Audio Birding

I redeployed my AudioMoth acoustic recorder (3^rd year in a row), continuing to explore my backyard soundscape during peak spring migration. Every morning during migration is different. Some seem almost quiet, while others reveal a chorus of resident birds and the occasional surprise visitor. As in previous years, I programmed the device to record for three hours each morning, from 5:30 to 8:30 am, beginning on May 12 and continuing through June 15. This recording window captures most of the dawn chorus, when the birds are most vocal. If you’d like more background on how I use the AudioMoth, check out my earlier blog post. Below are some of my initial findings from this year’s deployment.

d3 = require("d3@7")
format = d3.format(".01%")

// load data for 2026
bird_freq_26 = await FileAttachment("data/may_june_chorus_2026.csv").csv({typed: true});
bird_freq_26_filt = bird_freq_26.filter((d) => d.freq > 0.001).map(d => { d.year = Number(2026); return d; })

// load data for comparison
bird_pct = await FileAttachment("data/comparison_2026.csv").csv({typed: true});

// OcclusionY adds an initializer that shifts nodes vertically with a tiny force simulation.
occlusionY = ({radius = 6.5, ...options} = {}) => Plot.initializer(options, (data, facets, { y: {value: Y}, text: {value: T} }, {y: sy}, dimensions, context) => {
  for (const index of facets) {
    const unique = new Set();
    const nodes = Array.from(index, (i) => ({
      fx: 0,
      y: sy(Y[i]),
      visible: unique.has(T[i]) // remove duplicate labels
        ? false
        : !!unique.add(T[i]),
      i
    }));
    d3.forceSimulation(nodes.filter((d) => d.visible))
      .force("y", d3.forceY(({y}) => y)) // gravitate towards the original y
      .force("collide", d3.forceCollide().radius(radius)) // collide
      .stop()
      .tick(20);
    for (const { y, node, i, visible } of nodes) Y[i] = !visible ? NaN : y;
  }
  return {data, facets, channels: {y: {value: Y}}};
})

// d3 plot
Plot.plot({
  marginLeft: 145,
  fill: "#021d44",
  x: {axis: "top", percent: true, label: "Frequency"},
  y: {label: null},
  marks: [
    Plot.barX(bird_freq_26_filt, {x: "freq", y: "common_name", fill: "#00467F", sort: {y: "x", reverse: true}}), 
    Plot.text(bird_freq_26_filt, {x: "freq", y: "common_name", text: d => format(d.freq), textAnchor: "start", dx: 3, filter: d => d.freq <= 0.02, fill: "currentColor"}), 
    Plot.text(bird_freq_26_filt, {x: "freq", y: "common_name", text: d => format(d.freq), textAnchor: "end", dx: -3, filter: d => d.freq > 0.02, fill: "white" }),
    Plot.ruleX([0])
  ]
})

Figure 1: Relative frequency of detected bird calls from AudioMoth recordings.

To better understand how the soundscape changed between 2024 and 2026, I focused on species with a noticeable presence (those that made up at least 2% of the relative frequency in all years). This threshold helped to spotlight species that were consistently active or showed meaningful changes, while filtering out those that were incidental or rare.

So, what stood out?

// d3 plot
Plot.plot({
  style: { margin: "0 auto" },
  height: 540,
  marginLeft: 109,
  marginRight: 108,
  marginBottom: 5,
  x: {axis: "top", type: "point", tickFormat: "", label: null},
  y: {axis: null, inset: 20},
  marks: [
    Plot.line(bird_pct, {
      x: "year",
      y: "pct",
      z: "common_name",
      stroke: "common_name"
    }),

    Plot.text(
      bird_pct.filter(d => d.year === d3.min(bird_pct, d => d.year)),
      occlusionY({
        clip: false,
        x: "year",
        y: "pct",
        text: d => `${d.common_name} ${d.pct}%`,
        textAnchor: "end",
        dx: -5,
        fill: "common_name",
        fontSize: 14,
        radius: 8
      })
    ),

    Plot.text(
      bird_pct.filter(d => d.year === d3.max(bird_pct, d => d.year)),
      occlusionY({
        clip: false,
        x: "year",
        y: "pct",
        text: d => `${d.pct}% ${d.common_name}`,
        textAnchor: "start",
        dx: 5,
        fill: "common_name",
        fontSize: 14,
        radius: 8
      })
    )
  ]
})

Figure 2: Changes in the relative frequency of detected bird calls over time.

The biggest surprise this year was the sharp decline in Chipping Sparrows, which are usually a staple in my backyard. Then again, now that I think about it, I haven’t seen nearly as many this year. Maybe the breeding pair that typically nests in my backyard decided to call somewhere else home.

A couple of other things stood out to me. Some calls were classified as Fish Crow, which is certainly possible. However, it could also be the result of an update to BirdNET (the classifier I use to identify birds by their calls), as previous recordings were almost always identified as American Crow. Another interesting pattern was the relative increase in Blackpoll Warbler detections, which was a pleasant surprise.