Partial budgeting for acquiring and operating a ground-based optical crop sensor for variable rate nitrogen application

The financial evaluation presented in this paper is based on a field experiment executed between October 2019 and July 2020 at the experimental farm of the University of Natural Resources and Life Sciences, Vienna, which is located in Groß-Enzersdorf, east of Vienna (48°11′28.3″N, 16°33′39.5″E). It is located in the Marchfeld plain, an important crop production region in the northwestern part of the Pannonian Basin. The aim of the experiment was to assess the efficacy of a ground-based optical crop sensor mounted on a tractor in terms of fertilizer savings and yield increasing in a winter wheat (Triticum aestivum L.) field. The evaluated sensor was a CropXplorer (CNH Industrial Österreich GmbH, St. Valentin, Österreich) crop sensor manufactured by Fritzmeier (2020). Figure 1 shows the sensor mounted on a tractor used in the field study. A one-time fertilizer application was performed on April 17, 2020, applying calcium ammonium nitrate (CAN) with 27% N content (Schwaiger, 2021). The execution of fertilizer application was done based on two different scenarios: stimulus and compensatory. In the stimulus scenario, crops with lower yield expectations received more fertilizer and in the compensatory scenario, crops with more yield expectations received less fertilizer. The scenarios were compared to a conventional way of fertilizer application, where fertilizer is applied as business as usual without any sort of precision farming technology involved. Table 1 shows the description of scenarios. The stimulus and compensatory scenarios used the sensor for fertilization based on the original experimental design presented by Schwaiger (2021). Therefore, these two scenarios were chosen to be assessed. An additional third scenario to be evaluated in this study has been generated from the average of the stimulus and compensatory results.

Figure 1

The chosen methods to evaluate the financial implications related to the acquisition of the crop sensor for VRNA were partial budgeting and a payback period calculation. Partial budgeting is a tool used by enterprises to assess the financial implications of any possible change on a farm. This tool focuses on adding and reducing income and costs coming from future changes in a systematic way (Sahs and Doye, 2005). The formula given below is used to evaluate the future change in income (Tigner, 2018): Δ net income=(additional income + reduced costs)−(additional costs + reduced income) \matrix{{\Delta \,{\rm{net}}\,{\rm{income}} = ({\rm{additional}}\,{\rm{income}}\, + \,{\rm{reduced}}\,{\rm{costs}}) -} \cr {({\rm{additional}}\,{\rm{costs}}\, + \,{\rm{reduced}}\,{\rm{income}})} \cr}

An advantage of partial budgeting is that only the additional and reduced cash flows are considered without the necessity of doing a complete budget. Table 2 shows the structure (Rabin et al., 2014), together with the costs and income included in this study.

Results on the additional yield and reduced fertilizer application rates coming from the different scenarios were obtained from the VRNA field experiment executed in 2019–2020. The additional fertilizer rate was compared to the conventional way of fertilizer application (Table 3). Crop yields were not only influenced by the fertilizer application rates – the former determined by the sensor’s initial calibration –, but weather and soil pedoclimate characteristics also affected them. The average temperature was 10°C and the accumulated precipitation was 478 mm during the period of the experiment. The type of soil of the experimental field is calcareous gray with a pH range between 7.5% and 8% and a clay content of 14% (Schwaiger, 2021).

It is relevant to state that due to technical problems with the crop sensor, the second planned fertilizer application was not performed. Therefore, the amount of applied fertilizer was less than half the common rate used for winter wheat production in the Marchfeld plain, but resulted in high yields nevertheless (cf. Neugschwandtner et al., 2015; Moitzi et al., 2020).

The calibration of the sensor played a crucial role in the losses of grain yield in both scenarios. These losses are directly influenced by the amount of fertilizer and the yield obtained. Despite the losses, statistical significances were not found between the scenarios. The gain or loss of yield and the fertilizer application rates in both scenarios were multiplied with the market price in (€/t) of the milling wheat and the fertilizer costs to obtain values for partial budgeting. The milling wheat market price is based on the protein content of the harvested winter wheat. Laboratory results showed an average protein content of 9.6% in the grain (Schwaiger, 2021). This protein content is below the accepted value for milling wheat (12.5%) at the stock exchange for agricultural products in Austria (Gartner, 2018). Hence, a 12.5% protein content was assumed in the grain for purpose of this evaluation. The additional costs involved in the acquisition of the crop sensor were separated into three categories: equipment costs, information costs, and material costs.

Equipment costs: These costs refer to the acquisition of the crop sensor and the additional machinery that is needed for the correct functionality of the sensor. In this case, an AMAZONE ZA-TS 2000 mounted precision spreader and a laptop are the additional devices required. It is assumed that the farmer already has a non-precision spreader to perform conventional fertilization. In this analysis, the resell salvage value of the non-precision spreader is not considered, although a sensitivity analysis scenario reducing the salvage value of the conventional spreader from the initial investment of the precision spreader is going to be shown. The total costs of these devices are annualized based on the buying costs, salvage value of the equipment, and the real interest rate. It is assumed that a commercial financial institution would finance the acquisition of these devices. Annualized payment (r) and the interest rate (i_r) were calculated according to Tekin (2010) as follows: r=[C0−Cn]ir(1+ir)n(1+ir)n−1+Cn*ir r = [{C_0} - {C_n}]{{{i_r}{{(1 + {i_r})}^n}} \over {{{(1 + {i_r})}^n} - 1}} + {C_n}*{i_r} where r is the annual payment (€/year), C_o the purchase price of the equipment (€), C_n the resale value of the equipment (€), n the lifespan of the equipment (years), and i_r is the real interest rate (%). ir=in−ig1+ig {i_r} = {{{i_n} - {i_g}} \over {1 + {i_g}}} where i_r is the real interest rate (%), i_n the interest rate on loan capital (%), and i_g is the inflation rate (%).

The insurance costs are estimated to be 2% of the value of the machinery. To cover the annual repairs and maintenance costs, 3% of the value of machinery was used (ÖKL, 2019).

Information costs: These costs are associated with obtaining site-specific crop and soil data and the methods (software and hardware) to process these data. In this case study, the sensor is initially calibrated based on an NDVI map. Later, the vegetation reflectance measurement index given by the sensor is compared against this initial calibration quantity of the NDVI index. Based on this method, crops with lower/higher sensor-determined vegetation indexes receive higher/lower fertilizer rates. It is not necessary to have any sort of internet connection to operate the sensor; however, it facilitates data transfer and storage to create NDVI maps for further sensor calibration. At the same time, these data can be stored in the cloud, which the sensor’s manufacturer offers as a service. In this study, a laptop with access to a wireless network was included in the information costs section. A onetime payment of 75 € to obtain a wireless internet box and the internet monthly fee of 25 €/month (magenta, 2020) were included in these costs.

The creation of NDVI-based soil management zone maps to calibrate the sensor is also optional, although having these maps gives more accurate information about N deficiency in the crops. To create them – NDVI-based soil management zones maps – it is necessary to have access to satellite images. Precision farming software and/or companies help to determine the management zones and can create yield zoning maps based on NDVI. The yearly satellite mapping software fee is 257 €/year (moneysoft, 2020). The yield zoning map costs are based on the total area to be assessed, 60 € for the first 100 ha and 120 € for more than 100 ha (AgrarCommander, 2020). Although the creation of NDVI-based soil management zone maps is not necessary to utilize the sensor, the potential benefits are augmented by using these maps simultaneously with the sensor. In this study, the costs to create NDVI-based soil management zone maps were included in the information costs. Despite adding the information costs to operate the sensor in this study, a sensitivity analysis scenario without including these costs is presented in the section “Results and discussion.”

Material costs: These costs are related to the extra fertilizer applied to the field zones where yield expectations were lower in comparison to other zones. The application rate is based on the scenarios of the VRNA field experiment. In this case, the stimulus scenario required an extra 7 kg N/ha compared to the conventional scenario to compensate for the nutrient deficiency in the zones projecting low yields.

This amount was later multiplied by the market price of N in CAN. The price before taxes of CAN was 191 €/t.

A summary of some of the equipment and information precision farming costs are shown in Table 4.

It was not projected to reduce any sort of income by using the crop sensor on the farm. However, based on the yield results of the field experiment, a reduction in income was generated due to technical and environmental factors. For the same reason, a reduction in yield income was added to this partial budget, although this is not the norm. It is important to mention that the additional costs were annualized, therefore a present value (PV) annuity factor was used when calculating the yearly cost. The annuity factor was calculated according to Andrew and Gallagher (2007) as follows: pva = pmt*1−(1(1+i)n)i pva\, = \,pmt*{{1 - ({1 \over {{{(1 + i)}^n}}})} \over i} where pva is the PV of annuity (€/year), pmt the annual payment cost (€/year), n the time period (years), and i is the discount rate (%). The financial data utilized at the time of performing this financial analysis is presented in Table 5.

In addition to the results obtained in the field experiment (Table 3), three additional scenarios are presented, where the benefits of the additional yield of using a sensor are added to the partial budgeting presented here. These scenarios were taken from three field experiments using the same type of ground-based optical crop sensor for VRNA, although the Yara Hydro N was used instead of Fritzmeier sensor in the studies of Link et al. (2002) and Mayfield and Trengove (2009). The additional economic benefits (increase in yield in comparison to a conventional way of fertilizer application) due to the use of the crop sensor in these studies were added to the assessed partial budgeting (Table 6). The amount of nitrogen reduction applied in the mean scenario in this research was used to assess the other three studies; only the amount of yield input was changed. It is important to mention that not only the amount of fertilizer applied, but also other factors, such as climate and soil conditions, play a crucial role in the amount of yield obtained in different studies. However, these factors were not considered in this comparison; just the additional yield obtained by using the crop sensor relative to baseline was considered.

The compensatory scenario shows the highest negative change in income (Figure 2). This scenario presents an income of −271.8 €/ha/year on 25 ha of fertilized area and −83.3 €/ha/year on 250 ha of fertilized area. Although this scenario gives the highest amount of fertilizer reduction with −12 kg N/ha, the reduced costs did not cover the additional costs related to the acquisition and operation of the sensor. In addition, the negative amount of yield had a null impact on additional benefits resulting from buying the sensor, since it was the scenario giving a lower yield. On the other hand, the stimulus scenario shows the lowest negative change in income. This scenario presents −227.3 €/ha/year on a 25 ha farm and −38.9 €/ha/year on 250 ha of fertilized area. The stimulus scenario applies an additional amount of fertilizer (+27 kg N/ha), but gives the lowest yield reduction of −69 kg/ha. The third scenario represents the mean of these two strategies, the compensatory and stimulus. A sensitivity analysis of the third scenario was also included in the results. The first sensitivity analysis scenario takes the salvage value of the traditional spreader and reduces this amount to the initial investment of the precision spreader. In this case, the financial losses were −228.4 €/ha/year on 25 ha of fertilized area and −62.3 €/ha/year on 250 ha of fertilized area. This is around a 10% reduction on 25 ha of fertilized area and 4% reduction on 250 ha of fertilized area, relative to the mean scenario. In the second sensitivity analysis, the information costs were removed from the additional costs in partial budgeting. In this case, an 8% reduction on 25 ha of fertilized area and a 3% reduction on 250 ha of fertilized area relative to the mean scenario were achieved.

Figure 2

The negative results on this partial budgeting may be affected by the external factors disturbing the amount of attained yield in all of the scenarios. Either increasing the yield or reducing the usage of the fertilizer may have the same impact on partial budgeting. The reason for this premise is that the difference between the market price of fertilizer (191 €/t) and milling wheat (173 €/t) at the time of this evaluation was only 18 €/t; this is around 10% higher price of the fertilizer relative to the price of the milling wheat. The crucial element is either the amount of additional yield or the reduction of fertilizer or a combination of both, which could pay off the additional precision farming costs attributed to the sensor. In this case, a negative amount of additional yield carries all the additional precision farming costs to the savings coming from the reduction of fertilizer usage.

Results also show that bigger farms have significantly better results than smaller farms. The additional income gained from the increased yield and reduced costs grows in proportion to the size of the fertilized area. However, farms smaller than 250 ha cannot pay for the additional precision farming costs due to the negative amount of yield obtained in all the scenarios relative to the conventional scenario. These results are similar to the ones obtained by Meyer-Aurich et al. (2008) who reported that farms bigger than 250 ha may obtain greater income. The calculated payback period just to cover the cost of the crop sensor device without including the additional necessary components for its operation is shown in Table 7. The mean of the stimulus and the compensatory scenario was chosen to compute the payback period. The reduced costs due to fertilizing savings were used to calculate the time of repayment because the additional amount of yield was negative in all the scenarios. Results show that 25 ha of fertilized area would need 331 years to cover the cost of the sensor with the fertilizer savings, whereas 250 ha of fertilized area would need 31.1 years.

Figure 3 shows a comparison of the partial budgeting results with selected studies (Table 6). It is possible to perceive negative results in the studies of Link et al. (2002) and Mayfield and Trengove (2009), as in both studies, a minimal amount of yield increase occurred due to the use of the sensor, 0.04 t/ha in the study of Mayfield and Trengove (2009) and 0.16 t/ha in the study of Link et al. (2002). The amount of yield increased between conventional and VRNA scenarios in the study of Mayfield and Trengove (2009) was not statistically significant, whereas Link et al. (2002) did not provide results for statistical significance. Galambošová et al. (2015) presented the highest additional economic benefits among all the studies, as 1.25 t/ha additional yield was obtained in this field experiment by also using the same Fritzmeier crop sensor for VRNA. The reason behind the positive result was the higher doses of fertilizer given to the plants, an average 49 kg N/ha in this study versus around 32 kg N/ha in the study of Mayfield and Trengove (2009), for example. At the same time, Galambošová et al. (2015) combined yield potential maps with the Fritzmeier crop sensor to obtain better results, whereas none of the other studies utilized yield potential maps in their methods. A break-even point on a 24-ha fertilized area could be attained if the additional yield results of Galambošová et al. (2015) were used in the mean of stimulus and compensatory scenario on this research.

Figure 3