Detecting Rice Fields using Radar Remote Sensing
- Arpit Shah
- Jul 25, 2021
- 7 min read
Updated: Mar 26
INTRODUCTION
A majority of my previous work on Remote Sensing (as of today - July 2021) has involved processing a single Satellite Imagery dataset and interpreting its output. On a few occasions, particularly for workflows that needed Change Detection to be done, I have compared outputs derived from processing two or more Imagery datasets. For this post, I have processed 15 Sentinel-1 SAR Satellite Imagery datasets (Radar/Microwave Remote Sensing) acquired between April and December 2020 to detect the presence of Rice Fields (Paddy Farms) in the Mekong River Delta of Vietnam.

The Mekong River Delta, also known as Vietnam's Rice Bowl, produces more than half of the country's aforesaid crop, Vietnam being the world's 5th largest producer and the 3rd largest exporter of Rice at the time of writing this post (May 2021).
Monitoring Rice production, or the production of any other staple food crop for that matter, is essential. Not only does the cultivation of Paddy employ a significant number of farmers (over 15 million in Vietnam itself), but also information about crop health and growth is of significant importance to stakeholders across the Supply Chain. Just this year, due to production shortage, Vietnam was compelled to import Rice from its trading rival - India.
METHODOLOGY
Why are so many Imagery datasets needed to detect Rice?, you may wonder.
Well, consider interpreting this Optical Satellite Imagery below-

Can you detect the Rice Fields from this aerial view? Would you be able to state with certainty, for example, that the vegetated areas are all Paddy farms while the barren fields aren't?
Of course, not. It is not possible to arrive at that conclusion.
What if I were to give you an added piece of information - this imagery was acquired on a day within the Rice harvesting season.
Would you be able to detect the Rice fields now?
I do not think it would still be possible. Irrespective of the fact that you may be able to delineate the fields that have already been harvested from the ones that haven't, you still do not have sufficient information to delineate a Rice field from one that isn't purely on the basis of this evidence.
You've come to possess additional information about the harvesting characteristics of the Rice crop from a Vietnamese farmer. Now, if I were to multiple aerial photographs from different days within the Rice harvesting season, would you be able to detect the Rice Fields?
You will fare much better than your previous attempts, I suppose.
There are two aspects that I have attempted to drive at-
Having knowledge about the Crop Cycle is useful in order to detect its presence
Optical Imagery, a rendition of the Earth's surface obtained from the reflection of Solar Radiation, offers good but insufficient insights about the presence of Rice fields
Microwave/Radar Remote Sensing has unique characteristics that overcome some of the major limitations of Optical Remote Sensing. How Microwaves interact with surface features on Earth is significantly different than how Sunlight does - and this is of fundamental importance to our Rice Fields Detection workflow (Microwaves are transmitted actively to acquire Synthetic Aperture Radar Imagery whereas Solar Radiation is captured passively to render Optical/Multispectral Imagery).


From sowing to harvesting, Rice Crop typically takes 4 months to be cultivated, on an average. Paddy farms are inundated with standing water up until 7-10 days prior to harvesting to promote Crop Growth and control Weeds. As a result, only during the final stages of growth when the floods are eased does the crop become clearly visible to an observer (to a Satellite in our case). It is this unique Crop Cycle characteristic that will help me to distinguish Rice fields from other types of Farmlands and Land Covers from imagery.

If Microwaves emitted by Radar Satellites were to hit a flooded Rice field (the Rice crop being in a submerged state), the energy waves would hit the water surface and reflect away from the Satellite due to Specular reflection.
However, if the Microwaves were to hit the canopy/foliage of mature Rice Crops above water close to harvesting time, a significant quantity of energy would hit the crop and reflect back towards the Satellite due to Diffuse reflection. Figure 5 above depicts this vital peculiarity.
When energy waves reflect in the direction of its source, it is technically known as Backscatter and significant changes in Backscatter close to the Rice Harvesting season is what I will be looking to monitor in order to delineate the Rice Fields.

Figure 6 above depicts a raw SAR Satellite Imagery Dataset acquired over Mekong River Delta on 9th December, 2020 - the dark pixels represent those underlying areas from where low backscatter was received by the Satellite i.e. these areas are likely flooded (notice the East Vietnam sea on the right is full of jet black pixels) whereas the bright pixels represent those areas from where high backscatter was received by the Satellite i.e. these typically tend to be vegetated and/or urban areas. With the assistance from Optical Aerial Imagery from Google Earth, one can rule out the presence of urban areas in this scene - the Mekong River Delta is predominantly an agri-zone.
ESA's SNAP Software has been used for Imagery Processing and Visualization purposes

Several processing steps, namely: Apply Orbit File → Thermal Noise Removal → Calibration → Terrain Correction → Geographic Extent Subset → Band Subset → Stack Generation → Multi-temporal Speckle Filtering → Decibel Transformation (technicalities elaborated in this document from Page 14 onwards) need to be performed in order to transform the raw Radar Imagery dataset into an output that can be used for analysis and interpretation.
Do observe from the processed Imagery output depicted in Figure 7 above that-
the output has become more vivid this is due to Decibel Transformation of Backscatter readings
the Imagery's Geographic Extent has been clipped to the Mekong River Delta,
the Imagery has been transformed to a Map Projection i.e. how the terrain would appear on a 2D map (originally it was depicted in Radar Coordinates i.e. in the geometry the radar backscatter was captured by the signal receiver onboard the Satellite).

Because I have batch-processed the fifteen input datasets into a single output Imagery Stack i.e. all the processed SAR outputs clubbed in a single product, I am now in a position to prepare a Backscatter Evolution Profile, which is a graphical representation to observe how the Backscatter readings for a particular pixel in the Imagery Stack has evolved through the timeline of the dataset acquisitions (from April 2020 to December 2020).
From Figure 8 above, notice that there are three distinctive troughs in the graph i.e. low Backscatter readings - corresponding to the 2nd, 8th and 13th Imagery dataset (acquired on 13th April, 18th July and 15th November 2020 respectively) in the Stacked product. Also observe that the troughs are followed by periods of increasing Backscatter.
What could this imply?
In Vietnam, Rice is sown three times in a year - typically in spring, autumn and winter. Hence, these would be the times when the Backscatter readings would be at at its lowest. As the Rice crop grows, the Backscatter values would gradually increase as is indicated in the Backscatter Evolution Graph. Thus, I can deduce that the land underlying the selected pixel is being used to cultivate Rice.
The Backscatter Evolution Profile of a Rice Field is distinctive - other types of agricultural fields or other types of land cover would not have the same profile. For example, the Backscatter readings over a residential neighborhood would have minimal changes through time as the underlying surface feature (building/ house) would remain unchanged. If a Farmland is not used to cultivate Rice, then again, the variation in Backscatter would be minimal as the crop will not have the same submerged-to-visible four months-growing cycle as that of Rice.
Thus, the pixels that have large variation in Backscatter readings temporally are likely to be Rice Fields. Of course, Optical Imagery and Ground Truthing can be used to confirm and validate the Radar Imagery Output.
I will proceed to apply this delineating characteristic of Rice Fields to the entire Imagery Stack over the Mekong River Delta in order to render a map which would highlight the fields that are used to cultivate Rice. I will ask SNAP software to compute for each pixel-
a) its Maximum Backscatter value
b) its Minimum Backscatter value, and
c) the Range (Maximum - Minimum) Backscatter value
These computations will be saved as three individual bands of information in the Imagery Stack and I will use it subsequently to create a False-Color Composite as depicted in Figure 9 below-

The symbology applied ranges from blue to violet - the pixels with the former have low variation in Backscatter in our 8-month timeline of 15 Radar Imagery datasets whereas the pixels with the latter have high variation in Backscatter readings - these are likely to be Rice Fields/Paddy Farms.
Hope you enjoyed this study on using Remote Sensing for Rice Field Detection. Can you think of any other crops which can be detected using Radar Remote Sensing? Feel free to share.
ABOUT US
Intelloc Mapping Services, Kolkata | Mapmyops.com offers Mapping services that can be integrated with Operations Planning, Design and Audit workflows. These include but are not limited to Drone Services, Subsurface Mapping Services, Location Analytics & App Development, Supply Chain Services, Remote Sensing Services and Wastewater Treatment. The services can be rendered pan-India and will aid your organization to meet its stated objectives pertaining to Operational Excellence, Sustainability and Growth.
Broadly, the firm's area of expertise can be split into two categories - Geographic Mapping and Operations Mapping. The Infographic below highlights our capabilities-

Our Mapping for Operations-themed workflow demonstrations can be accessed from the firm's Website / YouTube Channel and an overview can be obtained from this brochure. Happy to address queries and respond to documented requirements. Custom Demonstration, Training & Trials are facilitated only on a paid-basis. Looking forward to being of service.
Regards,
Credits: RUS Copernicus