Suggestions for modifications of the landing page

[melisandeteng]

Landing page

• Shouldn't the logo at the top left be the colored one of the CCAI project ?

• The font of the text should be bigger

• Presentation of the project : It is not quite clear from the description sentence of the website what it actually does and what we should expect. What about a more precise injunction, something like "Type a location and see what its neighborhood will look like under climate change in 2050" ? - this is not the best formulation as we still want to be careful and warn about the model confidence -
  When reading "This website shows you how climate change could affect our daily lives." I expect to read facts about crops being destroyed, shortage of natural resources, ... it's hard to make a direct link between an image that looks like your neighborhood under heavy rain, and what will change in your daily life but that's my personal opinion.

• It should maybe be specified that we are only looking at floods for the moment.

Once a user has searched an address:

• "low risk of flooding" should be defined somewhere
• the level of precipitations in cm is not informative per se, as it is unlikely that users will all have an idea of what the "normal" precipitation level is --> put relative change in percentage rather than centimeters

Other things to think about:

• what if the address that was queried is in a "safe" area? What call to action should there be ?

[melisandeteng]

Conveying information on the "risk of flooding" efficiently

The challenge is to deliver a message with information on the risk of flooding:

• without scaring people
• making it clear that there is no such thing as 100% confidence on predictions
• and still compelling to them, even if the queried address is in a "low risk" region

How can risk of flooding be presented? Here are a few suggestions:

• Absolute values : low/high risk of flooding can be defined in absolute values (increase in precipitations in cm or in terms of a defined threshold (X% of increase))
  Now, the challenge would be to define what that threshold is. Do we have climate science inputs on what a sustainable increase in precipitation levels would be according to the geographic region ?
  --> + : interesting to have a "global" standard threshold that is backed by research with some authority. We can have a compelling message like : "this is X% higher than what the sustainable configuration is (define sustainable)" or we can convert X% in terms of "this X% increase is equivalent to Y months of rain currently" (better phrased)
  If there is low risk of flooding, we can still have a call to action by pointing to a nearby region
  --> - : How can we get this "absolute" criterion ?
• Relative value/ranking : compare the risk of flooding to other regions and say something like : "this place is in the X% regions of Canada (or a smaller scale e.g. Quebec) with highest increase in precipitations "
  --> + : "easier" not to mislead people. We do not necessarily have to define low risk of flooding and can talk about high risk of flooding and precipitation increase instead.
  More precisely, If risk is low, we can still deliver a message that is compelling AND a statement (no such thing as "manipulation" here): "This place is in the X% places in Quebec with lowest increase in precipitations. But there will be an increase in precipitations by Y% which corresponds to Z months of rain currently".
  --> - : what scale are we working at ? If we say numbers at a small scale, it feels closer to home so we might have a more compelling message but "relative" value is bad if we are ANYWAY in a region that will be completely flooded / more generally in a region with homogeous increase in precipitations

Other thoughts : we might want to customize the message depending on which type of region we are in. The impact of X% more precipitations might be different in a mountain region, or near a shore.
What data do we have on the impacts (general) of increase in water levels ?

Another point is that we want to be careful and be transparent about the confidence we have in our model. Therefore, we might want to put an emphasis on words like "risk", "the predictions", "confidence of the model" (this is probably a bad one). But at the same time, if we put something out it will be because we TRUST our model's predictions, so we would need to put a note on that.

[gcosne]

General comments on user experience:

• When we pin a place on the map we would like the picture to load directly
• The inference time is too long even for a demo (~22 seconds)
• The Slide action to display the flood should be from left to right
• We could improve the watermark (It's too blurry) painting it after the resize
• Google Logo should not be flooded ( crop -> flood -> add logo)
• Round prediction from the flood model (up to 3 digits)
• The geocoder return the closest location where there is a picture but doesn't move the pin accordingly.