Environmental Science, visualization

For self-study under COVID-19, we publish an education/research blog for environmental science. The blog premise is: there is a discussion with professors such as using e-mail; and teaching materials for self-study. So, there is less description in the blog, but many images are posted. They are processed by 7 imaging-parameters, and important patterns are emphasized. We explain meanings of the parameters.
The blog has been open to public since July 14, 2020. On Feb. 28, 2021, there are over 250 articles. They are about weather, dust, sea surface anomalies, and volcanic conditions on the Earth’s hemisphere that can be detected by a geostationary satellite at 140.7 degrees east longitude.

1. Introduction
This is a report of liberal arts under the COVID-19 pandemic. Nowadays, it is difficult to lecture liberal arts subjects in a large classroom. There is less choice but to have students learned by themselves in internet environments and submit their reports. We can support students’ self-studies; under such the view point, we propose followings.

2. Bases for learning
The environmental education requires assemble of many facts about meteorology, statistics, geology, information science, etc. It’s a comprehensive science. Students must know a lot of facts far from that of mass communications means. Important facts would be hidden in many kinds of databases (DB) for the environmental science. The students have to learn how to use/analyze them. Since there are various analysis approaches, we believe it is better for us to actually analyze and exemplify environmental events using the DB. Here; we assume the following DB [1~10].

3. Our approach
We adopt a blog [11] to make the examples known to students (and graduated peoples). Blog can update itself at any time, and colors are also available. It is not appropriate for long sentences. “twitter” is also unusable in this respect; and “note” is insufficient access from search engines. Since we separate complex facts and the explanations over many articles, the defect can be reduced.
We are writing in English for attending international students and Japanese living in overseas. Characters are garbled and cannot be read from overseas unless the Japanese environment is installed. Since elementary English is used, sentences can be converted to many other languages by using translation services implemented in browsers [14]. Thus; even native Japanese grown in Japan can recognize the articles.

4. Blog
Contents of the articles are facts after 2020, which are detected from a satellite Himawari-8 that is geostationary one on latitude 140.7 deg, height 35.8km. The observing wave length is from 0.47 to 13.3 μm. Sampling interval is 10 or 2.5 [min].
Thus, time series events can be traced. Moreover, by using image processing, much invisible phenomena are detected. Followings are about already published articles. We use Japan Standard Time (JST), world time +9h; Temperature unit is Celsius. We check Himawari-8 images every day with the following areas in mind.

4.1 China
1) Status of land waters,
2) Sea surface pollution along the coast (Balhae, east China sea),
3) Haze distribution, snowfield dirt distribution; yellow sand,
4) Cloud behavior during high-concentration haze; as one interesting phenomenon, iridescent cloud phenomenon,
5) Wildfires in the southern and remote areas, that would be impacted on global warming,
6) Rains in the Taklamakan and inner Mongolia desert, vegetation time-series changes.

4.2 Russia
1) Siberian large scale forest fire, soot flow to the Arctic. 

4.3 Australia
1) Black band on the Antarctic convergence line,
2) Eutrophication of the Timor Sea and northern Australia,
3) Large scale red dust flow in western Australia,
4) Northeastern wildfire,
5) The blue tide vortex of the Great Australian Bight,
6) Blue tide distribution in New Zealand waters,
7) Large forest fire in the southeast.
Forest fires and wildfires can be seen in various places on satellite images; but they are not so problematic at COP25 (2019.12). we think so strange.

4.4 India
1) Haze distribution in the east.

4.5 Volcanoes
1) Ogasawara, Nishinoshima,
2) Hawaii island, Kilauea volcano,
3) Luzon island, Taal volcano*.
Where “*” is an event happened in 2019. Volcanic eruptions are related to global climate change; thus, it cannot be overlooked.

4.6 Typhoon
We discuss it in Section 6.

4.7 Current status of the solar spherical on X-ray
The status of solar activity is related to global climate change. 

4.8 Current status of the COVID-19
COVID has nothing to do with earth science; but, it has to do with how teachers conduct classes. This is the first reason mentioned in the blog. The second reason is: considering the global environment including phenomena of human society, the research belongs to a new academic field. Mathematical quantification of social phenomena requires a kind of scientific and objective index. As a response to the index, we wish to detect hidden characteristics of human society.
We consider blogs not as thesis, but as a form of preliminary discussions. Using statistical orthodox approaches, under uncertain information, i.e., daily extreme numerical fluctuations, we are trying to detect the spread phenomenon of SARS-CoV-2 virus and effects in many kinds of human society. 

5. Parameterization of image processing
Himawari-8 is a satellite for the meteorological purpose; thus, clouds are detected clearly. But, under the strong clouds images, small other phenomena may be hidden. We wish to detect the hidden images. To detect many kinds of phenomena, we emphasize the originals about brightness, color, timing.
For reproducibility of the emphasis operations, operations are quantified and expressed as parameters.

5.1 Brightness
The brightness is increased or decreased by multiplying each pixel value and a positive real number given by the parameter-1. A pixel values is integer in most information engineering, which is defined by the image format. That’s not enough for physically meaningful processing. We represent 1-pixel value as a 32-bit real number corresponding to 3 primary colors of RGB. 

5.2 Color
We believe that any object has color components. White clouds have a white (R = G = B) part and difference parts from it. The difference parts are considered as color signals. 1 pixel is expressed by {Y(r),Y(g),Y(b)}.
W=min{Y(r),Y(g),Y(b)};
dY(r)=Y(r)-W, dY(g)=Y(g)-W, dY(b)=Y(b)-W;
Y’(r)=Q*W+dY(r), Y’(g)=Q*W+dY(g), Y’(b)=Q*W+dY(b).
Therefore; by the Q (parameter-2), color components of {Y’(r),Y’(g),Y’(b)} is increased/decreased from that of {Y(r),Y(g),Y(b)}.

5.3 Suppression of bluish haze
The satellite image has bluish coloring. Objects having strong brightness like clouds can be identified as they are.
However, patterns of ground surfaces, polluted water surfaces, volcanic plumes, and forest/wild fires are weakened by the bluish haze, and difficult to identify them.
The sunlight gives a dependent scattered light X (λ) that is proportional to the nth power of wavelength (λ) by haze particles. X (λ) = λ-n, n = 0 ~ 4.
The value of nth depends on the diameter of particles.
Now, let the typical wavelength of RGB be λr, λg, λb.
The brightness Y of a pixel is Y =λr-n + λg-n + λb-n.
In case of human’s eye, “λr=0.63, λg=0.55, λb=0.45 [μm]” is a standard. Thus; we multiplied 2 coefficients to GB components of each pixel. The coefficients are,
Kg=0.55-n/0.63-n, and Kb=0.45-n/0.63-n. The “n” is a parameter. Here; we write “amp=parameter-1, Q, n” as a set.

5.4 Bias
The brightness value of each pixel of an original image is depended with the position of sun. The gradient is remarkable at sunrise and sunset. Increasing the parameter-1 in Section 5.1 increases the gradient becomes noticeable. Here, the size of images is defined as 1366 * 768 pixels. A linear slope is sufficient for this size. Therefore, we introduce 3 parameters, bi, bE, and bN. They are real numbers. “bi” operates all pixels, “bE” is, Δx=bE*(1-X/1366), 0<X(integer)<1367.
“bN” is, Δy=bN*(1-Y/768), 0<Y(integer)<769.
X and Y are the number of pixels (natural number) from the left and top edges. Thus, we get a final brightness;
Y’(x,y)=Y(x,y)+Δx+Δy.
The Y’() is sames for RGB-components, and it can exceed the 8-bit range. In that case, following threshold is used.
IF(Y’()<0)Y’()=0, and IF(Y’()>255)Y’()=255.
Since Y’() is real number, finally we get,
Y”(x,y,R)=int{Y’(x,y,R)+0.5},
Y”(x,y,G)=int{Y’(x,y,G)+0.5}, ...
Those parameters are gathered, and it are written as “bi=bi,bE,bN”.

5.5 Clouds
Since meteorological satellites observe clouds, the brightness of clouds is high. We exclude the bright clouds, when we want to get information on sea/ground surfaces than that of clouds. We use a threshold “C”, and cut-off processing is done;
Y(x,y,average)={Y(x,y,R)+Y(x,y,G)+Y(x,y,B)}/3,
IF(Y(x,y,avarage)>C){Y(x,y,R)=0,Y(x,y,G)=…}.
At the same time, the “C” is defined a parameter, when “C>256” is inputted, non cut-off processing is executed. The process of "Y () = 0" means that nothing is done at the pixel position (x, y).
When there are multiple images in time series {Y (..., t0), Y (..., t1), ..}, and the clouds move, the cutoff process may be able to detect patterns below the clouds.
For the possibility, a counter is attached to each pixel to record how many images are accumulated. After accumulating them, the pixel value is divided by the counter value, and we obtain the averaged brightness as one accumulated image. In this way, clouds can be erased to some extent.

The parameters in Sections 5.1-5.5 are related each other. Pay particular attention to 3 parameters of "bi=...". Please learn the feeling while actually operating them. 

5.6 Moving objects
We need to Identify moving / non-moving patterns within a defined time interval. This is required to predict property of the detected pattern. We express 2 images as Y0 and Y1, which obtained at timing (t0, t1). Δt=t1-t0.
If the Δt is negligible, since the difference of brightness between the two images is less, “ΔY=abs(Y0-Y1)” indicates moving patterns.
When the ΔY image is calculated actually, it is ΔY ~ 0 at Δt = 10min. By using “amp=3,1,0, bi=110,0,0”, the 0 value pixels are raised closer to the median value 127, and average brightness is amplified 3 times. We save the processed image in png-format. It is desirable to enhance the contrast of the png-image with a commercial available image editor. It is so pale images.
The images have edges of moving clouds. The edge-normal indicates the moving direction of clouds. Clouds do not have a clear outline in a storm. In that case, no edge lines are visible and the clouds become a dark, dispersed area image. Using character of this image, we can detect wind direction of typhoon paddlewheels and storms near the eyes. We call ΔY “abs(sub)” image. The plain accumulation image is “PY=(Y0+Y1)/2”. “FY=PY-ΔY” corresponds with non-moving patterns. We call FY “alternative accumulation” image. Many kinds of phenomenon is understood by using FY and abs(sub) images.