Satellite-based remote sensing offers a useful source of information regarding the variability of vegetation and soil on individual fields. Earth observation satellites carry various sensors, which collect information about the earth on a continuous basis. This information has potential value to producers who want to consider the variable productivity of their fields in their management strategy.
An increasing number of earth observation satellites are in operation. This has created a wider choice of information sources along with a significant decline in the price of satellite images. As a result, a growing number of companies are now offering satellite data products to the agriculture industry. The trend currently points towards increased use of remote sensing for crop management. The main hurdle is bridging the knowledge gap between the technology and the information needed at the farmgate for its profitable application.
When you consider satellite data for any application, there are three key parameters to evaluate.
Spatial resolution refers to the smallest size object that the sensor can resolve on the earth's surface. A Landsat-5 TM image has 30 meter resolution and cannot resolve any object smaller than 30 x 30 meters in size. The panchromatic mode of Landsat-7 ETM sensor, launched in 1999, can resolve objects down to 15 x 15 meters in size. An IRS panchromatic image can resolve objects down to 5.8 x 5.8 meters in size. An IKONOS panchromatic image offers spatial resolution of 1 meter, which is comparable to the resolution of air photos. The spatial resolution of all these sensors is sufficient to map the variability of soil or vegetation on an individual field.
Satellite sensor spectral properties determine exactly what type of electromagnetic energy the sensor can measure. The sensors mentioned above all measure radiation intensity in the optical and near infrared part of the electromagnetic spectrum. In order to provide an image of the earth's surface, the atmosphere must be cloud-free; otherwise the sensors just see the top of the clouds. Cloud cover is the main problem that adversely affects these sensors.
Landsat TM measures reflected radiation in 7 different wavelength bands ranging from visible wavelengths to thermal infra-red wavelengths. SPOT can collect either 3 different bands (visible, near infra-red) or panchromatic (grayscale) images in the visible/near infra-red range. In the near future, there are launches planned for hyperspectral sensors which can measure reflectance at dozens of small wavelengths through the visible and near infra-red region. These sensors will provide very detailed spectral "fingerprints" of the earth's surface.
The temporal resolution of a satellite is its repeat cycle or how often the sensor passes over the same area. The IRS satellite repeat cycle is 24 days. For the Landsat satellites, the repeat cycle is 16 days. For SPOT the repeat cycle is 26 days, however, the sensor is "programmable", so it can be pointed off to the side and can image the same area on successive days. The IKONOS satellite is also programmable and has a 3 day repeat cycle. The more frequently the sensor can scan an area, the greater the possibility of acquiring a cloud-free image. More frequent repeat coverage is a definite advantage.
The value of remote sensing data depends on our understanding of how electromagnetic radiation interacts with a surface. Green vegetation has a unique reflectance pattern of visible and near infra-red light compared to soil, water, snow and ice (Figure 1). Chlorophyll absorbs visible light, especially red light, as a means of providing energy to the process of photosynthesis. As a result, a sensor that measures the reflectance of red light from green vegetation gets very low readings. However, green vegetation reflects near infrared energy very strongly. This is in contrast to the reflectance pattern of soil in the same wavelength range. Soils have stronger reflectance in the visible wavelengths than green vegetation but lower reflectance in the near infrared wavelengths. Thus, a sensor, which can measure the reflectance of red light and near infrared radiation separately, will show significantly different signal patterns for green vegetation and soil. The heavier the vegetative canopy, the stronger the red absorption and near infra-red reflectance. There is a quantitative relationship between green biomass and visible/near infra-red reflectance.
This distinctive reflectance pattern can be expressed quantitatively using a vegetation index. The most popular index is the Normalized Difference Vegetation Index (NDVI), expressed as:
NDVI = (NIR- red)/(NIR + red)
where NIR - reflectance measured in the near infra-red sensor band and
red - reflectance measured in the red sensor band.
NDVI is strongly correlated to vegetation density, which, for a crop, is correlated to grain yield potential. Therefore, satellite-borne remote sensing can provide quantitative measurements relevant to crop health and economic potential.
Also note the difference in reflectance between wet and dry soil (Figure 1). Wet soils show lower reflectance than dry soil throughout the visible and near infrared wavelengths. Thus, panchromatic sensors, which measure reflectance in this range tend to show dry soils as "brighter" (i.e. higher reflectance) than wet soils which are "darker" (i.e. lower reflectance).
However, trash cover also affects the reflectance pattern. Heavy, trash cover tends to lighten black and white images such as SPOT or IKONOS panchromatic or a Landsat-7 ETM. This can confound the interpretation of an image. A wet area, which would appear dark on an image without trash cover, will appear light if the area has been left with a heavy trash cover. Therefore, the interpretation of a black and white image is not simple. Knowledge of the previous year's crop can greatly improve the understanding of the patterns visible on an image.
Figure 1. Visible and near infra-red reflectance from a wheat canopy, dry soil and wet soil
Practical Applications for Crop Management
An NDVI map of an individual field made from a satellite image acquired between late June and early August will show the density of the vegetation on the field near the peak vegetative growth stage. Peak vegetative growth is strongly related to grain yield potential and, in effect, the vegetation density map is a proxy for yield. Side-by-side comparisons of NDVI maps and grain yield maps made from GPS-equipped yield monitors have shown repeatedly that areas with the heaviest vegetation density correspond to those with the highest grain yield. To date, this has been difficult to prove statistically but certainly the correlation is subjectively evident.
There are two important points about peak vegetation NDVI maps. First, the maps provide only a relative measure of yield potential. Actual yield data must be collected from points located precisely in the field in order to develop a conversion from NDVI to yield. This conversion is valid only for the field from which the ground measurements are taken. The NDVI-yield relationship obviously varies between crops but also varies between fields for the same crop. Therefore, extrapolation of an NDVI-yield formula from one field to the next is not valid.
The second point to note is that heavy vegetation can occasionally result from dense weed growth. In these situations, there can be locations on a field with high vegetation density as a result of weeds but grain yields are low. It is important to combine field knowledge with the NDVI map in order to make a sound assessment of relative yield potential.
Field size in Western Canada is large by world standards and the fields, understandably, vary with changes in soil type, topography and management practices. As a result, crop yields are not uniform across a field. Management practices appropriate for one area of a field may not be optimal in another area on the same field. Every field should be analyzed to determine the main factors limiting crop production and how they vary by location. A field may need to be split up into separate management zones, each with its own unique crop management program, in order to maximize the economic potential for crop production on that field.
Satellite image maps showing either the vegetation density at peak vegetative growth or grayscale soil surface color are useful tools for delineating boundaries between zones with different productivity potential. The maps provide a starting point for further analysis to determine the causes behind different growth and soil patterns. An understanding of the factors limiting productivity potential in each zone allows the design of a crop management program to maximize the economic return from each zone.
Every year there are thousands of soil samples taken from fields across Western Canada for the purpose of determining soil nutrient status and the most economic rate of fertilizer to apply. More agronomists are using GPS equipment for "benchmark soil sampling" which means they pinpoint the sample locations with a GPS unit so they can return to those locations in subsequent years. Satellite image maps have excellent utility for benchmark soil sampling.
Satellite image maps are geo-referenced, meaning that the information is tied to its actual location on the earth. With special software, a satellite image can be brought up on a computer screen with the location of a GPS unit overlaid. This allows soil samples to be taken from any specific location indicated on a satellite image map.
Soil fertility is strongly affected by surface drainage, soil type and crop growth. These characteristics are usually apparent on a satellite image map. An agronomist can use the satellite image map to determine where soil samples should be taken. The location will depend upon the objectives of the sampling. In some cases, the samples may need to avoid small anomalous areas so that the analysis is not skewed by the results from those locations. In other situations, the extreme locations may be the subject of analysis, so the satellite image map can guide the agronomist to those particular locations for sampling.
The same methods used to calculate fertilizer recommendations on a field basis can be adapted to determine the most cost-effective fertilizer rate for different zones within a field. Remote sensing capabilities can be used to determine the location of field management zones. The nutrient status of each zone can be determined with target soil sampling and the most cost-effective fertilizer rate for each zone is then calculated in the normal fashion. Crop inputs represent a significant portion of variable crop production costs. It is in every producer's best interest to attain the most efficient and effective use of every pound of fertilizer. Zone-based fertilizer application is a step towards this goal.
Zone-based fertilizer maps can be turned into variable-rate fertilizer prescription maps. A geo-referenced satellite image offers a significant advantage because a prescription map created from an image will also be geo-referenced. The exact location of every point on the field is already part of the image map. The fertilizer rate for each zone must be linked to the management zone map to create a prescription map.
There are many technical steps required to create a fertilizer prescription map from a satellite image. Currently, this part of the technology is not use-friendly. There are various software packages that can be used to create maps for a variable-rate controller, however, they usually have limitations on the type of file format they can handle and there are no established standards between various types of equipment and software. There is still a great deal of work to be done before this problem will be satisfactorily resolved to the point that people with a moderate level of technical competence will be able to create prescription maps.
Currently, the timeliness of remote sensing data delivery from satellite sensors to users is not sufficient to make it useful for real-time crop scouting. There are a number of reasons for this:
1) There may not be a satellite passing over a given field on a particular day.
2) It might be cloudy and the sensor cannot collect data from the earth's surface.
3) It takes several days for the satellite company to deliver an image.
4) It takes time to process an image and produce a map for a particular field.
Repeat coverage is becoming more frequent and data delivery is becoming more rapid. Further improvements will increase the utility of satellite-based remote sensing for crop scouting. The capability of the sensors is also improving. As spatial resolution and spectral capabilities improve, there will be opportunities to develop better crop monitoring capabilities. These could include more time-critical uses such as weed identification and spot spraying.
Remote sensing is not an entire solution. It is one tool and one piece of information that, when used with other management tools can improve crop management and add to your profits. The key to unlocking its benefits lies in combining the strengths of its geo-referenced field information with other sources of knowledge, so that knowledge can then be applied in a distributed fashion field-by-field.