing, it does not fully explain the precipitation response over the ocean. The fundamental constraint dictating the energy balance and resulting precipitation change over land and ocean lies in the surface boundary condition. Over land, the surface stores very little energy and the atmosphere is required to do all the work to compensate for the net TOA energy imbalance. Over the ocean, the surface storage is important. In fact, the net TOA energy imbalance is compensated almost entirely through a warming or cooling of the ocean surface and as a result, the change in the moist static energy of the atmosphere is very small. Therefore, it is the differing surface properties of the land and ocean that allow for very different atmospheric responses to the precessional forcing. The thermodynamic structure of the atmosphere becomes necessary to explain how precipitation can change over the ocean despite little change in the atmospheric energy budget.

41 Chapter 4: The Precessional Signal in the Tropical Atlantic In this chapter, we discuss the regional changes due to precessional forcing in the tropical Atlantic and Africa. A more localized analysis of the climate signal and energy budget can provide additional insight into the mechanism controlling precipitation changes. A better understanding of regional change in idealized and mid-holocene experiments is also very important for comparisons to the precessional signals found in proxy records from the tropical Atlantic. In the following sections, we discuss the regional changes in net TOA radiation, surface temperature, precipitation, and cloud cover in order to gain a more complete understanding of the processes controlling precessional changes in the tropical Atlantic. We first discuss the regional changes in surface temperature in the idealized precession simulations from GFDL-CM2.1 and CAM3-SOM. Then, we examine the regional precipitation changes in the PMIP3 mid-holocene experiments and discuss the resulting implications for model verification with paleodata Explaining Regional Changes in Surface Temperature Recall from section the land regions of isolated warming in DJF and widespread cooling in JJA that are unexplained by changes in insolation in both GFDL- CM2.1 and CAM3-SOM. These unexpected changes in surface temperature lead us to the question: Why is the change in surface temperature in some regions opposite to the change in insolation? Clouds play an important role in answering this question; they explain not only these regional changes in surface temperature, but also isolated changes in net TOA radiation. 29

42 30 Figures 4.1 and 4.2 show the regional changes in precipitation and surface temperature in GFDL-CM2.1 and CAM3-SOM. Note the region of surface cooling in JJA that corresponds with a region of increased precipitation over central Africa. Although not as drastic, there are also regions of isolated warming in DJF that correspond to large decreases in precipitation over Africa. These regional variations in the patterns of warming during JJA and cooling during DJF seem to be explained by changes in cloud cover (Fig. 4.3). The areas with increased precipitation show a regional increase in cloud cover, which cools the surface due to higher cloud albedo, while the areas with decreased precipitation show a regional decrease in cloud cover, which warms the surface. Thus, changes in cloud cover are an important feedback that can lead to a surface temperature response that is not expected from the initial forcing. Figure 4.1: Seasonal precipitation difference (mm/day) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM.

43 31 Figure 4.2: Seasonal surface temperature difference (K) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM. Figure 4.3: Seasonal cloud cover difference (%) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM.

44 32 Changes in cloud cover can also explain the regional changes in net TOA radiation that differ significantly from the precessional changes in insolation. To see this, we examine the region in the western Sahel during JJA in GFDL-CM2.1. The relationship between net TOA radiation and precipitation over land described in Eq. (4) explains the changes seen in both GFDL-CM2.1 and CAM3-SOM except for in this region. That is, while the rest of the domain has an increase in net TOA radiation, the region in the western Sahel is dominated by a decrease in net TOA radiation (Fig. 4.4). Figure 4.4: Seasonal net TOA radiation difference (W/m 2 ) between experiments (SS-WS) for GFDL-CM2.1 and CAM3-SOM. Boxes denote the regions in the eastern and western Sahel for Fig In order to understand this response, we examine the JJA change in cloud cover in two regions (boxed regions in Fig. 4.4) shown in Fig Region 1 represents the western Sahel where there is a decrease in net TOA radiation and region 2 represents the