Putting Yield Maps to Use

Here, you’ll explore how yield maps can be used as the starting point for precision management techniques to support sustainable arable farming. This is a summary of new techniques and digital technologies that’ll be covered in more depth over the rest of this week. Combining yield maps with other data and knowledge of the land and farm is a powerful tool for improving yields [1,2,3].

Yield maps can support:

• The development of fertilisation strategies and variable rate applications (VRA) [4]
• Yield forecasting and predicting future yields [5]
• The development of a planting strategy, such as variable planting or growing alternative crops [6]
• The timing of interventions at different points in the growing season to improve yields [7]
• The impact assessment of a new or different management strategy [8]

Variable Rate Fertilisation

Variations in yield can be due to a range of factors such as weather, soil compaction, disease, and soil nutrient status, for example, low soil nitrogen resulting in lower yields. Combining yield maps with other data, such as vegetation indices (which you’ll cover in the next Step) [4] and soil nutrient status [9,5] can highlight areas that would benefit from higher or lower doses of fertiliser. Using these data to generate a prescription map (explored in more detail later this Week) enables variable rate fertilisation within a field, leading to higher yields and less fertiliser wastage [10].

Yield Forecasting

Yield forecasting can be important both locally, for individual farms, and globally. At the farm level, yield forecasting can be used to inform a planting strategy or fertiliser application [2]. Yield forecasting can also be used for financial and insurance purposes, with yield estimates based on past yields being used to estimate future income [8].

On a national or global scale, various precision farming technologies can be combined to forecast yields. For example, bringing together yield maps from previous years, vegetation indices, and soil maps (which you’ll cover in the next Steps) can give a more accurate yield forecast that can be used to support food security – predicting where yields may be higher or lower than anticipated in order to mitigate food shortages and, conversely, food waste [5].

Case Study: Early Yield Prediction

This is an ongoing study being undertaken by Agricolus which demonstrates how yield maps and related precision farming technologies help with forecasting.

Winter cereals are widely cultivated and accurate early yield prediction could provide support for field management strategies, especially optimising nitrogen fertilisation. Accurate early yield predictions make it possible to calculate very precise prescription maps for nitrogen to be applied by variable rate technologies (VRT) directly to the field.

During 2018 and 2019, 39 fields (470 ha) were cultivated with winter cereals (barley, wheat and durum wheat). Yield data were collected through a Volume-flow sensor mounted on the harvesters. Different machine-learning models were tested using sowing day, variety and vegetation index data as yield predictors. The aim was to define which machine learning algorithm was most suitable for the early prediction of yield based on NDVI [Normalised difference vegetative index – to be cover later this Week], and to identify the best parameters to use. A preliminary analysis was carried out and the K-Nearest Neighbor Regression (KNN) algorithm was chosen.

The yield data were split in two sets: set one (training set) was used to train the model and to define the correct parameters. The predicted yield was calculated using the second part of the dataset (test set). The comparison between the predicted yield and actual yield gives a measure of the model’s reliability. Results were finally visualised in a map.

Yield prediction of winter wheat. A: Observed yield data map. B: Predicted yield map. © Agricolus

This study shows that machine learning is a promising technique for predicting early winter cereals yield. However, the model needs to be tested further in different environmental contexts and using data from different years to ensure it gives reliable results every time.

Variable Planting and Sowing

Yield maps used in combination with data such as vegetation indices and soil maps (covered later this Week) highlight possible reasons for less productive areas, so lowering the seeding density of those areas while increasing the seeding density in the more productive areas can increase yields [11] and, potentially, reduce waste and costs [12]. Variable sowing can also include changing the crop [6] or varying the hybrids grown [13].

Reducing Costs

Using yield maps can help to reduce the costs of farming. Yield maps can support targeted (reduced) fertiliser and seed use, leading to reductions in cost. In addition to the financial benefits, precision farming can also have environmental benefits such as reducing pesticide and fertiliser application [14]. Historical yield data combined with records of times and locations of fertiliser applications can also indicate the best time to add fertiliser for maximum yield gain [15].

You can explore some examples of how yield maps have been used to reduce costs in the ‘See Also’ section below.

Could you use yield maps in any of these ways to support your farming? Are there any aspects of the technology that you already use? How might the examples are shown here be used on your farm? Or adapted to fit your crops?