The leaves of the ginkgo resemble a small green fan. It's also, to a paleoclimatologist, a record of the air's condition when the leaf first grew. In 2017-18, that record was put to use as part of a citizen-science project run by the Smithsonian's National Museum of Natural History. Some 2,500 volunteers counted tiny pores on ginkgo leaves and made about 24,000 classifications and the data is now making a comeback in coverage this week.

The significance of the tree is due to its age. The ginkgo hasn't evolved much in some 270 million years, so leaves found in the fossil beds are very similar to those of today. Stomata (pores) are present on each leaf that allow the plant to absorb carbon dioxide and release water vapour and oxygen. A plant has to strike a balance between the two. If pores are open, the leaf allows more CO2 in but water is escaping too, hence the leaf responds. The tree will grow with fewer pores when the CO2 level in the atmosphere is high. If it is low, it will grow more.

This is converted into a number, the stomatal index, which is the percentage of stomatal cells on the surface of the leaf. Compare it with living leaf fossils grown at known CO2 levels and measure a fossil leaf—you can estimate what the atmosphere was like when the fossil was alive. Plant proxies like this fill in the deeper past since ice cores only go back ~800,000 years.

The difficulty is the counting. It is slow, is conducted cell-by-cell under a microscope on standardised patches of leaf. There are too many images to display to show the variability of the index from one tree to another, one place to another, and season to season. This variation is precisely what the method needs to take into account if it's to be believed.

The Smithsonian's solution: make the images available online via Zooniverse and have the public count them. The official test phase started on 20 December 2017. The early results indicated that the accuracy of the volunteers varied, as you would expect with individuals following instructions that were written by someone other than the individual performing the task.

The team then added an "unsure" option, which meant that people didn't have to guess, and constructed a training set, to allow new members to practise the system before they touched actual data. Despite the lack of consistency in the initial accuracy, the survival of the pooled counts was helpful, and the researchers noted that the crowd facilitated them overcoming the "bottleneck.

The other half of the project was the volunteers. The other half was in the grounds of the Smithsonian Environmental Research Center in Maryland where 15 mature ginkgo trees were maintained under tented chambers with controlled levels of CO2, up to 1000 parts per million. This is approximately two and one-half times the current rate. The trees that are exposed demonstrate the leaf reactions outside the range of currently available calibration data. The stomatal index/CO2 relationship is poorly constrained above 400 ppm, which is a weakness as we are getting closer to that threshold.

The science also has a warning. The Smithsonian team's work on cell size revealed that the size of ginkgo cells has not been consistent since the early Cenozoic. Pore size and density vary and do not stay in step, which means that fossil leaves are not perfect data substitutes for living leaves, and models that do not consider this can overestimate or underestimate ancient CO2. While the living ginkgo is an important guide, it should not be viewed as an unchanging witness.

Marry it all together, the project is a mini case study about the organisation of research. In many fields, from astronomy to ecology, data collection has become the bottleneck, and it is a good idea to distribute the job among many people and to incorporate the quality assurance mechanisms that make the data reliable. Volunteers provide scale and variety of local samples, as leaves were sent from all over the world, from Singapore to Colorado. The scientists provide the design, training and quality control. Neither can do it alone, and there is no better combination than it is with the same team at the same job that's paid.

A policy dimension is apparent and present. Climate models are now being tested against the behaviour of the planet in the past when governments and companies are making long-term decisions on carbon pricing, emissions targets and infrastructure spending. This kind of evidence is one of many used to create the ancient history of CO2 and temperature, and one of the evidences of the sensitivity of the climate to greenhouse gases. When the proxy is tightened, the science on which the rest of the argument is based is tightened. Fast-growing economies like India, where there is a large pool of scientific labour but research funds are limited, could emulate the model by engaging volunteers for specific jobs with proper training and error checking.