This page's content is no longer actively maintained, but the material has been kept on-line for historical purposes.
The page may contain broken links or outdated information, and parts may not function in current web browsers.

Research Results

Ocean-Atmosphere Models: Connecting Regional and Global Climate


Climate is the average condition of the weather of a particular region for a specified time. The climate is always changing due to forcings and chaos. All life in all forms on earth are affected by changes in climate, and vice versa. This is because availability of sunlight, food, oxygen, CO2 and other life sustaining factors depend on climate. So the climate forms a system in which all factors involved affect each other. However, since the last few decaded, it has been noticed that there has been more severe changes in the earth's climate than there has ever been for a long time. There has been a real, but irregular, increase in global surface temperature since the late nineteenth century. There has also been a marked, recession of the majority of mountain glaciers over the same period (Houghton, pg. 200). These, and other facts prove that the world has, indeed, been warming over the last century.

It is very important to understand how our climate system is changing and how it will continue to change in the future. Computer models are used to simulate these changes. A model is a computer program that includes a set of mathematical relationships between variables known to affect earth's climate. This model is then simulated for a time period in order to determine climate change. In this study, we focus attention on the new GISS (Goddard Institute for Space Studies) modelE. We will be conducting a fundamental validation study of the radiological and hydrological processes in the model. This is important since many climatic properties like the cloud cover and the heat budget are affected by these two properties. We will also try to understand how the hydrological and radioligical variables in the new model simulations compare with the observation; How are these variables related in all the different versions of the model? What scientific questions about climate can the different versions of the model answer?

Three different Ocean models are used because they all can help understand and simulate different processes. They are:

  • Atmospheric Model
  • Qflux Model
  • Coupled Model

Atmospheric Model

The Atmospheric Model is an Atmospheric General Circulation Model (AGCM). It is a model in which the Sea Surface Temperatures (SSTs) are held constant. The atmospheric variables however are allowed to change over time. Since the correct values of the SSTs are used in this model, it can be used to predict, for example, how the atmosphere responds to Ocean changes over a given time period. This is the simplest form of model that we will be considering and since all SSTs are fixed, this model will be the most similar to actual observation data.

Qflux Model

On the Ocean, two main changes affect how the Sea Surface climatic properties change over time. They are the incoming streams from the sides, and the changes from above and below the Ocean. The streams from the sides are mainly ocean currents. Ocean currents are huge amounts of ocean water moving independently of the rest of the ocean. They are driven by winds, and the corriolis effect affects the direction in which they move. The Corriolis effect is a deflection of air motion to the right of the wind flow in the northern hemisphere and to the left of the wind flow in the southern hemisphere due to the force of gravity (World's Oceans, pg.195). Changes to the ocean from the top is due to the amount of sunlight that is coming in and out of the ocean which is also affected by the amount of clouds present above the ocean regions.

In the Qflux Model, in addition to the changing atmosphere, the amount of heat energy from the top and the bottom of the ocean is allowed to change, while the incoming streams from the side are fixed. This type of model will allow us to answer questions such as how the Atmosphere or the Ocean responds to these fixed incoming streams from the side, or how the atmosphere affect and is affected by the changes on the Ocean surfaces. It will also show the response of the ocean surface temperature to changes in the heat budget (Duxbury pg.162)

Coupled Model

The Coupled model is a combination of an Atmospheric General Circulation Model (AGCM), and an Ocean General Circulation Model (OGCM). In this model, the atmosphere, the incoming streams as well as the Ocean fluctuations from above and below the Ocean surface are allowed to change over time. This model is expected to be the most different from the observations. Since all variables are allowed to change. It will explain much scientific reasoning about how the future climate will change.

All of these different versions of the model will be compared between themselves and also with Observation data that from satellites and data collection centers around the world. We will be testing this model to see how it compares with the observation data, as well has how the different versions compare between themselves.

It is important to carry out this task because, if climate models are going to predict future changes accurately, they need to be verified, to ensure that the different scientific phenomena that the simulate are represented accurately. In view of this, we carried out our research by considering the surface air temperature, the sea level pressure, and other radiological and hydrological variables and examined their behavior in the model and observation data.


The purpose of this research is to analyze the GISS GCM (Global Circulation Model) and determine the relationship between the three different versions of the new model and between the observation data.

Model data sets

The GISS GCM uses a 72×46 grid dimension array. If you take a look at any map of the world, you will observe imaginary lines called lines of longitude and parallels of latitude. These lines help divide each region of the world into different grid boxes. They run 180 degrees east to west and 90 degrees north to south. We refer to each of the rectangular intersections that they form as a grid box. A more coarse scale of this dimension — 72×46 is used by the GCM. Therefore, each grid box covers about 5×4 degrees. So, about from New York to parts of Canada will be represented by one grid box. Due to limited computing speed, and the huge set of data involved, it will be very time consuming as well as expensive to run a finer resolution.

What is the significance of the coarse resolution on the results that we obtain? Most climatological variables do not change significantly over a 5×4 degree dimension. For example, the temperature in the whole of New York does not vary excessively at different parts on any given day. Of course, a finer resolution will yield better results since some variables such as precipitation change a lot over small distance. However, since we are considering the climate change (not weather change) over the whole world for long periods of time (about 5 years) the coarse resolution does not significantly reduce the validity of our results.

Observation Data Sets

One of the observation sets was obtained from the National Center for Environmental Prediction (NCEP). Weather stations around the world gather data for NCEP use. However, since there are no weather stations for every grid box in the world, not even every region, the collected data is run in a model that approximates the data for grid boxes where no data was collected. So the NCEP data is a combination of model and observed data. The limitation of this is that if the model does not work accurately, then the data for the approximated regions will be incorrect. On the other hand, using NCEP data ensures that there is data for every region of the world, which makes it much easier to compare the model and the observation.

Some of the data sets that we used was collected from satellites orbiting the earth, e.g. the International Satellite Cloud Climatology Project (ISCCP), the Earth Radiation Budget Experiment (ERBE). Data collected this way is essential because there are not enough weather stations collecting data from the earth surface. So the satellite provides data for regions of the world where there are no people or weather stations. Most of this data is collected by remote sensing technology. Infrared radiation could be sent down to earth and the time taken for reflection is measured to calculate, for example, the height of cloud cover. So several remote sensing techniques are used along with several mathematical relationships to obtain observational data for different variables. Since the calculations are based on known scientific concepts, the satellite data is considered as a reliable data source.

Airplanes or balloons flying over specific regions of the earth collect another set of data. For example, flying over ice caps and taking aerial photographs of regions with snow cover can obtain snow cover data. Air balloons are also sent out from weather stations to measure several variables like precipitation or cloud cover.

Statistical Significance Of Results

To determine the differences between each version of the model, we made difference maps which are basically maps of values different regions subtracted from each other. However, there is always a change in the climate of the earth in real life and in the model as well every year. This is known as inter-annual variability. So, we decided to determine how much of the differences are due to the inter-annual variability, rather than to actual errors within the model. The Images below help illustrate the fact that there is inter-annual variability between the models for different years. They show the standard deviation for 5 years, which is an average measure of how the values change within each of these 5 years.

Figure 1

Figure 1: The Standard deviation for the SST model is close to 0, as you see a lot of purple colors, especially in the Oceans, since the sea surface temperatures are fixed

Figure 2

Figure 2: The Standard deviation values for the Qflux model are slightly higher than in the SST model. Notice more blue colors, showing standard deviation between 0.5 and 1.This is reasonable since only the streams are fixed in this model.

Figure 3

Figure 3: There are larger standard deviation values for the Coupled model. Everything is allowed to change here, so the increase in standard deviation is consistent with the science behind these models.

So as we can see from these images, there is indeed inter-annual variability for each of the different ocean models. The progressive rise in standard deviation with the lowest in the SST model and the highest in the Coupled model, is consistent with the science behind this models. The SST has the most fixed values, therefore, should have less standard deviation . The Coupled has the most changing values, and should have more deviation.

Are the differences that were observed in the model due to inter-annual variability or actual errors in the model? To answer this question, we conducted a T significance test. This statistical, test divides the difference between the model versions by the standard deviation for the different years. We used 5 years of data and did this for two variables-Surface Air Temperature & Sea Level Pressure. These two variables are important radiological and hydrological variables.

The differences compared are the Qflux - SST and the Couple - SST models. This will help us to understand how the Qflux deviates from the SST model and how the Couple deviates from the SST. This is also important since both versions of the model have the SST as their base model.


Surface Air Temperature

For the Qflux Model, most of the differences observed are not very significant, except for the south pole sea ice regions. From the image (Fig 4) we see that most of the world regions have color white, which shows significance values between +1 and -1. However, the differences in the south pole Sea Ice regions are actually very significant, being greater than +11. This may be due to the fact that in the SST model, the sea surface temperatures are fixed, but in the Q flux model, they are allowed to change, leading to significant differences.

For the Coupled model (Fig 5), we see a lot of significant difference in the Sea Ice region, the North Atlantic region, and the El Nino region. In the Sea Ice region the differences are very significant just as they are in the Qflux Model. In the North Atlantic region, the differences are also significant, with very significant negative values. Notice that despite the fact that that in the difference maps, there is slight of differences in the El Nino region, but these small differences are actually very significant.

Sea Level Pressure

For the Qflux model, no significant differences are observed. These shows that there is very little changes between the Qflux and the SST versions of the model. This is a positive results because it shows that for the sea level pressure variable the SST and the Qflux models are very similar in the results they simulate.

The Coupled model shows more significant differences. We see a lot of significant differences for the NAO region and this can be attributed to the known Northern Oscillation in this region. Also, in the equitoral region, we see a lot of significant differences. This can be attributed to the West Pacific Warm Pool, that causes changes in these regions and therefore, significant positive differences.

Figure 4

Figure 4: This shows the T significant values for Surface Air Temperature in the Qflux model. As you can see, the most significant values occur in the southern hemisphere sea-ice region.

Figure 5

Figure 5: These are the T significant values for the Coupled model, Surface Air Temperature. Significant T values are in the NAO, El Nino and Sea Ice regions. Notice the very significant values in the El Nino region.

Figure 6

Figure 6: These are the T significant values for the Qflux model, Sea Level Pressure. Most of the map has values from 1 to +1 (white regions). There are no significant differences here.

Figure 7

Figure 7: These are the T significant values for the Coupled model, Sea Level Pressure. Most significant values are in the NAO region and the West Pacific Warm Po l(equatorial) region.


Most of the significant results keep occurring in the NAO region, the El Nino region and the south pole sea ice regions. In the NAO and El nino regions, the a known modes of variability in the earth's climate, and this may be the reason for the significant differences that we observe. For the south pole sea ice region, we expected some significant differences since the sea surface temperatures are allowed to change in the Qflux and Coupled model, while the are seasonally fixed in the SST model. However, we did not expect high significance values ranging from +11 to 11 as seen in our results. It is yet unclear if these high significance values can be attributed to changing sea surface temperatures alone.

From our results we also get an idea of how two of the hydrological and radiological variables Sea level pressure and Surface air temperature, are related to each other in the models. The fact that for the two variables, we see high and low significant values in similar regions suggest that these hydrological and radiological are indeed represented in a similar fashion in the different ocean models.

For further study, we will try to understand more why significant differences are observed in the NAO and El nino regions and the sea ice. We will try to determine if these might be related to errors in the model. We will also study the significance of the observed differences for other variables such as Net Thermal Radiation, Absorbed Solar radiation, Cloud Cover, precipitation and geo-potential heights.


Duxbury, C Alyn. An Introduction to World's Oceans Wm.C.Brown Publishers: USA, 1994.

Houghton J.T. Climate Change: The IPCC Scientific Assessment Press Syndicate of the University of Cambridge: New York, 1993

National Centers for Environmental Prediction.