Gisel Booman
Jul 6, 2018 · 5 min read

By Taras Kazantsev & Gisel Booman

No-till farming has been widely used in crop fields for the purpose of enhancing soil quality and reducing the risk of erosion. As it turns out, no-till farming doubles as a powerful carbon sequestration management practice, revealing incredible potential of agricultural lands to combat climate change.

Monitoring and encouraging these kind of practices to achieve global carbon drawdown is one of the main aims of Regen Network. Under this premise, our team is developing a methodology to discriminate till and no-till farming from remote sensing and GIS, which will serve as a valuable tool to assess agricultural lands in our network and monitor long-term changes in soil health due to these management practices. The main challenge in this model is discriminating the two categories (including crop species, soil types, and other varying factors) from available, open, regularly-updated satellite data.

Taras Kazantsev and Gisel Booman, member of the Regen Network Sciene Team, give us a first insight into the analysis of Sentinel satellite data for pilot areas in Ukraine and Romania. In this analysis, they test the behavior of three vegetation indices: Normalized Difference Vegetation Index (NDVI), Normalized Difference Tillage Index (NDTI), and Crop Residue Cover Index (CRC).

What We Found

NDVI utilizes reflectance at red and near-infrared regions, and serves as a measure of plant growth activity. Hence, it can be used to separate crops from other land objects, and for detecting periods with the lowest growth activity (or absence of growth). In return, NDTI is based on reflectance at two microwave regions around 1500 and 2100 nm. The region at 2100 nm corresponds to cellulose absorption, thus making it sensitive to plants residue levels on the soil surface.

Preliminary NDTI index results show good discrimination of both practices at early vegetation growing stages for typical East-European crops including wheat, maize, soybean and sunflower (Fig. 1). Soils under tillage show similar characteristics to bare soil, leading to low values of NDTI, whereas values for no-tillage areas are higher. As we can see, annual variations of the NDTI index encompass crops development in areas under tillage, but the index has a more homogenous behavior for no-tillage systems during different growing seasons. This leads to the highest difference in behavior under minimal or absent plant growth (occurring around winter).

Figure 1. Mean NDTI values, averaged between all polygons (n=144) +/- 99% confidence intervals. Mean NDTI values for all polygons were calculated by zonal statistics in GIS by averaging all the pixel values within each polygon.

For the NDVI index, we can observe a good discrimination of both practices at early vegetation growing stages (Fig. 2). Similarly to NDTI, we see a big difference in behavior when observed under minimal or absent plant growth (winter-spring), where crop fields under tillage show lower mean NDVI values. This is consistent between different years, at least for the period under study, which makes it good for developing predictive models. In summer 2017, the relationship was reversed, with much higher NDVI values for crops under tillage. This observed difference would be good to discriminate till from no-till if it was maintained through different years, but we can see that there is more interannual variation at this stage of growth in the behavior of the indices, and thus is less likely to detect clear patterns for prediction.

Figure 2. Mean NDVI values, averaged between all polygons (n=144) +/- 99% confidence intervals. Mean NDVI values for all polygons were calculated by zonal statistics in GIS by averaging all the pixel values within each polygon.

CRC index depends on blue reflectance and SWIR1, which is reflectance from the shortwave infrared portion of the spectrum (1.6 micrometers). This index showed a great discrimination of till and no-till practices for dates around tilling and sowing, with higher CRC mean values for crops under tillage, and similar patterns through different years (Fig. 3). In other words, there is a detectable difference between till and no-till that we can measure using the Crop Residue Cover index, specially for dates around tilling and sowing.

Figure 3. Mean CRC values, averaged between all polygons (n=144) +/- 99% confidence intervals. Mean CRC values for all polygons were calculated by zonal statistics in GIS by averaging all the pixel values within each polygon.

More analysis will be performed as new pilot areas with historical information of land management (“ground truth”) are available in the upcoming weeks. From these results, we expect to define the value ranges for the NDTI, NDVI and CRC at early stages of crop growth that would clearly and accurately separate this two groups or “classes”.

Additionally, we are analyzing radar data from Sentinel I. Here we share some ongoing progress for the first year of data analyzed:

Figure 4. Mean VV-VH values, averaged between all polygons (n=144) +/- 99% confidence intervals. Mean VV-VH values for all polygons were calculated by zonal statistics in GIS by averaging all the pixel values within each polygon.

Looks like there’s a very nice discrimination between till and no till farming at almost all crop stages! Radar data looks very promising, although we need to study longer periods in order to be more conclusive.

Upcoming Challenges

The behavior of these indices can vary across different climate zones, soil types, and bioregions. Because of this, multiple value ranges may have to be defined for different global regions in order to discriminate till from no-till farming. This will be achieved by adding as much well-distributed pilot areas to the GIS as possible. From there, our team will need to perform GIS spatial analysis to discover zonifications.

We are trying also to analyze interannual variations in the patterns observed for the different indices, in order to see if threshold values can be defined in order to classify from remote sensing alone these two classes of land management. If too many interannual variations are present, we can analyze the relationships between these variations and climatic variables, like accumulated rain or temperature.

In Conclusion…

Our analyses of remote sensing results thus far indicate that we are on track towards identifying fields that practice no-till agriculture with high level of accuracy. These algorithms will be embedded in different Ecological State Protocols to determine ecological outcomes linked to the use of regenerative practices.

The ability to monitor and compare regenerative vs. non-regenerative practices like this unlocks vast opportunity to predicting longer shifts in carbon sequestration of land, even before any changes are evident at the ground level. Utilizing this data, we can discern what farms are adhering to regenerative practices, and what kind of sequestration results they will yield. When combined with Regen Ledger technology and our XRN token system, farmers will be financially rewarded on the spot instead of having to wait for detectable changes in soil carbon to prove their impact — impact that without sensing methodologies like this, would otherwise take up to 10 years to become tangible.


Regen Network

A blockchain network of ecological knowledge changing the economics of regenerative agriculture to reverse global warming. Learn more: https://regen.network

Gisel Booman

Written by

Landscape Ecologist - Science Lead at Regen Network.

Regen Network

A blockchain network of ecological knowledge changing the economics of regenerative agriculture to reverse global warming. Learn more: https://regen.network

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade