Precision agriculture is a farming management concept based on observing, measuring and responding to inter and intra-field variability in crops.
The goal of precision agriculture research is to define a decision support system for whole farm management with the goal of optimizing returns on inputs while preserving resources.
Precision agriculture also provides farmers with a wealth of information to:
A key element in precision agriculture and forest management is a vigor map (NDVI, etc.) of the plants that provides a quantitative measure of the plants’ health. LRRT provides reliable vigor maps to agronomists and forest managers by collecting multispectral aerial data using Unmanned Aerial Systems (UASs, also called “drones”).
The sunlight is processed by the plants (photosynthesis), and the reflected light bears a clear signature of the plant’s health. The spectral analysis of the light reflected by a healthy plant has clear peaks not only at the Green (550 nm) wavelength, but also at the Red Edge (725 nm) and NIR (Near Infrared, 800 nm) wavelengths.
Laboratorio Rocce e Ricerca Tonon (LRRT) transforms this multispectral data and aerial imagery into actionable data, allowing growers, forest managers and turf managers to detect and treat early-onset issues, resulting in long-term savings, improved crop or turf health, and water savings:
as described in the following documents:
This page is organized as follows:
MULTISPECTRAL AERIAL DATA ACQUISITION
LRRT acquires the reflected light from the plants at 4 narrow bands: Green (550 nm), Red (660 nm), Red Edge (725 nm), and NIR (800 nm). At the same time, we acquire the light input to the plants in that very moment; this is the same as correcting for the light intensity in standard photography by opening or closing the lens aperture. If we do not do this, the picture will either be overexposed or underexposed; likewise, if we do not correct in real time for the sunlight input to the plants, the vigor maps will provide either too high or too low a vigor degree.
The second important aspect is the sharpness of the multispectral sensor. Above is the comparison between the bands acquired by a sharp multispectral sensor such as the one used by LRRT and the bands acquired by a regular camera that has been converted to also capture multispectral data. This is similar to a bad quality radio that cannot be exactly tuned to a specified radio frequency so that a lot of noise and spurious signals are output by the radio. As a result, you cannot understand what the radio is saying or playing; likewise, the vigor map will be unreliable because it is impossible to discern the plant reflected spectra.
The third important aspect is the calibration of each sensor, which LRRT carries out before each flight by using a special standard reflector. This is equivalent to the white balance in standard photography.
Each time our multispectral camera is triggered, 4 spectra are acquired over each pixel on the plant canopy, which is typically 3 x 3 in (8 x 8 cm) in size, i.e. similar in size to a vine leaf.
Typically, we can cover 200 acres (0.31 sq mi) or 80 hectares in a 30 min flight!
The table gives comparative data for satellite multispectral data for agriculture and forest use. The following items are of importance:
Satellites with daily overpass and latency have a spatial resolution of 1 km, which is not enough for precision agriculture and special forest applications. On the other hand, higher resolution satellites such as Landsat 7 only reach a 30 m resolution that is not enough for precision agriculture, even more so when the ground is steep (hilly vineyards, orchards or woods), and plant canopies are close together when seen from above. These higher resolution satellites have an orbit frequency of about 2 weeks; however, if it is cloudy when the satellite visits a site, no information is gathered because no plant reflectance reaches the satellite sensors. Especially for agriculture applications where time is of the essence to make critical decisions (e.g., fertilizing or harvesting), this is a deal breaker!
MULTISPECTRAL DATA ANALYSIS
These are typical simultaneous shots taken by our multispectral sensor, which has 4 special eyes (one per band) and 1 standard eye for the visible (RGB) spectra. On the ground, each picture covers an area about 450 x 350 ft ( 140 x 110 m) in size. Although each picture looks like a black and white (grayscale) photo, the grayscale indicates the reflectance over that particular band (white = max reflectance, black = no reflectance). Notice that vines and trees exposed to the sun are very reflective on the Red Edge and NIR bands (see spectra above) but have no reflectance on the red band.
For each band, a suitable overlap between shots insures that a georeferenced orthophoto may be created. So, we generate 4 orthophotos on the multispectral bands and 1 orthophoto on the visible range. At this point, the agronomist or the forest manager provides LRRT with the desired vigor index, which is a function of the 4 acquired bands. For example, the most common index is the NDVI (Normalized Difference Vegetation Index), which was proposed by Rouse, Haas, Schell and Deering in 1973:
A substantial list of vegetation indexes is available HERE.
Shown above are: (a) an orthophoto in the visible (RGB) range, and (b) the vigor map for the NDVI over the same plot of land. Large NDVI variations are visible along each vine and across vines; these small scale variations make it clear that satellite data (whose best resolution is 30 m) are inapplicable for these purposes.
A typical detail of an NDVI map of a vineyard to appreciate the resolution of the map. Even at this scale, it is evident that large NDVI variations are visible along each vine and across vines.
Move your mouse over the picture to toggle between the NDVI vigor map and the visible range (RGB) orthophoto; the picture represents about a 60 m x 33 m (200 ft x 110 ft) area. Notice the extremely high level of detail and the perfect overlap between the NDVI vigor map and the visible range (RGB) orthophoto. Also notice that low vigor plants may correspond to plant with a smaller canopy, such as in the area marked with an “A” above, whereas in the area marked with a “B” above the plants look similar to one another in the orthophoto but their NDVI varies significantly along a vine line. In precision agriculture it is imperative that the vigor and RGB data be correctly georeferenced, otherwise any use of this data would be incorrect and even counterproductive, especially in fruit-tree and vine applications, where vigor may vary (as in this example) significantly within, say, 3 ft (1 m) both along a vine row and across vine rowes. Inaccurate georeferencing means that a treatment may be applied to the wrong vine row, that a plant protection product may be applied to the healthy vine whereas the needy vine is skipped, or that a diseased plant is not eliminated and the phytoplasm spreads throughout an entire field. Indeed, at this level of detail provided by LRRT, one may recognize which plants (vines) have been affected by diseases, such as:
Grapevine Pinot Gris Virus (GPGV)
This leads us directly to the use of vigor maps.
HOW CAN VIGOR MAPS (INCLUDING NDVI MAPS) BE USED?
The pictures above show the NDVI evidence in two instances taken from the same field: the picture on the left pertains to an area where NDVI = 0.3, and the picture on the right pertains to an area where NDVI = 0.7.
These four pictures refer to four instances taken from the same corn field: it is evident how NDVI directly correlates to the field productivity.
The following are the typical uses of the vigor maps (incl. NDVI maps):
3. Based on a vigor map, agronomists prepare prescription maps for:
These prescription maps may be:
4. Sampling may be optimized by directing it to the most significant areas and by establishing the geographical validity of each sampling.
5. Harvesting may be optimized based on the plant vigor: fruits are processed differently based on the vigor value of their plants (e.g., different wines are made from grapes picked from areas of different vigor). This dramatically increases the production value.
One may proceed in two different ways:
These are typical screenshots taken from “intelligent harvest” applications that are available to the operator as the equipment moves around a field. On the left, the equipment picks the fruit because the area has a low vigor value; on the right, the equipment does not pick any fruit because the area has a high vigor value.
Via Nazionale 206 - 38123 - Trento, Italy