👉 Our findings highlight how hard it is to separate the effects of photoperiod from those of a thermal optimum cue.
Much remains to be done to unravel the complexity of phenological cues, and understand the implications for forecasting plant responses to climate change!

💡 As we only used temperature data (and not photoperiod), this could suggest that plants rely on plastic thermal cues rather than a fixed photoperiod switch, such as the solstice.

🔎 But when we zoomed-in, we found significant local variation. Some sites had an optimum >20 days before sites, while other sites, especially in Scandinavia, had an optimum >20 days after solstice.

🌍 Using climate data from the past, present, and future, we found that the summer solstice tends to align with a thermal optimum during the growing season.
The period around the solstice represents on average a good trade-off between environmental predictability and growth potential.

💡 Takeaway? Don’t blindly trust good model fit! Always look into the biological processes, and use inverse calibration carefully. This is critical if we want to preserve the value of process-based models over less mechanistic approaches.

Inverse calibration with distribution data should be applied selectively, to calibrate parameters where we lack other sources of data.

⚠️ Compensation between processes leads to non-identifiability: different parameter sets can produce similar species distributions, but with different internal dynamics.

Inverse calibration can yield good predictions of species ranges—even under past climates. But it might do so for the wrong reasons.

💡 Takeaway? Don’t blindly trust good model fit! Always look into the biological processes, and use inverse calibration carefully. This is critical if we want to preserve the value of process-based models over less mechanistic approaches.

⚠️ Compensation between processes leads to non-identifiability: different parameter sets can produce similar species distributions, but with different internal dynamics.

Inverse calibration can yield good predictions of species ranges—even under past climates. But it might do so for the wrong reasons.

(Thanks to all the collaborators, including @fredsaltre.bsky.social !)

💡 The challenge ahead: finding the right balance between representing complex biological processes, diverse data sources to support them, and robust inference methods!

⚠️ That said, more complexity isn't always better. More processes involve additional parameters, which are usually harder to calibrate. And we still often rely on a single parameter set per species — neglecting to quantify and propagate uncertainty when making projections