RESUMO
Modern sensor technologies can provide detailed information about soil variation which allows for more precise application of fertiliser to minimise environmental harm imposed by agriculture. However, growers should lose neither income nor yield from associated uncertainties of predicted nutrient concentrations and thus one must acknowledge and account for uncertainties. A framework is presented that accounts for the uncertainty and determines the cost-benefit of data on available phosphorus (P) and potassium (K) in the soil determined from sensors. For four fields, the uncertainty associated with variation in soil P and K predicted from sensors was determined. Using published fertiliser dose-yield response curves for a horticultural crop the effect of estimation errors from sensor data on expected financial losses was quantified. The expected losses from optimal precise application were compared with the losses expected from uniform fertiliser application (equivalent to little or no knowledge on soil variation). The asymmetry of the loss function meant that underestimation of P and K generally led to greater losses than the losses from overestimation. This study shows that substantial financial gains can be obtained from sensor-based precise application of P and K fertiliser, with savings of up to £121 ha-1 for P and up to £81 ha-1 for K, with concurrent environmental benefits due to a reduction of 4-17 kg ha-1 applied P fertiliser when compared with uniform application. Supplementary Information: The online version contains supplementary material available at 10.1007/s11119-022-09887-2.
RESUMO
The prediction accuracy of soil properties by proximal soil sensing has made their application more practical. However, in order to gain sufficient accuracy, samples are typically air-dried and milled before spectral measurements are made. Calibration of the spectra is usually achieved by making wet chemistry measurements on a subset of the field samples and local regression models fitted to aid subsequent prediction. Both sample handling and wet chemistry can be labour and resource intensive. This study aims to quantify the uncertainty associated with soil property estimates from different methods to reduce effort of field-scale calibrations of soil spectra. We consider two approaches to reduce these expenses for predictions made from visible-near-infrared ((V)NIR), mid-infrared (MIR) spectra and their combination. First, we considered reducing the level of processing of the samples by comparing the effect of different sample conditions (in-situ, unprocessed, air-dried and milled). Second, we explored the use of existing spectral libraries to inform calibrations (based on milled samples from the UK National Soil Inventory) with and without 'spiking' the spectral libraries with a small subset of samples from the study fields. Prediction accuracy of soil organic carbon, pH, clay, available P and K for each of these approaches was evaluated on samples from agricultural fields in the UK. Available P and K could only be moderately predicted with the field-scale dataset where samples were milled. Therefore this study found no evidence to suggest that there is scope to reduce costs associated with sample processing or field-scale calibration for available P and K. However, the results showed that there is potential to reduce time and cost implications of using (V)NIR and MIR spectra to predict soil organic carbon, clay and pH. Compared to field-scale calibrations from milled samples, we found that reduced sample processing lowered the ratio of performance to inter-quartile range (RPIQ) between 0% and 76%. The use of spectral libraries reduced the RPIQ of predictions relative to field-scale calibrations from milled samples between 54% and 82% and the RPIQ was reduced between 29% and 70% for predictions when spectral libraries were spiked. The increase in uncertainty was specific to the combination of soil property and sensor analysed. We conclude that there is always a trade-off between prediction accuracy and the costs associated with soil sampling, sample processing and wet chemical analysis. Therefore the relative merits of each approach will depend on the specific case in question.
RESUMO
How well could one predict the growth of a leafy crop from reflectance spectra from the soil and how might a grower manage the crop in the light of those predictions? Topsoil from two fields was sampled and analysed for various nutrients, particle-size distribution and organic carbon concentration. Crop measurements (lettuce diameter) were derived from aerial-imagery. Reflectance spectra were obtained in the laboratory from the soil in the near- and mid-infrared ranges, and these were used to predict crop performance by partial least squares regression (PLSR). Individual soil properties were also predicted from the spectra by PLSR. These estimated soil properties were used to predict lettuce diameter with a linear model (LM) and a linear mixed model (LMM): considering differences between lettuce varieties and the spatial correlation between data points. The PLSR predictions of the soil properties and lettuce diameter were close to observed values. Prediction of lettuce diameter from the estimated soil properties with the LMs gave somewhat poorer results than PLSR that used the soil spectra as predictor variables. Predictions from LMMs were more precise than those from the PLSR using soil spectra. All model predictions improved when the effects of variety were considered. Predictions from the reflectance spectra, via the estimation of soil properties, can enable growers to decide what treatments to apply to grow lettuce and how to vary their treatments within their fields to maximize the net profit from the crop.
