This repository consists of materials that I produce on geospatial data. The works are linked to their source code.
Lake Suğla is a karst lake located in the southeast of Konya, Turkey. In dry seasons, the area of the lake shrinks; and the alluvial lake bed emerges as fertile land for agriculture. The lake was fed with the waters coming from Lake Beyşehir, which I was born. Since there was a change in the water transfer, Lake Suğla became drier and shrunk from 16k ha to 2.5k ha. Though the renewal plan in 1999 enlarged the size to 4k ha, the lake floods during the rainy years to get back to what it belongs. The eye of the lake wants you to hear its story with wet eyes.
| Konya Suğla |
|---|
 |
We like to simplify the world around us. Nature simply does not carry such a goal. When we look at a usual RGB satellite image, we unconsciously start to label structures such as mountains, agricultural fields, cities, etc on the image. We do not stop there; we immediately think if we can calculate the areas of these structures. We never stop there either; we want more specificity: what are the classes of agricultural fields? How many of them are cotton? Can we do it automatically?
You need to think a lot to find a class in your mind from the imagery. You are always looking at an RGB image with many assumptions you fill in your mind in your lifetime. The cotton fields do not label themselves as cotton fields, or trees do not know what you are thinking about them. You can only unbiasedly observe that these structures all vary in time and space, some are more, and some are less. Hint: If you can understand how much variance is going on within the class you label, you'll have fair classification results. The images below use different times’ NDVI values to show you the variance in time. Similar colors seem to act similarly, but still, be careful to assume their class.
| Manisa Turgutlu | Aydın Söke | Şanlıurfa Ceylanpınar |
|---|---|---|
 |
 |
 |
| Adana Ceyhan |
|---|
 |