(Page is in very early development)
The intended audience of this page
This page is aimed at diverse audiences. One is members of national meteorological services, from forecasters to senior administrators. These individuals can develop satellite imagery-based climatologies to help in their forecasting activities.
Then there are educators who may want to know more about their country’s climate. And any members of the public who might be interested in the climate of their country.
Finally, an important audience is the biogeography and conservation community that needs to know more about where organisms live and what areas need the highest protection.
This page describes procedures anyone with internet access can use to develop satellite imagery-based cloudiness climatologies of high spatial resolution. Here I describe the procedures I have recently applied to both GOES imagery and to MODIS imagery from sun synchronous polar orbiting satellites. Anyone with an Apple computer and Graphic Converter software can reproduce what I will show here. (Comparable image manipulation software for Windows computers exists to do the same functions.) No programming knowledge is even required. Much more can be done with programming skills and suitable manipulation of the imagery, but the basic aspects of developing cloud climatologies don’t require it.
Philosophy of downloading imagery. What is the objective?
What is purpose of developing cloud climatologies? What do they offer that rainfall climatologies do not? What are possible applications of such climatologies. Here we discuss why we should develop such climatologies.
Examples of the applications of cloud climatologies include:
1) identifying the least cloudy regions for solar power plants, airports and industrial activities that benefit from lack of clouds
2) identifying the location of cloud forests for possible conservation consideration
3) identifying the least cloudy regions for astronomical observatories (though nighttime cloudiness is more important for this application)
4) Identifying the least cloudy coastal locations for possible tourism activities.
For weather forecasters there are additional applications that might not be relevant to most users of cloud climatologies. The satellite imagery can be stratified by synoptic situation (essentially larger-spatial scale conditions) such as anomalously strong or weak trade wind conditions, anomalously cold or warm days, or the presence of anomalous vertical wind shear. There are essentially an infinite variety of meteorological conditions by which the satellite imagery can be stratified. Of course, the more stratifications there are of a given data set, the fewer the number of observations that will comprise the resultant average and the less the signal-to-noise of the final results. These so-called “synoptic climatologies” have been prepared for various weather phenomena and form the basis of conceptual models of atmospheric processes. Satellite imagery has been used to develop such models, but usually not with very high spatial resolution.
Types of satellite imagery available
Two types of satellites are useful for developing cloud climatologies: polar orbiting and geostationary. Many satellites provide this type of imagery but the availability of such imagery is another matter. Here I will use GOES imagery that covers most of the Western Hemisphere, as these satellites are managed by NOAA and the data is in the public domain. Probably the most convenient polar orbiting imagery is that from the NASA Terra and Aqua satellites that are still providing data since their launch almost 20 years ago. There are now a host of other such satellites but there are some problems with mixing imagery from different satellites.
Sources of GOES imagery
There are many websites that display GOES imagery. Most are very good for what they do. Most applications of GOES imagery are for operational weather forecasting objectives. These are not primarily research satellites. Unfortunately, most GOES imagery websites do not display imagery from much of the western Hemisphere – they tend to focus on the US. If they do cover the entire full disk perspective they do so at reduced resolution. A full resolution full disk GOES image (500m pixel at nadir) is about 20,000 pixels across!
Some websites provide full resolution imagery but the download speed or display capabilities are not suitable for rapidly downloading the many images needed for developing climatologies. Others websites have little archived data – images must be downloaded as they arrive.
The website used to download GOES imagery here is that from the NASA Marshall Spaceflight Center. Up to 50 images are archived for display, with newer imagery replacing older imagery. For 10 minute time resolution this amounts to a time window of just over 8 hours. Thus imagery must be downloaded twice daily to make sure no imagery is missed. The actual site has options for mage quality (jpg) and size of the domain and the specific location of the sector you wish to download. You select your domain, size of the domain in pixels (maximum unfortunately is 1400 by 1000 pixels), and the quality of the jpg. Also, you select whether you want a map or no map superimposed on the image.
The main problem with the GOES data download procedure sketched out above is that it is time consuming. Even with a fast download speed the time it takes to change the date, hit return and copy the image adds up. One sector (1400×1000 pixels) of 50 images (an 8hr 20min period with 10min interval imagery) takes, in my experience, about 2 minutes to download, depending on the internet speed at the time. This is not a problem – except that it must be done twice daily to capture all visible imagery from sunrise to sunset, and if you want to cover a large geographical domain many sectors must be downloaded. For example, if you only want imagery that covers the land area of Costa Rica a single sector suffices, but perhaps 10 sectors are required for larger countries like Mexico or Peru. Although this is quite manageable for a weather service or a research group with undergraduate students, but it is obviously better if such routine downloading could be automated, centralized and the products distributed via a website to any potential user. Better yet is to find a GOES imagery archive with full resolution imagery and automate the access and downloading of the imagery. This is suitable for a research group – but probably not a small forecast office or those with other limitations. Of course, finding a suitable online imagery archive is not easy – if one exists.
Despite the potentially tedious nature of developing a cloud climatology by downloading the individual images there are a few benefits of doing it at your own office, at least for smaller domains (for example a handful of sector downloads). Firstly, you actually see the imagery and develop a sense of the “synoptic climatology” – the phenomena that are important on a daily or weekly basis in modulating the cloudiness. Some days the forecaster will want to archive as good examples of a particular phenomenon. Secondly, you will become very aware of the diurnal cycle of cloudiness over your domain of interest. Any forecaster should have a solid appreciation of both the synoptic variability and diurnal variability of cloudiness over their area of interest. Researchers should also appreciate these variations – they are also likely to identify problems with the data set. For example, the imagery from the NASA site mentioned above is very reliable but their ground station has had power outages or other problems. Occasionally data is missing for certain periods – though usually only one or two image times.
Source for MODIS imagery from Terra and Aqua satellites
The most convenient source of MODIS imagery is probably the NASA Worldview snapshots site. This site allows one to see the entire imagery data set and select a subset of the imagery at a suitable spatial resolution. Like with the GOES imagery, a domain can be saved and used to sequentially download imagery for the same domain so that a climatology can be generated. The procedures are slightly different but yield the same results.
My motivation for working with satellite imagery in recent years is to relate vegetation to mean cloudiness. I have see very cloudy, and very cloud-free regions around the world and realize that the vegetation in these areas depends strongly on the mean cloudiness.
Extracting clouds from visible imagery
The procedures used to generate the cloud climatologies here have been described elsewhere, but I will briefly describe them here. An 8-bit grayscale visible image has 255 brightness levels, from 0 (pure black) to 255 (pure white). If a highly reflective surface is considered to be a cloud then clouds can be detected by applying a threshold to select pixels brighter than the specified threshold. The example below shows how applying different brightness thresholds yields different cloud detection results.
Of course a brightness threshold technique cannot distinguish clouds from salt flats, glaciers or bright sand or snow. For this reason the technique should be restricted to areas where snow is infrequent. Most areas in the mid-latitudes in the warm season satisfy this requirement and nearly all of the tropics and subtropics meet this criterion. Only on very high mountains in the tropics is ice present and these locations are well-known. The same applies to highly reflective salt lakes and sand dunes. These latter areas also tend to be minima in cloudiness in any case.
The threshold technique works very well over water bodies like the ocean since the background brightness is very low except under calm water conditions. Under certain conditions specular reflections occur that are very bright. However these vary with latitude throughout the year and require relatively calm conditions more often found on smaller lakes and rivers than on the open ocean (though this certainly occurs). While important on a given day, specular reflections don’t affect the mean cloudiness patterns to an appreciable degree.
The main challenge to the threshold technique is determining the best threshold to use to identify clouds. It very much depends on your applications. If you want to identify thin cirrus clouds that barely reflect sunlight you will need a relatively low threshold value. However, such a value will also detect the background over many land areas. Over the ocean it will work better. But if your objective is to detect only thick clouds that are highly reflective then a higher threshold is appropriate.
Cloudiness versus precipitation
Clouds cannot be directly related to precipitation arriving at the earth’s surface. Stratus is usually non-precipitating but can be very persistent along many subtropical coastlines. High cirrus, composed of ice crystals, may be precipitating but the precipitation will usually not reach the ground. However clouds are important on the underlying surface. The figure below shows the major impacts of an idealized stratus cloud near a coastline.
Islands and their mean cloudiness
Islands have an important role in the theory of biogeography. “Island biogeography” is the subject of how islands are colonized by new species and how speciation occurs on islands. Eventually there is a balance between new colonizing species and the available niches on an island. These niches are occupied both by speciation of organisms already on the island and those that are arriving. The number of niches is thought to be related to the diversity of habitats on an island. The larger and more topographically diverse the island the greater the diversity of organisms – provided comparable island ages. But the topographic diversity of an island affects the climatic diversity of the island. So it may be that the climatic diversity of an island is as important as the elevational or topographic diversity of an island. Here I am trying to show the climatic diversity of many important islands that lack dense climatic data via a partial surrogate of cloudiness.
Diurnal changes and understanding mean cloudiness patterns
Climatologies based on GOES imagery
The GOES imagery has the major advantage of being very frequent. Full hemispheric coverage is available every 10 minutes with the current GOES satellites. Thus, approximately 72 images (depending on latitude and time of year) are available for averaging. But there are complications. Near sunrise or sunset there are long shadows cast by clouds that can reduce the reliability of extracting cloudy pixels from visible imagery.
We present here examples of “short-period climatologies” for all of the countries of the western hemisphere tropics and subtropics to show the potential of the technique applied to the operational imagery that is widely available. The following are not meant to be accurate long-period climatologies of cloudiness, but rather examples of what can be accomplished by using only a few months of 10-minute imagery.
Where are short period climatologies unreliable? Over the ocean. Much longer averaging times are needed to obtain a reliable cloud climatology over the ocean because there is no strong focing due to topography and the diurnal cycle of land heating. Although there are topographic gradients of the ocean, they are very small and unimportant (except for oceanographic circulations) and less than 1 m so of no consequence in forcing atmospheric motions. Over flat land areas a longer averaging period is also needed relative to mountainous areas. In very dry/cloud free areas the average cloudiness may be affected by a few events per year, so longer averageing intervals are needed for building a more reliable climatology.
Selecting the domain
If your interest includes cloudiness over the ocean you will naturally want to download imagery over the ocean. But 500m resolution imagery would require many sectors and would not contain much information at that resolution. One can select lower resolution imagery for downloading and cover the required area of interest. For example, a GOES low resolution (2 km pixel) sector from the NASA Marshall GOES site covers much of the Caribbean Sea.