Binning and mapping

My favorite cartographer is John M. Nelson. In fact, he's the one who actually got me searching what 'cartography' really is. Fortunately, he's a mix of a storyteller/technical support analyst/designer. So, his techniques are the ones I have least trouble understanding. And this is by no means a comment meant to offend because really, I'm a little slow and John is a very 'generous' teacher when it comes to explaining things; even through replies in posts. You can witness his work first hand at his own blog posts here; https://adventuresinmapping.com/.

So, the first of his work that captured my attention is the Six Month Drought of the American Southeast map created using the binning method. I didn't even know what binning is, but the map was so pretty it had me announcing my loyalty to #cartography hashtags. 

So what is binning? According to GIS Lounge, binning is a data modification technique where original data values is converted into a range of small intervals called bins. Bins will then be replaced with a values that is representative of that interval to reduce the number of data points. 

Okay. It should be a no-brainer. But the data he used was the polygon shapefiles of droughts' extent and their severity. Although it is still unknown to me how USGS actually collect this data but his map is sang the deserving anthem to their hard work. But alas, I never had the chance to reproduce it. I do not have the knack of identifying interesting data of any sort, so I either am stuck with reproducing a redundant work or waste my time in a wild goose chase for data; I'm a noob with a tunnel-vision focus. I won't even vote myself if we have a jungle excursion that requires mapping cause we'll be stuck longer than necessary. 

Even so, one year later, precisely this moment...I found a valid reason to attempt this. And it's all because I need to validate satellite imagery classification some colleagues made to show hot spots of global deforestation. I am not a remote sensing wizard, but vector data...now that I can work with. 

Using the same binning technique, I can summarize the steps as follows:

Merge all the data of deforestation variables 

Generate hexagonal tessellation 

Create the hexagon centroids 

Use 'Spatial Join' to sum up the weights of overlapping polygon features of the merged data and join it with the hexagonal centroids  

Then configure symbology 

Visualizing was a herculean effort for my brain. The map John made is a bivariate map. And compared to his data which has 2 numerical variables to enable that, mine only had one and it is the summation of the ranking weight I ensued on the deforestation variables. He merged all the shapefiles of weeks after weeks of drought severity readings. Me...I just manage this >>>

My first attempt was to just visualize the probability of the deforestation using the centroid point sizes.

That wasn't much of a success because visually, it doesn't actually appeal to my comprehension. It looks good when you zoom in closer because it gives off that newspaper print feel with that basemap. From this whole extent, it's not helpful.

So, after I tried to no avail to make it work with toggling the size and the colors, I found that instead of trying to make it look nice, I better opt on answering the questions posed by my colleague; could you identify the areas of high likeliness of prolonged deforestation? For that purpose, only hexagonal mesh would do the trick. So based on the 10 km sq size of their hexagons that depicts the areas of deforestation based on image classification, I used 'Spatial Join' too again and join the centroids back their predecessor hexagons to carry the binned values. 

Et voila!

The weight summation was of the degree of prolonged deforestation likeliness and the values range all the way to 24. I made 4 intervals which gave a practical visualization. Eight intervals were pushing it and 6 was not pleasant. It could be my color palette choice that made them unappealing but too many intervals will defeat my purpose. 

Yay or nay...I'm not too sure about it. But I do believe that this summarizes the areas where conservationists should be on the alert with. 

After having a discussion with a colleague, yeah...this technique has a lot of gaps. 

ONE; this is not a point feature. Using the values where the centroid touches/overlays ONLY is not exactly a precise method. Although, it is not wrong either.

TWO; The merged polygonal data came off as OVERLAPPING polygonal features. 

Overlooking the shortcomings and just using it to visually aid cross-checking...yea maybe. Even then...it's not as laser-point precise as one would aspire. I stand humbled.