howsure/index.html at main · GeoCodeCrafter/howsure · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
<!doctype html>
<html lang="en">
  <head>
    <meta charset="utf-8" />
    <meta name="viewport" content="width=device-width, initial-scale=1" />
    <title>howsure — how much do the models actually agree?</title>
    <link rel="stylesheet" href="/src/ui/style.css" />
  </head>
  <body>
    <header>
      <h1>howsure</h1>
      <p class="lede">
        The temperature on your weather app is the middle of dozens of forecasts
        that disagree with each other. This shows you the disagreement.
      </p>

      <div class="controls">
        <label>
          Place
          <select id="place"></select>
        </label>
        <label>
          Model
          <select id="model"></select>
        </label>
      </div>
    </header>

    <main>
      <p id="status" role="status" class="status">Loading…</p>

      <section id="verdict" class="verdict" hidden>
        <p class="headline"><span id="now"></span></p>
        <p id="sentence" class="sentence"></p>
      </section>

      <canvas id="fan" aria-label="Fan chart of the ensemble forecast"></canvas>

      <p class="legend">
        <span class="key key--iqr"></span> half the models
        <span class="key key--band"></span> 8 in 10
        <span class="key key--outer"></span> full range
        <span class="key key--median"></span> median
      </p>

      <section class="notes">
        <h2>What you're looking at</h2>
        <p>
          A forecast model is run many times from slightly different starting
          conditions, because we never know today's weather precisely enough to
          be sure about next Tuesday's. Each run is a <em>member</em>. The number
          you normally see is the middle of them.
        </p>
        <p>
          The fan widens with lead time because small differences today become
          large ones later. That's not a flaw in the model — it's the atmosphere.
          What's a flaw is showing day seven in the same typeface as tomorrow.
        </p>
        <p class="caveat">
          One honest caveat: ensemble spread is <strong>not</strong> the same as
          forecast error, and it's known to be too narrow. Real errors are
          typically larger than the spread suggests, so treat the fan as a floor
          on the uncertainty rather than the whole of it.
        </p>
      </section>
    </main>

    <script type="module" src="/src/ui/main.ts"></script>
  </body>
</html>