Nerd For Tech
Published in

Nerd For Tech

Crop rotation in Iowa: application and insights

In our new article, we provide you with the results of analyzing agricultural lands in Carroll, Iowa, with the help of SoilMate!

Soybean rotation (Iowa)

A widely used practice among Midwest farmers, rotating corn and soybeans could potentially contribute to long-term soil organic matter declines.
Rotating corn and soybeans allow farmers to use less nitrogen fertilizer when growing corn. That benefits the environment and will enable farmers to save on input costs. However, studies have noted corn-soybean crop rotation leads to the lower organic matter in the soil compared to land that undergoes continuous corn production or when other crops are included in rotation along with corn and soybeans.
Soybean should be rotated every two or four years. So we decided to analyze agricultural lands during a 4–5 years period and compare NDVI (Normalized Difference Vegetation Index) of these fields to NDVI of the other areas to locate fields that do not follow crop rotation.

Data

For analysis of soybean fields, we used satellite images, processed them using SoilMate — a fully automatic land usage analytics tool. Using artificial intelligence and neural networks, SoilMate can detect agricultural lands on the larger areas, define field boundaries, crop type, yield growth anomalies, plant stress, and calculate their size.

For our task, SoilMate provides services:

  1. Detection of field boundaries.
  2. Classification of field crop types.
  3. Calculation of NDVI for fields of interest.

The data for analysis was collected for the last five years. As a study area, we decided to pick Carroll, Iowa.

Carrol county in Iowa is located on a single tile 15TUG of Sentinel-2 satellite. The area of the investigated zone is about 1350 sq. km.

Field Boundaries

The first task was to detect field boundaries for the selected area. We loaded AOI GeoJson to SoilMate and used the Plot Boundaries option for the selected area to detect field boundaries. The result of processing is shown in Figure 1.

Figure 1. Field geometries of the investigated area.

SoilMate performs boundary detection using neural networks on the selected area and returns JSON with the geometries of detected fields. These shapes will be used to split soybean fields from the others.

Crop Type Classification

The next part was to perform the classification of crops over the last five years. For Carroll county, we predicted crop type for all the fields using SoilMate.

SoilMate crop type classification algorithm is described here. Using SoilMate, it is possible only to choose an area of interest and run processing to receive results. This algorithm uses bands of Sentinel-1 satellite images and all 12 Sentinel-2 satellite images to predict crop type.

When processed, we receive an image where each crop class has its own color. See Figure 2.

Figure 2. Crop types of the fields in Carroll.

This figure is an image for the 2020 crop type classification. Yellow is for soybean, and green is for corn. Such images were created for the last five years from 2016.

NDVI

We tried to notice differences in NDVI for the fields with soybean and other fields in Carroll county. The Normalized difference vegetation index (NDVI) is a simple graphical indicator used to analyze remote sensing measurements, often from a space platform, assessing whether or not the target being observed contains live green vegetation. SoilMate provides the perfect tool for TCI (True Colour Image) and NDVI calculation. We split each year’s season from May till August into 12 sections (1 section is 10–11 days each month). Such a split was used due to the Sentinel-2 mission providing global coverage of Earth’s land surface every ten days with the first spacecraft, reducing to every five days once both are in orbit. So we will exactly receive NDVI images for these periods. We calculated the NDVI for all 12 sections of the selected area. Here is the example NDVI image in Figure 3.

Figure 3. NDVI image for the investigated area.

Analysis

Figure 4. Soybean growing rate in years for all the fields.

Firstly, we visualized a map with years for each field where soybeans were grown. As shown in Figure 4, there is soybean on some fields for 5 and 6 years. These fields are the area of interest in our case.
Using boundaries for SoilMate, we split crop type images and NDVI images into a set of fields. For each field, we created a time series with mean NDVI values for all periods.
The task was to compare fields with 5–6 year soybean growing and areas where soybean is growing for the first time in recent years. Here is the figure for all these time series (Figure 5).

Figure 5. NDVI time series for all the detected fields.

In time series, there are some points where NDVI for all fields is 0, so we decided to drop these time intervals to analyze only valuable data (Figure 6).

Figure 6. NDVI time series for each field with dropped zero

We filtered a few groups from the initial set of fields. The first group is for fields with 5 or 6 year soybeans growing, the second group with areas with no soybean at all, and the third group is a group where soybean was growing for the first time during the investigated time.

The figure compares the time series for these three groups in Figure 7 (a,b,c).

(a).
(b).
(c.)

Figure 7. (a) NDVI time series for fields with 5 year soy, (b) NDVI time series for fields without soy for the recent 5 years, © NDVI time series for fields with one-year soy.

So, there is no difference in time series for these groups.

We decided to calculate the correlation between time series for soybean fields and others. Here is the table with correlation coefficients.

As you can see, all sequences are highly correlated.

The possible cause of such behaviour could be that NDVI shows the quality of the field only on the surface and does not show the quality of soil under plants. The other reason could be that soil moisture degraded for a longer period.

Conclusion

Consequently, our analysis of soybean fields in Carroll, Iowa, shows that NDVI on the fields with many years of soybean is nearly the same as on the others.

As a result of our research, there are the following findings:

  1. SoilMate can return information about NDVI, crop type, and field boundaries for a certain area in a certain period.
  2. There is no significant difference between fields with soybean for five years, fields without soybean for recent years, and fields with first-year soybean, according to NDVI.
  3. The possible causes of the absence of difference were described.

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Soil Mate

AI-powered automation tool for collecting analytics data from agricultural fields all around the world!