Spring is in the air. And for deciduous trees, spring means putting out new green leaves. Scientists still can’t predict with absolute accuracy the date that little green buds appear on trees, but they’ve figured out some of the things that matter: spring temperatures, amount of daily sunlight, and something called “winter chilling,” which can be defined in a number of ways. Essentially, some species may need a certain amount of coldness before they switch into put-out-leaves mode.
When we study spring green-up at the landscape or regional scale, we need to figure out how all the different trees (and other plants) contribute to overall “greenness.” It could be that the first species to put out leaves in the spring are the most important for the regional greenness signal. Or it could be an average over all species. Or it could be something more complicated.
In Season Spotter, we have a particular question for spring green-up, so we can better determine the progression of spring for sites with deciduous trees. In particular, we want to label two points: the “start of spring,” which is the day the first green leaf buds appear, and the “end of spring,” which is the day when the leaves have all fully expanded.
To do this, we show you two images side-by-side of the same scene. The images are taken either 1, 3, or 7 days apart and can be in either order. (In other words, the image taken first is not always on the left.) We ask you to tell us which image has “more or bigger leaves”. We are essentially asking you, “which of the two images comes second in spring?”
Why do we do this, when we know which image came first? The answer is that we want to see if people can visually see a change in leaf progression between the two days shown. If the leaves haven’t yet started to develop in either image, everyone should say that the two images have the same amount of green leaves. And if the leaves are fully developed in both images, everyone should also say that the two images have the same amount of green leaves. It is only in the in-between area when leaves are developing that people should note a difference.
We chose 7 days difference because we know you will be able to see a difference in leaf development in images taken a week apart. We chose 3 days difference because it is common for on-the-ground observers to take phenology measurements two or three times per week – or every 3 days on average. And we chose 1 day difference because we are not sure if people will be able to see minor changes on a day-to-day basis. By analyzing each of the different time lags we will be able to figure out how camera data compares to on-the-ground observations, and what time lag is most efficient for analyzing camera data.
To analyze your classifications, we will line all the days up for a site. We will look at what day everyone starts saying they see a change in leaves and what day is the last day they see changes in the leaves. That will give us the “start of spring” day and the “end of spring” day. We can compare these dates with those that we get from automated processing, from satellite data, and from on-the-ground observations. And we can start to include these dates in statistical models so we can better understand what factors matter for spring green-up at the landscape and regional scale.
In this made-up example, each green dot is one person’s classification. If a dot is at the bottom, then that person thought the two images had leaves at the same development stage . If a dot is at the top, then that person said the leaves in one image were ahead of the leaves in the paired image. We can use the agreement among classifications for each pair to determine the first and last days of spring.