Table of Contents
- 1. Research Setting
- 2. Primary Research Question
- 3. Investigation Plan
- 4. Data Access and Visualization Methods
- 5. Preliminary Analysis
- 6. Refinement of Analysis
- 7. Statement of Results
- 8. Discussion of Results
- 9. Statement of Conclusions
- 10. Questions for Further Investigation
1. Research Setting
The research setting for this tutorial will return to a well-known region, the North Atlantic Ocean. Refer to the Educational Module "5. Seasonal Variability" for some monthly images of the North Atlantic Bloom, the increase in phytoplankton productivity that occurs there each spring.
2. Primary Research Question
The primary research question for this tutorial will concern an initial examination of data that has recently been made available in Giovanni: output data from the NASA Ocean Biogeochemical Model (NOBM). NOBM utilizes remotely-sensed chlorophyll datato adjust the model-generated "fields", which are the values of variables in the model. The first model variable that will be examined is Total Chlorophyll (CT). CT is based on calculations in the model which generate phytoplankton populations, and these population estimates provide an initial value for chlorophyll concentration. The available satellite data is used to adjust the chlorophyll concentrations generated by the model. The adjusted chlorophyll concentrations are then used bythe model for the next daily calculation step. This data assimilation process is intended to keep the model consistent with the actual state of the oceanic environment.
So the primary research question can be stated as:
How can total chlorophyll from the NOBM improve our understanding of the dynamics of a highly variable oceanic process like the North Atlantic Bloom (NAB)?
3. Investigation Plan
Because the primary research question concerns the North Atlantic Bloom, it can be quickly inferred that the study area is the North Atlantic Ocean. This study area is a significant one for CT for two reasons; one, during the NAB, distributions of phytoplankton and the chlorophyll within them changes rapidly, and two, the North Atlantic frequently has a lot of clouds over it! CT is particularly useful for cloudy areas. The time period for the study is spring in the Northern Hemisphere.
First, the images below illustrate why CT might be a useful data product. These are browse images of a SeaWiFS imaging swath acquired on July 24, 2005. The "true color" image is on the left, and the chlorophyll concentration image is on the right. Greenland and Iceland are visible in the true color image, but Newfoundland was completely covered with clouds. A few small areas where the ocean surface could be glimpsed by the sensor are highlighted.
Our investigation plan will examine standard ocean color data products generated for a period during the NAB by Giovanni, and compare them to the CT images for the same period and region.
4. Data Access and Visualization Methods
The data access system we will be using is Giovanni, which provides access to SeaWiFS 8-day data products and to the NOBM monthly and daily data products. Both data sets can be accessed here:
Ocean Color Giovanni
5. Preliminary Analysis
First, let's look at a SeaWiFS 8-day image of the North Atlantic for the period July 20 to July 27, 2005.
Despite the fact that this is an 8-day image -- meaning that it shows all the data acquired by the sensor over the entire 8-day period -- much of the region is still not visible, due to pervasive cloud cover. Now let's take a look at the NOBM CT image for the same period of time (the method for computing CT will be explained later in the tutorial). The same color scale is used for both images.
This image shows what the assimilation method does -- it fills in the areas where the data was missing, based on the data that was acquired by the sensor. Notice that the spatial resoluton is lower, giving the image a "blocky" appearance, and that most of the coastal areas are excluded. Because the accuracy of chlorophyll concentration estimated by the sensor is lower in areas with shallow or turbid water, the model doesn'ttry to estimate chlorophyll concentrations in shallow coastal waters.
Now let's examine the basics of how CT is computed in the model. At the end of this tutorial is a link to a paper describing the entire NOBM model -- to summarize it briefly, the model includes ocean circulation, nutrient data, data on the amount of light reaching the ocean surface (radiation), sea surface temperatures, wind speeds, etc., all used to compute the populations of four major phytoplankton groups. The populations of the phytoplankton groups are used to generate a value of total chlorophyll concentration. The chlorophyll data from the satellite are used to keep the model "honest", i.e., if the model chlorophyll concentrations were much too high or much too low for a given region, the assimilation of the satellite data corrects the model in the proper direction.
So the model generates daily chlorophyll concentration values based on oceanic and biological dynamics. Data assimilation with satellite data can be accomplished whenever there is actual satellite data -- which explains the importance of the "glimpses through the clouds" of the ocean surface. Each of these data points is used in the assimilation method to adjust the model according to the actual observations. There are a few rules that the model follows to minimize errors in the satellite data:
- Any daily observational data value that is 2 times greater than the monthly average value for any location is excluded;
- The data is "weighted", giving 25% of the weight to the monthly average data value, 75% of the weight to the daily data value;
- Any data acquired where the model indicates the presence of sea ice is excluded; and
- The model chlorophyll concentrations are given a greater "weight" for areas where known conditions may produce inaccurate estimates of chlorophyll concentration from the satellite.
In summary, the NOBM produces estimates of total chlorophyll concentration that are adjusted by satellite observations of chlorophyll concentration.
6. Refinement of Analysis
To provide an improved picture of how this works, here are some images of the same area covered by the observational satellite swaths shown above. The map projection used here makes Greenland look a lot bigger than in the actual view from space. The left image shows the 8-day SeaWiFS data for July 20-27, and the right image shows the CT data averaged over the same period of time (the NOBM output is daily, so the output was averaged over the period July 20-27).
Remember that the NOBM data output has been modified by the assimilation of any actual observational data acquired by SeaWiFS. So even though it was mostly cloudy, in the places that SeaWiFS was able to acquire data, the chlorophyll concentration values will be similar.
One advantage of NOBM is that it provides a sense of the changes occurring daily. The image sequence below shows the NOBM output for each day in the period July 20-27, 2005.
7/20 7/21 7/22
7/23 7/24 7/25
Looking at this sequence, there are a few intriguing highlights. To the west of Greenland, the chlorophyll concentrations in the Labrador Sea (north of Newfoundland) are decreasing. Southeast of Greenland, chlorophyll concentrations are increasing. The chlorophyll concentrations that can be seen south of Iceland in the SeaWiFS image are slightly higher than in the NOBM data.
7. Statement of Results
The brief introduction shown here indicates that the NOBM CT data provide a way to observe and estimate chlorophyll concentrations in areas with extensive cloud cover. For regions and periods with significant variability on short time-scales (i.e., days to weeks), the CT data allow estimation of the changes occuring where satellite observations were prevented by atmospheric conditions.
Clearly, the NOBM CT data do allow a "view under the clouds" of the changing patterns of chlorophyll concentration during the North Atlantic Bloom on a daily basis. The question immediately arises: how accurate is the NOBM CT data? The paper provided at the link below describes several ways that the accuracy of the model data was evaluated. One of the methods was to use the daily assimilated chlorophyll values to calculate a monthly average assimilated chlorophyll value, and then compare this value to the monthly average SeaWiFS chlorophyll value. For the North Atlantic region in March 2001 (which is before the main period of the NAB), the agreement between the model and SeaWiFS chlorophyll was good.
It should be remembered that the data assimilation process uses observational data values to adjust the model chlorophyll concentration values. Therefore, for areas in the model that are close to where an observational data point was acquired, the model values in those areas will be adjusted closer to the observational data value than for areas farther from the observational data point. Also, the lower model resolution compared to the SeaWiFS resolution of 9km means that a single SeaWiFS data value will affect a larger area in the model.
Finally, the North Atlantic is not a region where there are known conditions that affect the accuracy of SeaWiFS data. So this is an area where a reasonable agreement between the model and the SeaWiFS data would be expected.
9. Statement of Conclusions
In summary, we can state:
- The NOBM CT data provide a useful way to observe daily changes in the highly variable North Atlantic region during the NAB, an area where satellite observations may be obscured by cloud cover;
- The NOBM CT data is likely to be in reasonable agreement with SeaWiFS chlorophyll concentration data for this region and time period; and
- The NOBM CT data indicate increasing and decreasing patterns of chlorophyll concentration in the study area, which are not observable in the SeaWiFS 8-day data and difficult to discern in SeaWiFS daily data.
10. Questions for Further Investigation
The paper which provides a full description of the NOBM is:
Assimilation of SeaWiFS Chlorophyll Data into a Three-Dimensional Global Ocean Model
The observational and model data can be examined for various regions and time periods using Giovanni. Choose a region and a time period using the daily NOBM data, and generate an average for a given month. Save the image and also save the data in ASCII format. Then go to the SeaWiFS monthly data, choose the same area and the same month, and perform the same procedure. Compare the images -- where are they alike? Where are they different? Then look at the data for approximately the same values of latitude and longitude in the ASCII output (the latitudes and longitudes won't be exactly the same because the output resolutions are different). How closely do they compare? How could the differences be evaluated statistically?