What value we create?
Figure 1. An average of all of the satellite’s cloud observations between July 2002 and April 2015. Colors range from dark blue (no clouds) to light blue (some clouds) to white (frequent clouds). Based on data collected by the Moderate Resolution Imaging Spectroradiometer (MODIS) on the Aqua satellite. (Source: NASA Earth Observatory Website)

One of the biggest problem in remote sensing are clouds. It is estimated (King et al, 2013) that about 67 percent of Earth’s surface is typically covered by clouds. Especially cloudy conditions dominate around mid-latitudes and areas near the equator as seen in figure 1.Optical satellites (for example Landsat, MODIS Terra/Aqua, Sentinel-2, DigitalGlobe series, etc) are not capable of seeing through clouds which creates shorter or longer gaps in time series of images when observing some phenomena. As stated by the unofficial Murphy's law invented by KZ, the most important events or trends happen during the gaps in the time series.  We can fill the gaps with data from radar satellite Sentinel-1. Sentinel-1 has a synthetic-aperture radar (SAR), which can acquire data in all-weather, day or night. Although SAR data might not substitute optical data in all cases, there are several examples from agriculture domain, civil engineering and environmental monitoring, where SAR data is  a valuable supplement to optical data.

Figure 2. NASA MODIS Terra image (left) and Sentinel-1 SAR image (right) from July 31, 2015.


Figure 3. Conseptual illustration of Sentinel-1 time-series generation.
How do we do it?

We have developed an automated processing chain for Sentinel-1 satellite images to produce time series for several parcel-related parameters. These are the main steps in our process:
  • Validate input geometry (your Area of Interest - AOI)

  • Calculate Sentinel-1 parameters for each parcel/polygon

  • Save the result to database

  • Share time series in appropriate format (database dump, .csv) or through API

Which parameters and statistics we can calculate?

All our parameters are parcel-related, which means they are calculated over the area of your interest (one or many polygons).

ParameterExplanationStatistics
Coh VH
Coh VV 
6-day repeat pass interferometric coherence in VH and VV polarizationmean
VH/VV ratioVH and VV polarization back-scatter ratiomean, median, min, max, stdev
s0VH
s0VV
Calibrated and noise corrected back-scatter (sigma0) in VH and VV polarizationmean, median, min, max, stdev

Why our Sentinel-1 time series are best on the market?
  • Figure 4. Detecting cultivation and harvesting events from Sentinel-1 timeseries (blue lines). Green line is NDVI index from Sentinel-2 optical images.
    Covering smaller parcels – we can reliably process 35 to 50%  more parcels than other services on the market. Depending on the parcel shape we can go down to 0.5 ha parcels.
  • Handling the noise correction professionally – compared to the output of common processing software e.g. ESA SNAP, the parameters are less biased resulting in more accurate and reliable estimation models.
  • Solid performance for operational services – in 2017 we enabled the first country-wide mowing detection system (Estonia) and have kept it operational since then. 

Where time series from satellite data is already used?
Figure 5. Near real-time mowing detection system has been operational in Estonia since 2017.


Our time series together with machine learning algorithms are being successfully used in nation-wide system for automated monitoring of mowing events on grassland. This near real-time system has been operational in Estonia since 2017. In addition to working solution in Estonia we have performed several user trials in Sweden, Denmark and Poland to enhance the existing cutting and grazing detection methodology. Check out our demo map. Different parameters have been used to monitor activities in peat production fields, mapping flood areas and monitoring characteristics of buildings as well as ice coverage, de-forestation, etc.Satellite data is more and more used in precision farming, providing both small and large scale insights into crop health and soil moisture, recommendations for harvesting date and fertilization need. Sentinel-1 time series can improve models which at the moment solely base on optical data. 

Where are our and Sentinel-1 limits? 

Spatial resolution: We use Sentinel-1 products (SLC IW) which have spatial resolution approx. 5x20m.
We can process smaller than 1 ha geometries depending on the shape of the field. Maximum time span: From year 2017 to present. Coverage: Sentinel-1 revisit frequency over Europe, parts of Antarctica, Hawaii islands and Galapagos islands is 6 days.
This coverage is ensured from same, repetitive relative orbits. Over other parts of world the frequency is 12 days.

How are the results delivered?

At the moment we have a very personal approach for each case and we deliver the data in a custom format. We suggest following formats:
• API service
• Postgres database .dump file
• .csv file
• .txt
• .xls fileThis doesn’t suit for you? Definitely we can find the appropriate solution and help with access via API, visualization and web-map creation for larger projects.

We have also capability to process Sentinel-2 images and provide NDVI and other indices time-series in addition. 

Feel free to contact us info@kappazeta.ee.
--------------------------------------
References
NASA Earth Observatory Website: https://earthobservatory.nasa.gov/images/85843/cloudy-earth
Sentinel-1 SAR User Guide Introduction: https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar
King et al, (2013, January 28) Spatial and Temporal Distribution of Clouds Observed by MODIS Onboard the Terra and Aqua Satellites. IEEE Transactions on Geoscience and Remote Sensing, 51 (7), 3826-3852.
K. Koppel, K. Zalite, K. Voormansik & Thomas Jagdhuber (2017) Sensitivity of Sentinel-1 backscatter to characteristics of buildings, International Journal of Remote Sensing, 38:22, 6298-6318, DOI: 10.1080/01431161.2017.1353160
Tamm, T.; Zalite, K.; Voormansik, K.; Talgre, L. Relating Sentinel-1 Interferometric Coherence to Mowing Events on Grasslands. Remote Sens. 2016, 8, 802.
Voormansik, K., Praks, J., Antropov, O., Jagomägi, J. and Zalite, K., 2013. Flood mapping with TerraSAR-X in forested regions in Estonia. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7(2), pp.562-577.