Method for fertilising cultivated plants

Abstract

The subject of the invention is the development of phosphorus fertilisation recommendations for cultivated plants, comprising steps in which, during the early growth stage of plants, defined on the BBCH scale between BBCH 10 and BBCH 32, the phosphorus content in plant leaves is determined; if the phosphorus content is below 4 g / kg, an aqueous solution of ammonium phosphate at a concentration of 2% is applied at a rate of 300 litres per hectare, allowing for the absorption of phosphorus by the plants.

Description

[0001] METHOD FOR FERTILISING CULTIVATED PLANTS

[0002] The subject of the invention is a method for fertilising cultivated plants. The invention is applicable to foliar fertilisation of plants, particularly in intervention fertilisation.

[0003] In cereal crops, there is a very high demand for nutrients during the early growth stages. At the same time, there is a rapid development of the root system, which requires large amounts of phosphorus. The peak phosphorus uptake occurs in the early growth stages, making the timely supply of this nutrient crucial for optimal crop development.

[0004] Known methods for detecting phosphorus deficiencies involve analysing soil and plant tissue samples, which must be manually collected from the field. These samples are then examined in specialised diagnostic laboratories, which introduces limitations in terms of availability, efficiency, and speed of obtaining results. These factors hinder the precise assessment of plant nutritional status in a timely manner. A lack of complete knowledge regarding phosphorus levels in plants prevents the correct selection of the timing and composition of intervention fertilisation mixtures. Fertilisation with phosphorus outside the optimal time frame does not compensate for existing deficiencies, resulting in lower yields compared to the maximum cultivation potential and leading to crop losses.

[0005] Intervention fertilisation methods are also known in the prior art. For example, document CN 107056437 A discloses an ammonium polyphosphate foliar fertiliser for mulberry leaves and a method for its manufacture. The ammonium polyphosphate foliar fertiliser for mulberry leaves is prepared from the following raw materials in parts by weight: 10 to 12 parts ammonium polyphosphate, 25 to 26 parts urea, 14 to 16 parts potassium nitrate, 0.04 to 0.06 parts calcium nitrate, 0.05 to 0.06 parts magnesium sulphate, 0.5 to 1.5 parts molasses, 0.1 to 0.2 parts sodium dodecylbenzenesulphonate, 15 to 25 parts insecticide, 0.3 to 0.4 parts trace elements, 0.5 to 1 part threonine, 0.5 to 1 part lysine, and 140 to 170 parts water. The prepared ammonium polyphosphate foliar fertiliser for mulberry trees has the advantage of containing a rich and balanced chemical composition. It allows for the rapid replenishment of nutrients by the mulberry tree. The absorption effect is good, the utilisation rate is high, and the fertiliser's effectiveness is strong and long-lasting. After applying the foliar fertiliser to the mulberry tree, proper tree growth is promoted, resistance to diseases and pests improves, leaf quality improves, and the function of increasing production and income is fulfilled.

[0006] Document US6436165B1 discloses foliar fertilisers containing a phosphate ester solution in water at a concentration greater than one molar. Phosphate esters are beneficial polyhydroxy alcohol phosphate esters. An example of such an ester is ethylene glycol phosphate ester. Fertiliser solutions or mixtures are supplied with a water-miscible hygroscopic agent, which retains the applied fertilisers on the leaves in liquid form by extracting moisture from the atmosphere. The hygroscopic agent is preferably supplied by an appropriate hygroscopic alcohol, such as those used in ester production or their substitutes. The esters may be formed into substituted salts using metals or other elements that exhibit nutritional value for plants. Methods for the production of esters and fertilisers have been described. Additionally, innovative methods for application and fertilisation using foliar fertiliser compositions have been presented.

[0007] The objective of the invention is to provide a method for diagnosing phosphorus deficiencies during critical growth stages of winter and spring cereals, with particular emphasis on winter rye. This is achieved using remote sensing methods and the analysis of multi-source hyperspectral data, implemented via computational software, to recommend the timing of intervention fertilisation and crop supplementation. The invention is defined in patent claim 1.The essence of the invention is a method for fertilising cultivated plants, comprising steps in which, during the early growth stage defined on the BBCH scale between BBCH 10 and BBCH 32, the phosphorus content in plant leaves is determined. If the phosphorus content falls below 4 g / kg, an aqueous solution of 2% ammonium phosphate is applied at a rate of 300 litres per hectare, allowing its absorption by the plants.

[0008] Preferably, the invention is characterised in that for maize, phosphorus content is determined at an early growth stage within BBCH 10 to 16.

[0009] Preferably, the invention is characterised in that for cereals, phosphorus content is determined at an early growth stage within BBCH 15 to 32.

[0010] Preferably, the invention is characterised in that fertilisation is repeated 10 to 14 days after the initial application.

[0011] Preferably, the invention is characterised in that the phosphorus content in plant leaves is determined through a sequence of steps in which hyperspectral data is collected. This data consists of multiple electromagnetic wavelength ranges representing cultivated fields (100). The hyperspectral data is then normalised to a common value range and converted into an input vector of a length corresponding to the number of hyperspectral data ranges. The input vector is then fed into a Feedforward Neural Network (FNN). The feedforward neural network (FNN) is trained using steps in which a training dataset is collected, consisting of hyperspectral data comprising multiple electromagnetic wavelength ranges representing cultivated fields (CF) and measurement data representing the phosphorus content in plant leaves from cultivated fields (CF); the hyperspectral data is then normalised to a common value range, after which the training data is converted into an input vector of a length corresponding to the number of hyperspectral data ranges; the dataset is then divided into a training set and a test set; the converted training data is then input into a neural network FNN consisting of three layers: an input layer containing 64 neurons with a ReLU activation function, each receiving an input vector; a hidden layer with a ReLU activation function, consisting of 32 neurons, fully connected to the input layer; and an output layer with a single neuron performing a regression function. The neural network FNN is trained so that the output data, representing the predicted phosphorus content in plant leaves, corresponds to the measured phosphorus content in plant leaves. The trained neural network FNN is then validated for accuracy in diagnosing phosphorus content using the test dataset, and subsequently, in the output block, the neural network processes hyperspectral data consisting of multiple electromagnetic wavelength ranges representing cultivated fields (100) to generate a phosphorus content value for the leaves of cultivated plants.

[0012] The application of the fertilisation method according to the invention has a beneficial effect on increasing yields. Field trials conducted on selected cultivated fields test group and at the Experimental Station of the Institute of Agriculture, SGGW in Warsaw, located in Skierniewice (control group) demonstrated that the timely application of the recommended fertilisation on fields exhibiting phosphorus deficiency increased yield by up to 57% compared to an identical crop where no intervention fertilisation was applied. A key advantage of the method and the computational tool is the ability to remotely detect nutrient deficiencies, preferably in near real-time, at all growth stages of plants, particularly in early developmental phases. This enables the identification of emerging deficiencies and the application of precise intervention fertilisation, thereby increasing yields. The increased yield of cultivated plants, in turn, helps to meet the nutritional needs of a growing population, while precision fertilisation reduces the negative impact of agriculture on the planet's ecosystem.The subject of the invention is further illustrated in a preferred embodiment in the figures:

[0013] Figure 1 a a block diagram of the method according to the invention;

[0014] Figure 1b a block diagram of the computational tool according to the invention;

[0015] Figure 2 structure of the neural network used in the invention;

[0016] Figure 3 contribution of drone-based spectral bands using Sentinel-2 bands in the phosphorus level prediction model for winter rye;

[0017] Figure 4 correlation matrix of the interrelationships between nutrient and metal dynamics in plants and soils;

[0018] Figure 5 scatter plot of the predicted and measured phosphorus content in winter rye using drone-based hyperspectral data;

[0019] Figure 6 scatter plot of the predicted and measured phosphorus content in winter rye using drone-based hyperspectral data corresponding to Sentinel-2 bands.

[0020] Foliar fertilisation of plants with phosphorus is a form of intervention fertilisation. This means that phosphorus is applied when deficiency symptoms of this element appear on the plants in the form of anthocyanin discolouration on the older leaves. In cereal crops (wheat, rye, barley, triticale, rice, oats, and maize), these symptoms occur near the leaf sheath on leaves positioned closest to the soil surface (the first and oldest leaves). Phosphorus deficiency symptoms can be easily identified. Phosphorus is an element that has limited mobility through leaf tissue. Additionally, phosphorus compounds have relatively low solubility in water. For this reason, ammonium phosphate, which has the best solubility and contains 18% nitrogen (N) and 46% P2O5, was selected for foliar application. Ammonium phosphate is a compound fertiliser with the formula (NF ^HPC + NH4H2PO4.

[0021] For spraying, a 2% aqueous solution of ammonium phosphate is used, meaning 2 kg of fertiliser per 100 litres of water. The fertiliser should first be dissolved in warm water and then added to the sprayer and mixed. The spraying should be carried out at a rate of 300 litres of working solution per hectare. The application should be performed in two stages: the first application should take place at the end of the tillering phase in cereals. The second application should be conducted between 10 and 14 days after the first treatment. The key to foliar spraying is to apply the deficient element as early as possible so that the plant can quickly absorb it and incorporate it into its organic structures. Phosphorus plays a crucial role in plants, particularly in the development of generative organs. In cereal crops, these are ears or, in the case of maize, cobs. These structures are formed relatively early. For example, in maize, cobs and the grains they contain develop at the 4 to 6-leaf stage, when the plant is only about 40 cm tall. If a phosphorus deficiency occurs, the plant either produces significantly fewer grains (cobs are only partially filled with kernels) or, in extreme cases, reduces the number of cobs to a minimum. In both cases, this leads to a significant reduction in yield.

[0022] Therefore, the earliest possible application of phosphorus in foliar fertilisation is a key factor in ensuring optimal plant growth. The second aspect is the concentration of the working solution. Prefer The concentration of ammonium phosphate should not exceed 2%., which is determined by the plant's ownphysiology. The concentration of elements in the plant itself is approximately 2.5%. If a higher concentration of the working solution is applied, for example 4% or 8%, the plants will suffer from burn damage. In such a case, in addition to the stress caused by the nutrient deficiency, the plant will also experience additional stress from the burn damage.

[0023] In the prior art, phosphorus fertilisation is typically carried out using granular phosphorus fertilisers, such as superphosphates. However, this approach is inefficient, as phosphorus fertilisers in solid form are absorbed by plants over an extended period, usually taking several months. When phosphorus deficiency symptoms appear, it is crucial to supply the element to the plant as quickly as possible, in contrast to the methods known in the prior art, where the absorption of phosphorus from granular fertilisers takes from several weeks to several months.

[0024] Field trials were conducted in 2023 at two locations in Poland. The samples collected from these fields formed a standardised database, which was used to obtain analytical results. The first location was the Experimental Station of the Institute of Agriculture, SGGW, in Skierniewice (51°57'535N, 20°9'254E), while the second location were test fields near Polczyn-Zdroj in the West Pomeranian Province (53°46.73274'N, 15°57.35202’ E) (Figure 1).

[0025] The Skierniewice Experimental Station operates under the auspices of the Institute of Agriculture at the Warsaw University of Life Sciences. The station has a rich history of over 100 years of long-term static fertilisation experiments, making it the oldest facility of its kind in Poland and one of the oldest in Europe. The unique research infrastructure of the Professor Marian Gorski Experimental Station facilitates the assessment of long-term environmental changes resulting from different fertilisation and crop rotation systems. Through experimentation, the station provides insight into the effects of these agricultural practices on yields and the quality of various plant species. As a leading agricultural research institution, the station serves as an important centre for the advancement of agricultural sciences and for informing farmers about sustainable agricultural practices.

[0026] Two fertilisation experiments, conducted unchanged since 1923, were used in the study. These experiments were established on Luvisol soil (World Reference Base for Soil Resources, 2014), with each experimental plot measuring 36 m2, replicated three times. The initial soil properties are presented in Table 1.

[0027] Table 1. Characteristics of the physical and chemical properties of the soil

[0028] Property Unit Value

[0029] Soil texture % sand 76, silt 8, clay 16

[0030] Total organic carbon (TOC) gkg18.40

[0031] Total nitrogen (TN) 0.76

[0032] pH - 4.75

[0033] mg

[0034] Available phosphorus (P) kg-138.57Available potassium (K) 70.60

[0035] 26.10

[0036] Available magnesium (Mg)

[0037] The soil pH was determined in 1M KCI using the potentiometric method with an automatic pH meter (ISO 10390) ("Soil, treated biowaste and sewage sludge - Determination of pH”, 2021). The total nitrogen content in the soil was determined using the Kjeldahl method (ISO 11261) ("ISO 11261:1995. Soil quality -Determination of total nitrogen by the modified Kjeldahl method”, 1995). The total organic carbon (TOC) content in the soil was determined using a Vario Max CHNS analyser (ISO 10694) ("ISO 10694:1995. Soil quality. Determination of organic and total carbon after dry combustion (elemental analysis)”, 1995). The available forms of phosphorus and potassium were determined using the Egner-Riehm method (PN-R-04023) ("PN-R-04023:1996 Chemical and agricultural analysis - Determination of the content available phosphorus and potassium in soil”, 1996), and magnesium was determined using the Schachtschabel method (PN-R-04020) ("PN-R-04020:1994 / Az1:2004. Chemical and agricultural analysis-determination of the content available magnesium in soil”, 1994).

[0038] Two experiments were conducted for measurements. The first experiment involved a 100-year-old four-field crop rotation with potatoes, spring barley, winter rye, and oats. The following fertiliser combinations were applied: control without fertilisation (0), NK, NP, PK, and NPK. The experiment was conducted under conditions with and without manure application.

[0039] In the second experiment, a 100-year-old monoculture of rye was used. The experiment included the following fertiliser combinations: control without fertilisation - Ca, as well as CaNP, CaKN, CaPK, and CaNPK.

[0040] Plant material samples were collected at three time points from an area of 0.25 m2. At the first time point, the rye was in the late tillering stage (BBCH 24). At the second and third time points, the rye was in the stem elongation stage (BBCH 31 and BBCH 37, respectively). The plant material was then dried at a temperature of 50 °C and ground using a Retsch mill (Katowice, Poland) at a speed of 5000 revolutions per minute. The nitrogen content was determined using the Kjeldahl method, and the phosphorus content was determined using the molybdenum-vanadium method (BN-81 0520-15) after prior mineralisation in a mixture of HNO3and HCIO4.

[0041] A comparative analysis of light reflectance coefficients in various spectral bands from Sentinel-2 and drone images with spectral measurements using a FieldSpec radiometer was performed. Differences in light reflectance between satellite data and in situ measurements may result from various factors, such as crop type, plant growth stage, weather conditions, and even technical aspects of the measurements. Therefore, understanding these differences and their impact on the quality and accuracy of the data is essential for proper interpretation in the context of scientific research and environmental monitoring.

[0042] The study utilised the FieldSpec 4 Standard-Res spectroradiometer, which features a spectral resolution of 3 nm in the VNIR range and 10 nm in the SWIR range. This device is ideal for characterising spectral properties with a resolution of up to 10 nm and is widely used in remote sensing, agricultural analyses, and snow and ice research.

[0043] Before each day of measurement, the FieldSpec device was started at least 30 minutes in advance to ensure proper sensor warm-up. Before each measurement point, or at least every 20-30 minutes, sensor calibrationwas also performed using a white reference panel (spectralon), and measurement readings were verified in the software to confirm stable and uniform values across the entire recorded electromagnetic spectrum. The experiments at the second location were conducted on a 10-hectare field in Polczyn-Zdroj, where the soil is classified as podzolic soil, characterised by a distinct eluvial horizon, which is low in nutrients. Such soils, with limited fertility, require specific agrotechnical treatments and appropriate fertilisation, making this location an excellent site for studying the effectiveness of different cultivation and fertilisation strategies. Winter rye {Secale cereale) was sown in the field as its adaptability to less fertile conditions made it a suitable species for research on podzolic soils. The field was divided into three equal sections. After conducting preliminary soil analyses and identifying phosphorus deficiencies, different types of fertilisers were applied to each section to evaluate their impact on crop yield and plant development. An equal number of samples were collected from measurement points. Measurements were taken during the same three growth stages in both Skierniewice and Polczyn-Zdroj. The lighting conditions varied due to partially changing cloud cover, but the measurement probe used ensured stable lighting conditions. The measurements focused on a single crop type - winter rye - at different growth stages, allowing analysis of how these factors influenced the spectral measurement results.

[0044] At each experimental plot, measurements were taken from three samples of winter rye leaves, with five FieldSpec readings per sample. Measurements were performed using a specialised measurement probe -the ASD FieldSpec 4 Standard-Res spectroradiometer, ensuring precise and repeatable results. During the measurements, time, date, and detailed observations regarding environmental conditions and plant status were recorded. The raw spectral data was processed using the ViewSpecPro software. The original files saved in .asd format were converted into a single table containing reflectance coefficient values. The study examined the prediction of phosphorus content in winter rye, using drone data to simulate PlanetScope spectral bands. PlanetScope, a constellation of over 180 CubeSats known as Doves, provides near-daily global coverage with a spatial resolution resampled to 3 metres. The satellites capture data in multiple bands, including blue, green, red, near infrared, and additional bands, such as Coastal Blue, Green I, Yellow, and Red Edge. PlanetScope data, due to its high temporal resolution and better spatial coverage, especially in cloudy regions, serves as a valuable complement to Sentinel-2 data for vegetation monitoring.

[0045] FieldSpec, a highly precise and reliable portable field spectroradiometer, was used to collect spectral data from test plots. This instrument is known for its ability to measure reflectance spectra, transmittance, radiance intensity, and irradiance intensity across the solar spectrum range from 350 nm to 2500 nm, making it perfectly suited for environmental and agricultural research.

[0046] The object of the invention is also to provide a method for training a diagnostic model used to analyse this data, based on a neural network calibrated using a calibration matrix, which was developed from longterm fertilisation experiments conducted at the Professor Marian Gorski Experimental Station in Skierniewice.

[0047] The invention utilises a method for diagnosing phosphorus deficiencies in crops, based on remotely acquired hyperspectral data, which undergoes a unique computational analysis using custom-developed software, making the process independent of manual field sampling and laboratory diagnostics.

[0048] The objective of the invention is to recommend the timing of intervention fertilisation. It concerns remote observation and analysis of nutrient deficiencies during critical plant growth stages. A key aspect of this solution is the collection of hyperspectral data within precisely defined electromagnetic wavelengthranges, specifically within intervals indicating water content in the plant and its structural properties, as well as chlorophyll absorption and photosynthetic activity. These data facilitate the determination of chemical element content in plant leaves at a given growth stage.

[0049] Figure 1 a presents a block diagram representing the process of creating the computational tool using neural network training, applied to the remote diagnosis of phosphorus content in the leaves of winter rye of the Dahkowskie Zlote variety.

[0050] The cultivated field 100 was regularly photographed from the air using drones 10 and satellites 11. The photographic data was stored in a database 12, from which it could be retrieved by entering geographic coordinates 14. The input module 50 for image processing performs the standardisation process; for example, each sample represents the result of a hyperspectral measurement at a single point in the field (numerical response values for different wavelengths at a single image point). Standardisation consists of subtracting the mean value for each channel and scaling the data to unit variance. Additionally, the phosphorus (P) content in plants from the corresponding section of the field is measured. The measurement data is input into module 60, and at the output of module 60, an input vector is obtained. The set of input vectors for different points in the field forms a training dataset. After standardisation, each sample in the training dataset consists of a numerical vector with a length corresponding to the number of spectral channels and the measured phosphorus content value. Module 70 schematically represents the training module of the FNN network, which is described in detail below with reference to Figure 2. In module 70, the steps related to training the neural network, as described in reference to Figure 2, are carried out. At the output of the process, a trained FNN network model, designated as 80, is obtained, which responds to an input vector and returns the predicted phosphorus content in the leaves, in this case, of winter rye of the Dahkowskie Zlote variety.

[0051] Figure 1b presents a block diagram of the computational tool according to the invention, utilising neural network training for the remote diagnosis of phosphorus content in winter rye leaves of the Dahkowskie Zlote variety. In Figure 1b, the same numerical designations are used to indicate identical elements. The computational tool presented in Figure 1 b receives images from drones 11 or satellites 11 , either directly or through a database collecting image datasets for a given area subjected to aerial or orbital monitoring. The photographic data may be stored in database 12, from which it can be retrieved after entering geographic coordinates 14. The input module 50 for image processing performs the standardisation process. For example, each sample represents the result of a hyperspectral measurement at a single point in the field (numerical response values for different wavelengths at a single image point). Standardisation consists of subtracting the mean value for each channel and scaling the data to unit variance. At the output of module 50, an input vector is obtained, consisting of a numerical vector with a length corresponding to the number of spectral channels in which images were taken for a single point in the cultivated field. In module 90, the input vector is fed into the neural network 80, which was obtained through the process presented in Figure 1a. In response, the predicted phosphorus (P) content in the leaves of the photographed plants is generated. It is essential that the standardisation process, the number of hyperspectral channels, and the wavelengths in these channels are identical to those used in the training of the FNN 80 during the tool development stage.

[0052] Figure 2 presents an example architecture of the FNN neural network used in the solution according to the invention. This is a feedforward neural network (FNN) built using Keras in TensorFlow, consisting of three layers: an input layer with 64 neurons using a ReLU activation function, each receiving an input vector; a hidden layer with 32 neurons using a ReLU activation function, fully connected to the input layer; and an output layer with a single neuron performing a regression function.The first step in this method involves collecting hyperspectral data related to light reflectance from Sentinel-2 spectral channels representing cultivated fields in Redlo near Polczyn-Zdroj, along with measurement data representing phosphorus content in the leaves of winter rye (Dahkowskie Zlote variety). These data were obtained from two field experiments, conducted between the sowing period on 23 September 2022, at a planting density of 550 plants / m2, and the harvest period on 2 August 2023, in the cultivated fields in Redlo near Polczyn-Zdroj.

[0053] The field experiments were conducted at three locations: the Experimental Station of the Institute of Agriculture, SGGW in Warsaw, the experimental fields in Redlo near Polczyn-Zdroj, and the experimental fields in Kuklowka-Zarzeczna. The first experiment involved a four-field crop rotation with potatoes, spring barley, winter rye, and oats, in which the following fertilisation combinations were applied: control without fertilisation (0), NK, NP, PK, and NPK. The experiment was conducted with and without manure application, in three replications. The second experiment utilised a 100-year-old monoculture of rye and included the following fertilisation combinations: control without fertilisation (Ca only), as well as CaNP, CaKN, CaPK, and CaNPK. This experiment was also conducted in three replications. In both experiments, fertilisation was applied at the following rates: 90 kg N ha’1- CO(NH2)2, 26 kg P ha’1- Ca(H2PO4)2, 91 kg K ha’1- KCI, and FYM (manure) at 20 Mg ha1every four years. Plant material samples were collected from an area of 0.25 m2at the time when rye was in the late tillering stage (BBCH 24), followed by another sampling when rye was in the stem elongation stage (BBCH 31 and BBCH 37, respectively). The plant material was then dried at a temperature of 50 °C and ground using a Retsch mill (Katowice, Poland) at a speed of 5000 revolutions per minute. The phosphorus content was determined using the molybdenum-vanadium method after prior mineralisation in a mixture of HNO3 and HCIO4. The availability of phosphorus in the soil in its assimilable forms was determined using the Egner-Riehm method (PN-R-04023). Measurements were carried out during the same growth stage of winter rye.

[0054] Figure 3 presents the contribution of drone-based spectral bands using Sentinel-2 bands in the phosphorus level prediction model for winter rye.

[0055] In this study, the first set of results focused on using comprehensive hyperspectral drone-based measurements to predict phosphorus content in winter rye using feedforward neural networks (FNN). The objective was to determine which spectral bands from drone data were most effective in predicting phosphorus levels, providing a detailed understanding of how specific wavelength ranges contribute to nutrient assessment.

[0056] The spectral bands were divided into several ranges, and their respective contributions to the predictive model were quantified, see Figure 3. First and foremost, the 950-1000 nm spectral range had a significant impact, contributing 13.45% to the predictive model. These wavelengths, located in the near-infrared region, are crucial for assessing vegetation status and biochemical composition. Next, the 900-950 nm range also demonstrated notable importance, with a contribution of 10.08%. This range includes wavelengths commonly associated with plant water content and structural characteristics, highlighting its significance in nutrient status assessment.

[0057] Additionally, spectral channels in the 500-550 nm range had a substantial impact on the predictive model, with a contribution of 13.48%. These spectral ranges correspond to chlorophyll absorption and photosynthetic activity, providing valuable insights into plant vigour and nutrient uptake. Conversely, less significance was observed for ranges, such as 400-450 nm, 550-600 nm, and 600-650 nm, with contributions of 7.43%, 7.37%, and 4.25%, respectively. These wavelengths correspond to regions associated with pigment variations and specific physiological processes in plants.Based on the measurements, a comparative analysis of light reflectance across different spectral bands was conducted, and the bands B2 - 490 nm, B4 - 665 nm, and B1 - 443 nm were selected as input data. These bands also appear in the spectral bands of satellite imagery from the Sentinel-2 mission, allowing the integration of data from different sources. In other preferred embodiments, hyperspectral data representing cultivated fields is collected within the spectral ranges of 500-550 nm and 650-1000 nm.

[0058] In the next step, the hyperspectral data is normalised to a common value range using a computational module, for example, by applying the StandardScaler library.

[0059] The training data for the neural network consists of a set of samples. Each sample represents the result of a hyperspectral measurement at a single point in the field (numerical response values for different wavelengths at a single image point) along with the measured phosphorus content in plants from the corresponding section of the field. Before being fed into the neural network, the hyperspectral measurement values are standardised - for each spectral channel, the mean value is subtracted, and the data is scaled to unit variance. After standardisation, each sample in the training dataset consists of a numerical vector with a length corresponding to the number of spectral channels and the measured phosphorus content value.

[0060] The training data is then converted into an input vector with a length corresponding to the number of hyperspectral data ranges, after which the dataset is split into a training set and a test set in an 80:20 ratio.

[0061] The converted training data is fed into a feedforward neural network built using Keras in TensorFlow, consisting of three layers. The input layer receives a vector of size n, corresponding to the number of hyperspectral measurement channels, with each value being processed by each neuron in the input layer. The input data, derived from hyperspectral measurements, is thus represented as a vector with n values, each corresponding to signal intensity for a specific wavelength after dimensionality reduction. The input layer consists of 64 neurons, each fully connected to every element of the input vector. This means that each of the n input values is transmitted to every neuron in the input layer, and the output of each neuron is calculated as a combination of the input values, incorporating weights and a bias term.

[0062] The feedforward neural network FNN, consisting of an input layer with 64 neurons using a ReLU activation function, each receiving an input vector; a hidden layer with 32 neurons, fully connected to the input layer and using a ReLU activation function; and an output layer with a single neuron performing a regression function, is trained such that the output data, representing the predicted chemical element content in plant leaves, corresponds to the actual chemical element content in plant leaves. The trained FNN network is then validated for its accuracy in diagnosing phosphorus content using the test dataset.

[0063] The training process lasts 10,000 epochs, in batches of 32 samples, optimising the network parameters using the MSE loss function with the Adam optimiser, ensuring optimal convergence between predicted and actual phosphorus values, while using validation data to monitor progress.

[0064] Upon completion of neural network training, the results generated by the network are evaluated on the test dataset to assess prediction accuracy. The MSE and the coefficient of determination (R2) are calculated, and the data is visualised in a scatter plot, indicating predicted phosphorus deficiencies.

[0065] Figure 4 presents the correlation matrix, allowing an in-depth analysis of relationships between different plant and soil parameters in the context of the conducted experiments. These correlations are key to understanding the interactions and relationships that influence nutrient uptake by plants and their overall health. The following analysis describes the observed correlations, highlighting significant relationshipsand their potential implications for agronomic practices. In Figure 4, the “r” values represent correlation coefficients, indicating the strength and direction of linear relationships between variables.

[0066] The matrix reveals a strong positive correlation (r = 0.87) between the nitrogen (N) content and the phosphorus (P) content in the plant. This indicates that as nitrogen content increases, phosphorus content also increases. This relationship suggests synergistic nutrient uptake, which is crucial for plant growth and development.

[0067] The nitrogen content also shows a moderate positive correlation with the zinc (Zn) content in the plant (r = 0.47) and a moderate negative correlation with the copper (Cu) content (r = -0.27). The positive correlation with zinc may indicate that nitrogen fertilisation enhances zinc uptake, whereas the negative correlation with copper suggests an antagonistic interaction between these two nutrients.

[0068] Interestingly, the phosphorus content exhibits similar trends, with a moderate positive correlation with zinc content in the plant (r = 0.44) and a weaker negative correlation with copper content (r = -0.21). These correlations reinforce the idea that nutrient management strategies must consider the interactions between different elements to optimise plant health.

[0069] The correlation between potassium (K) content in the plant and other plant nutrients is generally weaker, with the highest being a moderate positive correlation with copper content (r = 0.23) and manganese (Mn) content (r = 0.35). This suggests that potassium content in plants is less dependent on other macronutrients, and its uptake may be regulated by different physiological mechanisms.

[0070] The correlations between soil parameters also provide valuable insights. The carbon (C) and nitrogen (N) content in the soil are strongly correlated (r = 0.97), indicating that these two parameters often change together, likely due to their shared source in organic matter. Carbon content also shows a strong positive correlation with zinc content in the soil (r = 0.60) and ammonium nitrogen (NH4+) content in the soil (r = 0.67), suggesting that higher organic matter content in the soil may increase the availability of these nutrients.

[0071] The ammonium nitrogen (NH4+) and nitrate nitrogen (NO3“) content in the soil are almost perfectly correlated (r = 0.99), reflecting their role as primary forms of available nitrogen in the soil. This strong correlation highlights the importance of nitrogen management in maintaining soil fertility and supporting plant growth.

[0072] The soil pH exhibits moderate positive correlations with several nutrient contents in both plants and soil, including nitrogen content in plants (r = 0.37) and zinc content in the soil (r = 0.81). The positive correlation with zinc content in the soil suggests that as pH increases, the availability of zinc may also increase, which is crucial for various metabolic processes in plants.

[0073] The correlations between available phosphorus content in the soil and other parameters further emphasise its critical role in soil fertility. The content of these phosphorus forms is moderately correlated with zinc content in plants (r = 0.47) and with the ammonium nitrogen NH4+content in the soil (r = 0.73), indicating that the availability of plant-available phosphorus may influence the uptake of other essential nutrients. Notably, the plant-available manganese content in the soil exhibits a strong positive correlation with ammonium nitrogen NH4+content (r = 0.85) and nitrate nitrogen NO3 content (r = 0.85). This relationship suggests that the availability of plant-available manganese is closely linked to the dynamics of nitrogen transformation in the soil, which may affect plant growth and nutrient utilisation.Figure 5 presents a scatter plot comparing predicted vs. actual phosphorus content, with a reference line and a prediction accuracy percentage for values below 4 g / kg, which is identified as the threshold for phosphorus deficiency.

[0074] The position of a point in the scatter plot in Figure 5 represents the result of comparing the predicted values of available phosphorus content in the soil, as determined by the neural network, with actual in situ measurements obtained under laboratory conditions. The scatter plot of predicted vs. measured phosphorus content exhibits a high degree of correlation, with an R2value of 0.36 and a Root Mean Squared Error (RMSE) of 0.84 g / kg. For an ideal example of the functioning of a neural network, i.e., a case in which all predicted values would correspond to the measured values, the points in Figure 4 chart would lie on the diagonal reference line marked with a dashed line.

[0075] Figure 6 shows the result of comparing the predicted values of available phosphorus content in the soil, as determined by the neural network, with actual in situ measurements obtained under laboratory conditions. The coefficient of determination (R2) for the FNN model was 0.57, indicating a moderate correlation between the predicted and actual phosphorus values. The RMSE was 0.69 g / kg, reflecting the average magnitude of model prediction errors. Additionally, the model achieved 58% accuracy in predicting phosphorus content below the 4 g / kg threshold.

[0076] The first embodiment of the invention, concerning the method for developing a computational tool for remote diagnosis of nutrient deficiencies in plants using machine learning in a neural network, has been presented above in the context of phosphorus content measurement. In another preferred embodiment, the element whose content is determined in plant leaves is nitrogen.

[0077] In a preferred embodiment of the invention, the cultivated plant used for neural network training is winter rye of the Dahkowskie Zlote variety. In other preferred embodiments, element content is determined for other cereal species, particularly maize, wheat, other rye varieties, oats, or rice.

[0078] Preferably, the method according to the invention is implemented for the diagnosis of nutrient deficiencies during the following plant growth stages: BBCH 10-16 for maize, BBCH 15-32 for other crops.

[0079] For the diagnosis of nutrient deficiencies, a computational method is applied, enabling the remote determination of nutrient content in plants using deep learning in a neural network according to the invention. The computational tool implements a method comprising steps in which: hyperspectral data is collected, consisting of multiple electromagnetic wavelength ranges representing cultivated fields (CF), the hyperspectral data is normalised to a common value range, the normalised hyperspectral data is converted into an input vector with a length corresponding to the number of hyperspectral data ranges, the input vector is fed into a neural network (FNN) developed according to the invention, the output block returns the predicted chemical element content in the leaves of cultivated plants, based on the processing of hyperspectral data consisting of multiple electromagnetic wavelength ranges representing cultivated fields (CF).

[0080] In preferred embodiments, the hyperspectral data is obtained from aerial or satellite imagery, depending on availability. Furthermore, in other preferred embodiments, the hyperspectral data in the form of images is automatically retrieved from a hyperspectral image database, with data being selected based on the geographic coordinates of the cultivated field.

Claims

1. Patent claims1. A method for fertilising cultivated plants, comprising steps in which:during the early growth stage of plants, defined on the BBCH scale between BBCH 10 and BBCH 32, the phosphorus content in plant leaves is determined,if the phosphorus content is below 4 g / kg, an aqueous solution of ammonium phosphate at a concentration of 2% is applied at a rate of 300 litres per hectare, facilitating absorption of the solution by the plants.

2. The method according to claim 1 , characterised in that for maize, phosphorus content is determined at an early growth stage, defined on the BBCH scale between BBCH 10 and BBCH 16.

3. The method according to claim 1 , characterised in that for cereals, phosphorus content is determined at an early growth stage, defined on the BBCH scale between BBCH 15 and BBCH 32.

4. The method according to any of claims 1 to 3, characterised in that fertilisation is repeated within 10 to 14 days after the first application.

5. The method according to any of claims 1 to 3, characterised in that the phosphorus content in plant leaves is determined through steps in which:hyperspectral data is collected, consisting of multiple electromagnetic wavelength ranges representing cultivated fields (100), and thenthe hyperspectral data is normalised to a common value range;the normalised hyperspectral data is converted into an input vector with a length corresponding to the number of hyperspectral data ranges;the input vector is fed into a neural network (FNN) trained using steps in which:a training dataset is collected, consisting of hyperspectral data with multiple electromagnetic wavelength ranges representing cultivated fields (CF) and measurement data representing phosphorus content in plant leaves from cultivated fields (CF);the hyperspectral data is normalised to a common value range;the training data is converted into an input vector with a length corresponding to the number of hyperspectral data ranges;the dataset is split into a training set and a test set;the converted training data is fed into a feedforward neural network FNN consisting of three layers:an input layer with 64 neurons with a ReLU activation function, each receiving the input vector,a hidden layer with a ReLU activation function, with 32 neurons, fully connected to the input layer andan output layer with a single neuron performing a regression function, where in the neural network FNN is trained so that the output data, representing the predicted phosphorus content in plant leaves, corresponds to the measured phosphorus content in plant leaves,the trained neural network FNN is then validated for accuracy in diagnosing phosphorus content using the test dataset, and subsequently,in the output block, the neural network processes hyperspectral data consisting of multiple electromagnetic wavelength ranges representing cultivated fields (100) to generate a phosphorus content value for the leaves of cultivated plants.

Images