Crop Classification and Monitoring
When we use satellite radar imagery for classifying and monitoring agricultural landscapes, it is very important to have a clear understanding how microwave energy interacts with the ground surface, particularly the soil and the crop canopy. In the following video Dr. Heather McNairn explains the principal radar backscatter mechanisms and how AAFC uses research and know-how to map Canada’s agricultural landscapes.
SAR Interaction with crop canopies
Optical and SAR images see differently
Cloudy skies can challenge the use of optical satellite sensors which operate at relatively short electro-optical wavelengths. Throughout the growing season soils and crops change rapidly; cloud cover leads to gaps in optical satellite data records. These data gaps can lead to lower accuracies when generating satellite data products, or inability to deliver space-based information when it is needed to identify and mitigate risks to agriculture.
SAR sensors operate at much longer microwave wavelengths which are not affected by cloud cover. The ability of SAR satellites to image the Earth regardless of the presence of clouds is an important advantage when tracking changes in crop and soil conditions.
However, satellite-based SARs do more than “seeing” through cloud cover: they see crops differently than optical sensors. Microwave wavelengths are also ideal for estimating the amount of water in crops and soils. Because of these advantages, Agriculture and Agri-Food Canada has been relying on SAR satellite data to assist in mapping and monitoring Canada’s vast agricultural landscapes.
Here is an illustration of optical and SAR images and how microwave and optical wavelengths “see” differently. These images were acquired over an area of intensive agricultural cultivation in Manitoba, the Prairie region of Canada. The image at the bottom is a RADARSAT-2 C-Band SAR acquisition from June 23, 2016, with three HV, HH and VV polarizations displayed in red, green and blue, respectively. The optical Sentinel-2 image at the top was taken just two days earlier on June 21, 2016, with the Near IR, Red and Green bands displayed in red, green and blue, respectively. Note the cloud cover and cloud shadows in the south-west portion of the Sentinel-2 image. This is an area of sandy loam soils where fields are irrigated. The clouds obscure many of these irrigated fields.
Crop growth stages
[Q: where did this text originate?]
As crops grow, their structures change. Crop canopies accumulate leaves, extend stems and stalks, and develop flowers, seeds and fruit. As crop canopies change through the growing season, so does their radar backscatter behaviour in terms of the intensity, or strength.
In this example, data from the RADARSAT-1 SAR is measuring the intensity of HH backscatter at C-Band (~5 cm wavelength). Crops in this region of Canada include small grains (wheat, barley), soybeans, field peas and canola.
It is important to build a time series of SAR images that capture microwave responses to changes in the crop structures over the season. Based on this record of microwave scattering, analysis of multi-temporal SAR imagery can be very useful for classifying crops and tracking their development.
Here is an example of how SAR can be used to estimate crop growth stage.
Take a look at the crop growth estimator for a test site in Manitoba, Canada. Using C- and X-Band SAR data, and Growing Degree Days, this tool can accurately determine the growth stage of canola in daily increments; the tool can also predict the date when each of these 4 fields of canola will enter into different stages of growth.
Sensor frequency and crop canopy
Crops vary in terms of the size of the components of their canopies (leaves, stems, stalks and fruit). The total above-ground-biomass also differs among crops. At peak growth, for example, corn can have more than four times the weight of wet biomass (kg/m2) compared with soybeans.
[[[ next sentence needs work – don’t understand… ]]]
As well over time, crop cover changes with minimal biomass right after emergence and peak biomass accumulations typically occurring prior to senescence.
The use of SAR images taken at different microwave frequencies is very beneficial for increasing crop classification accuracy and tracking crop development. Longer wavelengths, like L-Band (~25 cm), penetrate deeper into canopies and tend to increase radar scattering as they interact with more canopy components. In some cases, these longer wavelengths will also reach the soil, such that part of the backscatter can be attributed to it. Higher X-Band frequencies (~3cm wavelength) will predominantly scatter in the very top of the crop canopy.
[[[ where does this next sentence come from? shaky logic & diction..! ]]]
Because canopies vary from field to field, and over the growing season, integrating [[??]] SARs with different frequencies provides the best [[??]] solution for crop identification and monitoring. The SAR image animation is capturing differences in backscatter from a variety of crop canopies. This example is from an Agriculture and Agri-Food Canada test site in Manitoba (Canada) where the main crops are wheat, soybeans, canola and corn. Note how the VH intensity differs from X-Band, C-Band and L-Band and how this can render unique information on crops. Multi-frequency imagery can provide excellent separation among crop types.
[[ Re: next caption .. does SAR actually “interact” with soil moisture? Or does SAR “respond” / (SAR reponse to..) or is “sensitive to” soil moisture? ]]
SAR interaction with soil moisture
Wet soils in fields of flowering canola
Canola is an oilseed crop from which vegetable oil as well as high protein animal feed can be produced (Canola Council of Canada, 2017). Global production of canola has grown rapidly over the last 40 years, and canola is now the second largest oil crop (United States Department of Agriculture, 2017).
Prolonged wet soil conditions, when present during critical crop development stages, can significantly elevate the risk of some crop diseases. Wet soils in fields of flowering canola are a concern with respect to the development of sclerotinia, a pathogen that feeds on the petals of the canola flower. This fungus can potentially cut crop yield by 50 per cent. In 2010, excessively wet conditions led to widespread infection of sclerotinia in canola crops, costing western Canadian growers C$600 million (Pioneer. 2012).
Whenever canola is in bloom during periods of high moisture, it is important to decide whether or not action is required to mitigate this disease. Given the importance of this crop to Canadian agriculture, Agriculture and Agri-Food Canada and A.U.G. Signals Ltd. (https://augsignals.com/) developed a method to determine when canola crops bloom. This approach used both C-Band RADARSAT-2 and X-Band TerraSAR-X SAR data and a dynamic filtering framework to estimate canola growth stages (McNairn et al., 2018).
SAR interaction with surface roughness [[ same comment/ Q as above re ‘interaction’..]]
RADARSAT-2 soil moisture map
As a rule, radar backscatter tends to increase with increasing soil moisture. Agriculture and Agri-Food Canada applies a physical model – the Integral Equation Model (IEM) – to retrieve surface (0-5 cm) soil moisture information using C-Band SAR data.
In this example, the IEM model is applied to RADARSAT-2 SAR imagery collected over an agricultural region in southern Manitoba where large soil moisture variations can occur. Wetter soils are colour-coded in shades of blue; circular pivot irrigation fields are clearly visible.
[[[ WHY the sudden switch from Canada to Chile ???Also: syntax and logic is very shaky..!I leave ‘as is’ to discuss. DW ]]
Chile is a heavily irrigated country and climate change dictates efficient use of water resources. Chicory is an important field crop for Chile. The tiny chicory seedlings must survive in the top few centimetres of soil for the first 40 days of development and thus, available soil water is imperative. In this example, the IEM model is applied to three RADARSAT-2 images. The uniformly dry soil conditions are obvious on October 20. Three days later, the sandy-textured half of this chicory field is irrigated, with wetter soils colour coded in blue. Drying of the irrigated half of the field is apparent a week later. However, drying is not uniform given the variability in clay and sand fractions in the soil. This type of SAR-based information could be used to assess when and where irrigation is required.
Further references and links related to this sections are located here.
Data collection and processing
Next topic explains how AAFC collects, processes and integrates image and field data to create Annual Space-Based Crop Inventory.
Please complete the quiz below to advance to the next lesson.
Click the ‘Submit’ Button to rate this topic.