How SIFcam Works

The fluorescence spectrum emitted by a photosynthesising leaf has two characteristic peaks: a smaller red peak near 685 nm and a broader far-red peak near 740 nm (see figure below). They come from the two protein complexes of the photosynthetic machinery. The red peak is produced almost exclusively by photosystem II, the regulated workhorse where most photoprotection happens. The far-red peak carries contributions from both photosystem II and photosystem I, so it integrates a different and more complete view of canopy photosynthetic activity [1].

At canopy level the two peaks behave very differently. Chlorophyll absorbs strongly between roughly 640 and 700 nm, exactly where the red peak emits, so light emitted from a leaf deep inside the canopy is largely reabsorbed by the leaves above before it can escape. Above 700 nm chlorophyll absorbs very little, so the far-red peak is essentially transparent to canopy structure and reaches an instrument above the canopy more or less intact [2]. As a result, the surviving far-red signal carries information from the entire productive canopy, not only the sunlit top, and has been shown to track gross primary production more closely than the red peak in field studies, particularly in cereal canopies during active growth [3].

Even with the far-red peak as the target, fluorescence is faint, only about 1 to 5 percent of the light leaving a sunlit canopy at the same wavelengths. To separate it from reflected sunlight, an instrument has to look inside a very narrow part of the spectrum where the reflected component is naturally suppressed. The oxygen A-band, a narrow region near 760 nm, provides exactly that window. Oxygen in the atmosphere absorbs incoming sunlight strongly there and leaves a deep, dark feature in the solar spectrum reaching the ground. Against that dark background the plant's emission stands out and becomes measurable [4]. The O₂-A band sits only about 20 nm from the surviving far-red peak, close enough to read it directly. It is the same physical principle, in the same wavelength region, that the European Space Agency's FLEX satellite mission will use to map fluorescence from orbit [1].

SIFcam reports the result in absolute physical units, milliwatts per square metre per nanometre per steradian, rather than as a relative index. Working in physical units is what makes one measurement comparable across flights, sites and seasons, and provides a consistent quantitative basis for downstream analysis, including machine learning.

SIFcam captures a whole scene in a single instant. Rather than building an image up line by line, as a scanning push-broom imaging spectrometer would, it records the full frame at once. The optics are stationary while the frame is taken, with no rotating filter, no scanning mirror, and no moving parts in the imaging path.

That stationary design lets SIFcam acquire calibrated fluorescence frames at rates of up to 10 per second. At that cadence the camera can do more than collect single frames for mapping: it will be able to stream live SIF imagery from a fixed tower, or record short videos that show how photosynthetic activity evolves over a canopy through the day. Every frame still arrives with the position and orientation it was taken at, so the same data can be stitched into a static map after a flight [5].

SIFcam is designed to fly on standard survey drones and can also sit on a fixed tower for continuous monitoring. The retrieval depends on stable sunlight, so it operates in daylight under clear or near-clear skies; acquisitions made under rapidly shifting cloud cover are flagged so they can be treated with appropriate caution.

From a flight altitude of 25 metres, a single snapshot covers only a small patch of ground, roughly 12 by 12 metres. To map a whole field, the drone flies a planned grid pattern and captures hundreds of overlapping images, each one a small piece of the larger scene.

Those pieces are then assembled into one seamless map. Each frame undergoes a series of pre-programmed corrections, including lens distortion, spectral calibration and radiometric normalisation [6].

It is then placed onto a common ground reference using the position and orientation recorded in flight. Where frames overlap, they are aligned to one another and blended so that the joins disappear. The result is an orthomosaic: a single, geometrically corrected, georeferenced image of the whole survey area [7].

Because every frame carries a calibrated fluorescence value, the finished orthomosaic is more than a picture. It is a SIF map in physical units, in which each pixel reports the photosynthetic signal at that point on the ground.

At a flight altitude of 25 metres, a SIFcam orthomosaic resolves detail down to roughly 2.6 centimetres per pixel, fine enough to see variation within a single field or breeding plot.

The orthomosaic is delivered as a standard georeferenced raster, so it drops directly into GIS tools and into farm-management or field-trial software. Plot boundaries can be laid over the map to extract an average SIF value for every plot, turning an image into a table of numbers that is ready for analysis.

Repeating flights over the same area can turn individual maps into a time series, making it possible to observe how fluorescence patterns change over time. Today, SIFcam is most reliable for comparing relative differences within a single acquisition, such as variation between plants or plots under the same illumination conditions. Longer-term comparisons across dates and seasons will become increasingly meaningful as corrections for changing incoming light are further developed.