Hyperspectral Analysis of Leaf Pigments and Nutritional Elements in Tallgrass Prairie Vegetation

Spectroscopy-based chemometrics combined machine learning modeling predicts cashew foliar macro- and micronutrients

Precision nutrient management in orchard crops needs precise, accurate, and real-time information on the plant's nutritional status. This is limited by the fact that it requires extensive leaf sampling and chemical analysis when it is to be done over more extensive areas like field- or landscape scale. Thus, rapid, reliable, and repeatable means of nutrient estimations are needed. In this context, lab-based remote sensing or spectroscopy has been explored in the current study to predict the foliar nutritional status of the cashew crop. Novel spectral indices (normalized difference and simple ratio), chemometric modeling, and partial least square regression (PLSR) combined machine learning modeling of the visible near-infrared hyperspectral data were employed to predict macro- and micronutrients content of the cashew leaves. The full dataset was divided into calibration (70 % of the full dataset) and validation (30 % of the full dataset) datasets. An independent validation dataset was used for the validation of the algorithms tested. The approach of spectral indices yielded very poor and unreliable predictions for all eleven nutrients. Among the chemometric models tested, the performance of the PLSR was the best, but still, the predictions were not acceptable. The PLSR combined machine learning modeling approach yielded acceptable to excellent predictions for all the nutrients except sulphur and copper. The best predictions were observed when PLSR was combined with Cubist for nitrogen, phosphorus, potassium, manganese, and zinc; support vector machine regression for calcium, magnesium, iron, copper, and boron; elastic net for sulphur. The current study showed hyperspectral remote sensing-based models could be employed for non-destructive and rapid estimation of cashew leaf macro- and micro-nutrients. The developed approach is suggested to employ within the operational workflows for site-specific and precision nutrient management of the cashew orchards.

Fluorescence and Hyperspectral Sensors for Nondestructive Analysis and Prediction of Biophysical Compounds in the Green and Purple Leaves of Tradescantia Plants

The application of non-imaging hyperspectral sensors has significantly enhanced the study of leaf optical properties across different plant species. In this study, chlorophyll fluorescence (ChlF) and hyperspectral non-imaging sensors using ultraviolet-visible-near-infrared shortwave infrared (UV-VIS-NIR-SWIR) bands were used to evaluate leaf biophysical parameters. For analyses, principal component analysis (PCA) and partial least squares regression (PLSR) were used to predict eight structural and ultrastructural (biophysical) traits in green and purple Tradescantia leaves. The main results demonstrate that specific hyperspectral vegetation indices (HVIs) markedly improve the precision of partial least squares regression (PLSR) models, enabling reliable and nondestructive evaluations of plant biophysical attributes. PCA revealed unique spectral signatures, with the first principal component accounting for more than 90% of the variation in sensor data. High predictive accuracy was achieved for variables such as the thickness of the adaxial and abaxial hypodermis layers (R2 = 0.94) and total leaf thickness, although challenges remain in predicting parameters such as the thickness of the parenchyma and granum layers within the thylakoid membrane. The effectiveness of integrating ChlF and hyperspectral technologies, along with spectroradiometers and fluorescence sensors, in advancing plant physiological research and improving optical spectroscopy for environmental monitoring and assessment. These methods offer a good strategy for promoting sustainability in future agricultural practices across a broad range of plant species, supporting cell biology and material analyses.

Water content estimation of conifer needles using leaf-level hyperspectral data

Water is a crucial component for plant growth and survival. Accurately estimating and simulating plant water content can help us promptly monitor the physiological status and stress response of vegetation. In this study, we constructed water loss curves for three types of conifers with morphologically different needles, then evaluated the applicability of 12 commonly used water indices, and finally explored leaf water content estimation from hyperspectral data for needles with various morphology. The results showed that the rate of water loss of Olgan larch is approximately 8 times higher than that of Chinese fir pine and 21 times that of Korean pine. The reflectance changes were most significant in the near infrared region (NIR, 780-1300 nm) and the short-wave infrared region (SWIR, 1300-2500 nm). The water sensitive bands for conifer needles were mainly concentrated in the SWIR region. The water indices were suitable for estimating the water content of a single type of conifer needles. The partial least squares regression (PLSR) model is effective for the water content estimation of all three morphologies of conifer needles, demonstrating that the hyperspectral PLSR model is a promising tool for estimating needles water content.

A Multi-Target Regression Method to Predict Element Concentrations in Tomato Leaves Using Hyperspectral Imaging

Recent years have seen the development of novel, rapid, and inexpensive techniques for collecting plant data to monitor the nutritional status of crops. These techniques include hyperspectral imaging, which has been widely used in combination with machine learning models to predict element concentrations in plants. When there are multiple elements, the machine learning models are trained with spectral features to predict individual element concentrations; this type of single-target prediction is known as single-target regression. Although this method can achieve reliable accuracy for some elements, there are others that remain less accurate. We aimed to improve the accuracy of element concentration predictions by using a multi-target regression method that sequentially augmented the original input features (hyperspectral imaging) by chaining the predicted element concentration values. To evaluate the multi-target method, the concentrations of 17 elements in tomato leaves were predicted and compared with the single-target regression results. We trained 5 machine learning models with hyperspectral data and predicted element concentration values and found a significant improvement in the prediction accuracy for 10 elements (Mg, P, S, Mn, Fe, Co, Cu, Sr, Mo, and Cd). Furthermore, our multi-target regression method outperformed single-target predictions by increasing the coefficient of determination (R2) for elements such as Mn, Cu, Co, Fe, and Mg by 12.5%, 10.3%, 11%, 10%, and 8.4%, respectively. Hence, our multi-target method can improve the accuracy of predicting 10-element concentrations compared to single-target regression.

Rapid Prediction of Nutrient Concentration in Citrus Leaves Using Vis-NIR Spectroscopy

The nutritional diagnosis of crops is carried out through costly foliar ionomic analysis in laboratories. However, spectroscopy is a sensing technique that could replace these destructive analyses for monitoring nutritional status. This work aimed to develop a calibration model to predict the foliar concentrations of macro and micronutrients in citrus plantations based on rapid non-destructive spectral measurements. To this end, 592 'Clementina de Nules' citrus leaves were collected during several months of growth. In these foliar samples, the spectral absorbance (430-1040 nm) was measured using a portable spectrometer, and the foliar ionomics was determined by emission spectrometry (ICP-OES) for macro and micronutrients, and the Kjeldahl method to quantify N. Models based on partial least squares regression (PLS-R) were calibrated to predict the content of macro and micronutrients in the leaves. The determination coefficients obtained in the model test were between 0.31 and 0.69, the highest values being found for P, K, and B (0.60, 0.63, and 0.69, respectively). Furthermore, the important P, K, and B wavelengths were evaluated using the weighted regression coefficients (BW) obtained from the PLS-R model. The results showed that the selected wavelengths were all in the visible region (430-750 nm) related to foliage pigments. The results indicate that this technique is promising for rapid and non-destructive foliar macro and micronutrient prediction.

A Novel Method for Estimating Chlorophyll and Carotenoid Concentrations in Leaves: A Two Hyperspectral Sensor Approach

Leaf optical properties can be used to identify environmental conditions, the effect of light intensities, plant hormone levels, pigment concentrations, and cellular structures. However, the reflectance factors can affect the accuracy of predictions for chlorophyll and carotenoid concentrations. In this study, we tested the hypothesis that technology using two hyperspectral sensors for both reflectance and absorbance data would result in more accurate predictions of absorbance spectra. Our findings indicated that the green/yellow regions (500-600 nm) had a greater impact on photosynthetic pigment predictions, while the blue (440-485 nm) and red (626-700 nm) regions had a minor impact. Strong correlations were found between absorbance (R2 = 0.87 and 0.91) and reflectance (R2 = 0.80 and 0.78) for chlorophyll and carotenoids, respectively. Carotenoids showed particularly high and significant correlation coefficients using the partial least squares regression (PLSR) method (R2C = 0.91, R2cv = 0.85, and R2P = 0.90) when associated with hyperspectral absorbance data. Our hypothesis was supported, and these results demonstrate the effectiveness of using two hyperspectral sensors for optical leaf profile analysis and predicting the concentration of photosynthetic pigments using multivariate statistical methods. This method for two sensors is more efficient and shows better results compared to traditional single sensor techniques for measuring chloroplast changes and pigment phenotyping in plants.

Dissection of Hyperspectral Reflectance to Estimate Photosynthetic Characteristics in Upland Cotton (Gossypium hirsutum L.) under Different Nitrogen Fertilizer Application Based on Machine Learning Algorithms

Hyperspectral technology has enabled rapid and efficient nitrogen monitoring in crops. However, most approaches involve direct monitoring of nitrogen content or physiological and biochemical indicators directly related to nitrogen, which cannot reflect the overall plant nutritional status. Two important photosynthetic traits, the fraction of absorbed photosynthetically active radiation (FAPAR) and the net photosynthetic rate (Pn), were previously shown to respond positively to nitrogen changes. Here, Pn and FAPAR were used for correlation analysis with hyperspectral data to establish a relationship between nitrogen status and hyperspectral characteristics through photosynthetic traits. Using principal component and band autocorrelation analyses of the original spectral reflectance, two band positions (350-450 and 600-750 nm) sensitive to nitrogen changes were obtained. The performances of four machine learning algorithm models based on six forms of hyperspectral transformations showed that the light gradient boosting machine (LightGBM) model based on the hyperspectral first derivative could better invert the Pn of function-leaves in cotton, and the random forest (RF) model based on hyperspectral first derivative could better invert the FAPAR of the cotton canopy. These results provide advanced metrics for non-destructive tracking of cotton nitrogen status, which can be used to diagnose nitrogen nutrition and cotton growth status in large farms.

Classification and Prediction by Pigment Content in Lettuce (Lactuca sativa L.) Varieties Using Machine Learning and ATR-FTIR Spectroscopy

Green or purple lettuce varieties produce many secondary metabolites, such as chlorophylls, carotenoids, anthocyanins, flavonoids, and phenolic compounds, which is an emergent search in the field of biomolecule research. The main objective of this study was to use multivariate and machine learning algorithms on Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR)-based spectra to classify, predict, and categorize chemometric attributes. The cluster heatmap showed the highest efficiency in grouping similar lettuce varieties based on pigment profiles. The relationship among pigments was more significant than the absolute contents. Other results allow classification based on ATR-FTIR fingerprints of inflections associated with structural and chemical components present in lettuce, obtaining high accuracy and precision (>97%) by using principal component analysis and discriminant analysis (PCA-LDA)-associated linear LDA and SVM machine learning algorithms. In addition, PLSR models were capable of predicting Chla, Chlb, Chla+b, Car, AnC, Flv, and Phe contents, with R2P and RPDP values considered very good (0.81−0.88) for Car, Anc, and Flv and excellent (0.91−0.93) for Phe. According to the RPDP metric, the models were considered excellent (>2.10) for all variables estimated. Thus, this research shows the potential of machine learning solutions for ATR-FTIR spectroscopy analysis to classify, estimate, and characterize the biomolecules associated with secondary metabolites in lettuce.

Remotely Sensed Spatiotemporal Variation in Crude Protein of Shortgrass Steppe Forage

In the Great Plains of central North America, sustainable livestock production is dependent on matching the timing of forage availability and quality with animal intake demands. Advances in remote sensing technology provide accurate information for forage quantity. However, similar efforts for forage quality are lacking. Crude protein (CP) content is one of the most relevant forage quality determinants of individual animal intake, especially below an 8% threshold for growing animals. In a set of shortgrass steppe paddocks with contrasting botanical composition, we (1) modeled the spatiotemporal variation in field estimates of CP content against seven spectral MODIS bands, and (2) used the model to assess the risk of reaching the 8% CP content threshold during the grazing season for paddocks with light, moderate, or heavy grazing intensities for the last 22 years (2000–2021). Our calibrated model explained up to 69% of the spatiotemporal variation in CP content. Different from previous investigations, our model was partially independent of NDVI, as it included the green and red portions of the spectrum as direct predictors of CP content. From 2000 to 2021, the model predicted that CP content was a limiting factor for growth of yearling cattle in 80% of the years for about 60% of the mid-May to October grazing season. The risk of forage quality being below the CP content threshold increases as the grazing season progresses, suggesting that ranchers across this rangeland region could benefit from remotely sensed CP content to proactively remove yearling cattle earlier than the traditional October date or to strategically provide supplemental protein sources to grazing cattle.

Quantifying Nutrient Content in the Leaves of Cowpea Using Remote Sensing

Although hyperspectral remote sensing techniques have increasingly been used in the nutritional quantification of plants, it is important to understand whether the method shows a satisfactory response during the various phenological stages of the crop. The aim of this study was to quantify the levels of phosphorus (P), potassium (K), calcium (Ca) and zinc (Zn) in the leaves of Vigna Unguiculata (L.) Walp using spectral data obtained by a spectroradiometer. A randomised block design was used, with three treatments and twenty-five replications. The crop was evaluated at three growth stages: V4, R6 and R9. Single-band models were fitted using simple correlations. For the band ratio models, the wavelengths were selected by 2D correlation. For the models using partial least squares regression (PLSR), the stepwise method was used. The model showing the best fit was used to estimate the phosphorus content in the single-band (R² = 0.62; RMSE = 0.54 and RPD = 1.61), band ratio (R² = 0.66; RMSE = 0.65 and RPD = 1.52) and PLSR models, using data from each of the phenological stages (R² = 0.80; RMSE = 0.47 and RPD = 1.66). Accuracy in modelling leaf nutrients depends on the phenological stage, as well as the amount of data used, and is more accurate with a larger number of samples.

Near-Infrared Spectroscopy (NIRS) and Optical Sensors for Estimating Protein and Fiber in Dryland Mediterranean Pastures

Dryland pastures provide the basis for animal sustenance in extensive production systems in Iberian Peninsula. These systems have temporal and spatial variability of pasture quality resulting from the diversity of soil fertility and pasture floristic composition, the interaction with trees, animal grazing, and a Mediterranean climate characterized by accentuated seasonality and interannual irregularity. Grazing management decisions are dependent on assessing pasture availability and quality. Conventional analytical determination of crude protein (CP) and fiber (neutral detergent fiber, NDF) by reference laboratory methods require laborious and expensive procedures and, thus, do not meet the needs of the current animal production systems. The aim of this study was to evaluate two alternative approaches to estimate pasture CP and NDF, namely one based on near-infrared spectroscopy (NIRS) combined with multivariate data analysis and the other based on the Normalized Difference Vegetation Index (NDVI) measured in the field by a proximal active optical sensor (AOS). A total of 232 pasture samples were collected from January to June 2020 in eight fields. Of these, 96 samples were processed in fresh form using NIRS. All 232 samples were dried and subjected to reference laboratory and NIRS analysis. For NIRS, fresh and dry samples were split in two sets: a calibration set with half of the samples and an external validation set with the remaining half of the samples. The results of this study showed significant correlation between NIRS calibration models and reference methods for quantifying pasture quality parameters, with greater accuracy in dry samples (R2 = 0.936 and RPD = 4.01 for CP and R2 = 0.914 and RPD = 3.48 for NDF) than fresh samples (R2 = 0.702 and RPD = 1.88 for CP and R2 = 0.720 and RPD = 2.38 for NDF). The NDVI measured by the AOS shows a similar coefficient of determination to the NIRS approach with pasture fresh samples (R2 = 0.707 for CP and R2 = 0.648 for NDF). The results demonstrate the potential of these technologies for estimating CP and NDF in pastures, which can facilitate the farm manager’s decision making in terms of the dynamic management of animal grazing and supplementation needs.

Evaluation of effective quantum yields of photosystem II for CO2 leakage monitoring in carbon capture and storage sites

Vegetation monitoring can be used to detect CO₂ leakage in carbon capture and storage (CCS) sites because it can monitor a large area at a relatively low cost. However, a rapidly responsive, sensitive, and cost-effective plant parameters must be suggested for vegetation monitoring to be practically utilized as a CCS management strategy. To screen the proper plant parameters for leakage monitoring, a greenhouse experiment was conducted by exposing kale (Brassica oleracea var. viridis), a sensitive plant, to 10%, 20%, and 40% soil CO₂ concentrations. Water and water with CO₂ stress treatments were also introduced to examine the parameters differentiating CO₂ stress from water stresses. We tested the hypothesis that chlorophyl fluorescence parameters would be early and sensitive indicator to detect CO₂ leakage. The results showed that the fluorescence parameters of effective quantum yield of photosystem II (Y(II)), detected the difference between CO₂ treatments and control earlier than any other parameters, such as chlorophyl content, hyperspectral vegetation indices, and biomass. For systematic comparison among many parameters, we proposed an indicator evaluation score (IES) method based on four categories: CO₂ specificity, early detection, field applicability, and cost. The IES results showed that fluorescence parameters (Y(II)) had the highest IES scores, and the parameters from spectral sensors (380–800 nm wavelength) had the second highest values. We suggest the IES system as a useful tool for evaluating new parameters in vegetation monitoring.

Monitoring the Foliar Nutrients Status of Mango Using Spectroscopy-Based Spectral Indices and PLSR-Combined Machine Learning Models

Conventional methods of plant nutrient estimation for nutrient management need a huge number of leaf or tissue samples and extensive chemical analysis, which is time-consuming and expensive. Remote sensing is a viable tool to estimate the plant’s nutritional status to determine the appropriate amounts of fertilizer inputs. The aim of the study was to use remote sensing to characterize the foliar nutrient status of mango through the development of spectral indices, multivariate analysis, chemometrics, and machine learning modeling of the spectral data. A spectral database within the 350–1050 nm wavelength range of the leaf samples and leaf nutrients were analyzed for the development of spectral indices and multivariate model development. The normalized difference and ratio spectral indices and multivariate models–partial least square regression (PLSR), principal component regression, and support vector regression (SVR) were ineffective in predicting any of the leaf nutrients. An approach of using PLSR-combined machine learning models was found to be the best to predict most of the nutrients. Based on the independent validation performance and summed ranks, the best performing models were cubist (R2 ≥ 0.91, the ratio of performance to deviation (RPD) ≥ 3.3, and the ratio of performance to interquartile distance (RPIQ) ≥ 3.71) for nitrogen, phosphorus, potassium, and zinc, SVR (R2 ≥ 0.88, RPD ≥ 2.73, RPIQ ≥ 3.31) for calcium, iron, copper, boron, and elastic net (R2 ≥ 0.95, RPD ≥ 4.47, RPIQ ≥ 6.11) for magnesium and sulfur. The results of the study revealed the potential of using hyperspectral remote sensing data for non-destructive estimation of mango leaf macro- and micro-nutrients. The developed approach is suggested to be employed within operational retrieval workflows for precision management of mango orchard nutrients.

Grazing Intensity Alters Leaf and Spike Photosynthesis, Transpiration, and Related Parameters of Three Grass Species on an Alpine Steppe in the Qilian Mountains

The effect of grazing on leaf photosynthesis has been extensively studied. However, the influence of grazing on photosynthesis in other green tissues, especially spike, has remained poorly understood. This study investigated the impact of different grazing intensities (light grazing (LG), medium grazing (MG), and heavy grazing (HG)) on leaf and spike photosynthesis parameters and photosynthetic pigments of three grass species (Stipa purpurea, Achnatherum inebrians, and Leymus secalinus) on an alpine steppe in the Qilian Mountains. Grazing promoted leaf photosynthesis rate in S. purpurea and L. secalinus but reduced it in A. inebrians. Conversely, spike photosynthesis rate decreased in S. purpurea and L. secalinus under intense grazing, while there was no significant difference in spike photosynthesis rate in A. inebrians. The leaf and spike net photosynthetic rate (Pn) and transpiration rate (Tr) in S. purpurea were the greatest among the three species, while their organ temperatures were the lowest. On the other hand, grazing stimulated leaf chlorophyll biosynthesis in S. purpurea and L. secalinus but accelerated leaf chlorophyll degradation in A. inebrians. Furthermore, spike chlorophyll biosynthesis was inhibited in the three species under grazing, and only L. secalinus had the ability to recover from the impairment. Grazing had a positive effect on leaf photosynthesis parameters of S. purpurea and L. secalinus but a negative effect on those of A. inebrians. However, spike photosynthesis parameters were negatively influenced by grazing. Among the three species investigated, S. purpurea displayed the greatest ability for leaf and spike photosynthesis to withstand and acclimate to grazing stress. This study suggests that moderate grazing enhanced leaf photosynthetic capacity of S. purpurea and L. secalinus but reduced it in A. inebrians. However, spike photosynthetic capacity of three grass species decreased in response to grazing intensities.

Identification of plant leaf phosphorus content at different growth stages based on hyperspectral reflectance

BACKGROUND

Modern agriculture strives to sustainably manage fertilizer for both economic and environmental reasons. The monitoring of any nutritional (phosphorus, nitrogen, potassium) deficiency in growing plants is a challenge for precision farming technology. A study was carried out on three species of popular crops, celery (Apium graveolens L., cv. Neon), sugar beet (Beta vulgaris L., cv. Tapir) and strawberry (Fragaria × ananassa Duchesne, cv. Honeoye), fertilized with four different doses of phosphorus (P) to deliver data for non-invasive detection of P content.

RESULTS

Data obtained via biochemical analysis of the chlorophyll and carotenoid contents in plant material showed that the strongest effect of P availability for plants was in the diverse total chlorophyll content in sugar beet and celery compared to that in strawberry, in which P affects a variety of carotenoid contents in leaves. The measurements performed using hyperspectral imaging, obtained in several different stages of plant development, were applied in a supervised classification experiment. A machine learning algorithm (Backpropagation Neural Network, Random Forest, Naive Bayes and Support Vector Machine) was developed to classify plants from four variants of P fertilization. The lowest prediction accuracy was obtained for the earliest measured stage of plant development. Statistical analyses showed correlations between leaf biochemical constituents, phosphorus fertilization and the mass of the leaf/roots of the plants.

CONCLUSIONS

Obtained results demonstrate that hyperspectral imaging combined with artificial intelligence methods has potential for non-invasive detection of non-homogenous phosphorus fertilization on crop levels.

Leaf Nitrogen Concentration and Plant Height Prediction for Maize Using UAV-Based Multispectral Imagery and Machine Learning Techniques

Under ideal conditions of nitrogen (N), maize (Zea mays L.) can grow to its full potential, reaching maximum plant height (PH). As a rapid and nondestructive approach, the analysis of unmanned aerial vehicles (UAV)-based imagery may be of assistance to estimate N and height. The main objective of this study is to present an approach to predict leaf nitrogen concentration (LNC, g kg−1) and PH (m) with machine learning techniques and UAV-based multispectral imagery in maize plants. An experiment with 11 maize cultivars under two rates of N fertilization was carried during the 2017/2018 and 2018/2019 crop seasons. The spectral vegetation indices (VI) normalized difference vegetation index (NDVI), normalized difference red-edge index (NDRE), green normalized difference vegetation (GNDVI), and the soil adjusted vegetation index (SAVI) were extracted from the images and, in a computational system, used alongside the spectral bands as input parameters for different machine learning models. A randomized 10-fold cross-validation strategy, with a total of 100 replicates, was used to evaluate the performance of 9 supervised machine learning (ML) models using the Pearson’s correlation coefficient (r), mean absolute error (MAE), coefficient of regression (R²), and root mean square error (RMSE) metrics. The results indicated that the random forest (RF) algorithm performed better, with r and RMSE, respectively, of 0.91 and 1.9 g.kg−¹ for LNC, and 0.86 and 0.17 m for PH. It was also demonstrated that VIs contributed more to the algorithm’s performances than individual spectral bands. This study concludes that the RF model is appropriate to predict both agronomic variables in maize and may help farmers to monitor their plants based upon their LNC and PH diagnosis and use this knowledge to improve their production rates in the subsequent seasons.

High resolution leaf spectral signature as a tool for foliar pigment estimation displaying potential for species differentiation

Optical leaf profiles depend on foliar pigment type and content, as well as anatomical aspects and cellular ultrastructure, whose effects are shown in several species. Monocotyledon and Dicotyledon plants presenting natural pigment content variations and anatomical alterations were analyzed. Each plant species displays its own spectral signatures, which are, in turn, influenced by foliar pigment class (composition) and concentration, as well as anatomical and ultrastructural plant cell characteristics. Plants with no anthocyanin displayed increased reflectance and transmittance in the green spectral region (501-565 nm), while values decreased in the presence of anthocyanin. At wavelengths below 500 nm (350-500 nm), strong overlapping signatures of phenolics, carotenoids, chlorophylls, flavonoids and anthocyanins were observed. Using a partial least squares regression applied to 350-700 nm spectral data allowed for accurate estimations of different foliar pigment levels. In addition, a PCA and discriminant analysis were able to efficiently discriminate different species displaying spectra overlapping. The use of absorbance spectra only was able to discriminate species with 100 % confidence. Finally, a discussion on how different wavelengths are absorbed and on anatomical interference of light interaction in leaf profiles is presented.

A Machine Learning Framework to Predict Nutrient Content in Valencia-Orange Leaf Hyperspectral Measurements

This paper presents a framework based on machine learning algorithms to predict nutrient content in leaf hyperspectral measurements. This is the first approach to evaluate macro- and micronutrient content with both machine learning and reflectance/first-derivative data. For this, citrus-leaves collected at a Valencia-orange orchard were used. Their spectral data was measured with a Fieldspec ASD FieldSpec® HandHeld 2 spectroradiometer and the surface reflectance and first-derivative spectra from the spectral range of 380 to 1020 nm (640 spectral bands) was evaluated. A total of 320 spectral signatures were collected, and the leaf-nutrient content (N, P, K, Mg, S, Cu, Fe, Mn, and Zn) was associated with them. For this, 204,800 (320 × 640) combinations were used. The following machine learning algorithms were used in this framework: k-Nearest Neighbor (kNN), Lasso Regression, Ridge Regression, Support Vector Machine (SVM), Artificial Neural Network (ANN), Decision Tree (DT), and Random Forest (RF). The training methods were assessed based on Cross-Validation and Leave-One-Out. The Relief-F metric of the algorithms’ prediction was used to determine the most contributive wavelength or spectral region associated with each nutrient. This approach was able to return, with high predictions (R2), nutrients like N (0.912), Mg (0.832), Cu (0.861), Mn (0.898), and Zn (0.855), and, to a lesser extent, P (0.771), K (0.763), and S (0.727). These accuracies were obtained with different algorithms, but RF was the most suitable to model most of them. The results indicate that, for the Valencia-orange leaves, surface reflectance data is more suitable to predict macronutrients, while first-derivative spectra is better linked to micronutrients. A final contribution of this study is the identification of the wavelengths responsible for contributing to these predictions.