IntechOpen

Home > Books > Grasslands - Conservation and Development [Working Title]

Open access peer-reviewed chapter - ONLINE FIRST

Determination of Grass Quality Using Spectroscopy: Advances and Perspectives

Written By

Manuela Ortega Monsalve, Tatiana Rodríguez Monroy, Luis Fernando Galeano-Vasco, Marisol Medina-Sierra and Mario Fernando Ceron-Munoz

Reviewed: 23 August 2023 Published: 14 November 2023

DOI: 10.5772/intechopen.112990

Grasslands - Conservation and Development IntechOpen
Grasslands - Conservation and Development Edited by Muhammad Aamir Iqbal

From the Edited Volume

Grasslands - Conservation and Development [Working Title]

Dr. Muhammad Aamir Iqbal

Chapter metrics overview

48 Chapter Downloads

View Full Metrics
Advertisement

Abstract

Spectroscopy is a promising technique for determining nutrients in grasses and may be a valuable tool for future research. This chapter reviews research carried out in recent years, focusing on determining the quality of grasses using spectroscopy techniques, specifically, spectrophotometry. The chemical methods used to determine the nutritional quality of grasses produce chemical residues, are time-consuming, and are costly to use when analyzing large crop extensions. Spectroscopy is a non-destructive technique that can establish the nutritional quality of grass easily and accurately. This chapter aims to describe the techniques focused on the use of spectroscopy and machine learning models to predict and determine the quality of grasses. A bibliographic review was conducted and recent research articles were selected that showed spectroscopic techniques applied to grasses. Different methods and results focusing on the quality of the grasses were compiled. In general, this review showed that the most commonly used spectroscopic method is near-infrared analysis. Spectroscopy is a very effective tool that opens the way to new types of technologies that can be applied to obtain results in determining the quality of pastures, leaving behind the use of traditional methods that represent higher costs and disadvantages compared to traditional methods based on precision agriculture.

Keywords

  • electromagnetic spectrum
  • nutrient prediction
  • precision agriculture
  • sustainable production
  • vegetation index

1. Introduction

Grasses and forages are essential for feeding herbivorous animals, especially ruminants that produce milk and meat. In recent years, the growth of the world’s population and the resulting need to increase animal food production, have forced grass production to generate a greater amounts of biomass and to seek new technologies that maximize production efficiency. Therefore, it’s important to remember that for optimal grass crop production, the soil on which these crops are grown must have good nutrient levels and stable physical characteristics. Therefore, soil characteristics can directly affect the quality and productivity of grass crops. For example, soils that are deficient in one or two nutrients will experience a decrease in grass production [1].

The most important step in determining soil or grass quality is to know the amount of nutrients present. The best way to obtain this information is through a chemical analysis, which indicates nutrient excesses or deficiencies and allows for the development of appropriate fertilization programs [2]. However, the methods used to determine the chemical properties require time, skilled labor, and the use of chemical reagents that are contaminated and dangerous [3]. In addition, according to [4], laboratory analysis presents a number of challenges, such as the analysis of a reduced number of samples and that it can only be performed in the laboratory by destroying the original samples. It also requires a specialized laboratory, a demanding task that is not practical for people working in the field [5]. For this reason, technological alternatives are being considered that are capable of predicting grass properties without generating negative impacts on the environment, and that can be used to control the growth and development of grasses [6]. New calibration techniques and laboratory and remote sensing (RS) belonging to the spectroscopy can predict grasses components without chemical analysis [7].

This chapter aims to describe the techniques focused on the use of spectroscopy and machine learning models that allow the prediction and determination of the quality of grasses. The collection of information on investigative articles was carried out from January 2022 to March 2023 in the databases Google Scholar, ScienceDirect, and SpringerLink. The search terms used were: grasses, soil, spectroscopy, nutrient prediction, agriculture, vegetation index, and machine learning models. Research articles that had a clear methodology and results of using the technique were selected, collecting important information from each of them.

Advertisement

2. Use of spectroscopy in grasses

Spectroscopy captures different portions of the electromagnetic spectrum (ES). The light variations in the ES regions cover the visible spectrum (VIS) consisting of bands like red (R), green (G), and blue (B) and the infrared (IR) region which is divided into the near-infrared (NIR), short wave infrared (SWIR), medium wave infrared (MWIR) and long wave infrared (LWIR) spectra as shown in Figure 1. The VIS and IR region of the ES are the most used equipment that detects pasture quality by spectroscopy.

Figure 1.

Visible and infrared regions of the electromagnetic spectrum. Adapted to [8].

Among the spectroscopic techniques is spectrophotometry, a branch based on the use of different parts of the electromagnetic spectrum and the presence of light to detect different elements according to their chemical composition. Among the methods using spectrophotometry we can find cameras with NIR, RGB, multispectral (MSI), hyperspectral (HSI) and RS bands. NIR works with vibrations [9, 10]; and according to [11], this technique has gained attention because it’s widely used in different areas of science, proving to be successful in obtaining information that is difficult to obtain with competing techniques. Now, spectroscopy uses the MWIR zone, which is known as a vibrational method that uses the light emitted in a sample to measure the transmission, absorption and reflection of light. This technique detects the structures of the sample as the molecules have chemical bonds that generate vibrational energy [12, 13]. In the same way, HSI detects signals or contiguous bands that contain a very narrow spectral bandwidth [14], greatly facilitate the detection of sample elements [15]. However, the use of HSI has not been sufficiently explored due to it’s high acquisition cost and the large amount of spectral data that must be processed [16, 17]. The difference between these methods is the range of the spectrum they use to determine different types of features related to the spectral bands.

Cameras that record the spectrum’s bands can be mounted on unmanned aerial vehicles (UAVs) or placed permanently in a laboratory to record data from samples. UAVs are widely used to collect data in the field due to their simplicity of operation and on-site data collection. For this reason, [7] used UAVs with HSI cameras to evaluate quality characteristics in grasses and demonstrated that this method can predict forage quality parameters.

Figure 2 shows two ways to analyze the quality of grasses through different forms that use spectrophotometry.

Figure 2.

Different forms to apply spectroscopy to determine the quality of grasses: (a) hyperspectral images in the laboratory and (b) UAV drone with RGB, multispectral or hyperspectral moving cameras.

The components of a sample can be analyzed from great distances using satellites; this is known as RS. RS has been used to study the spectral, spatial, and temporal variations of electromagnetic waves, and revealed the correlations between them and the proprieties of different terrestrial materials [18]. In other words, it focuses on identifying the materials present on the earth’s surface and on understanding the phenomena that occur on it through their spectral signature, that is, it seeks to establish a relationship between the properties of the light reflected or emitted by terrestrial objects and the intrinsic properties of these objects [19]. To perform this type of data analysis, RS analysis and processing techniques are used, using satellites, to capture electromagnetic radiation at different wavelengths [20, 21]. All of this provides valuable information about the composition and dynamics of the land surface allowing producers to detect crop variations and facilitate a more precise application of fertilizers [22] nutritional composition and general condition of plants.

It’s important to note that plants exhibit dynamic behavior in ES due to phenological changes in the plant, lighting conditions associated with topography (slope and orientation), sun position over the year, and soil moisture conditions, which can result in significant variations in the spectral response pattern [23].

The spectral signature of plants is characterized by how they reflect light in the visible spectrum (400–700 nm). Chlorophyll absorbs strongly in the visible spectrum around the B and R bands, resulting in a low reflectance, and the G spectral band, the reflectance is higher. It’s in this region of the spectrum that photosynthesis occurs, a process in which energy is absorbed [24]. On the other hand, in the NIR region, the vegetation shows its highest reflectance in the NIR between 700 and 1300 nm, around 45–50% (Figure 3). This phenomenon is the result of diffusion caused by the refractive indices of the intracellular fluid and the intercellular spaces present in the plant mesophyll, especially in the spongy mesophyll. In contrast, at wavelengths between 1300 and 2500 nm, the reflectivity of the sheet it’s controlled by water absorption, resulting in reflectance values of 10–20% [18]. When the vegetation is mature or under stress due to disease, insect attack, or low humidity, changes in the spectral characteristics of the leaves occur [25]. In general, these changes occur simultaneously in the VIS and IR regions, but there are greater alterations in IR. This behavior explains the great utility of ES for the study of vegetation.

Figure 3.

Changes in the spectral signature of healthy and diseased vegetation in the visible and infrared portions of the electromagnetic spectrum. Adapted to [19].

Figure 3 shows the spectral signature of the vegetation under different phenological conditions. It can be observed that the reflectance is influenced by the concentration of pigments in the leaves, mainly chlorophyll and carotenoids. In the VIS region of the spectrum, reflectance and transmittance are low due to strong absorption by leaf pigments.

2.1 Vegetation indices in grasses

From the study of spectroscopy, numerous vegetation indices (VI) have been developed. These indices are defined as parameters calculated using reflectance values at different wavelengths, to extract information related to vegetation [26]. In this context, VI is generated by mathematical calculations involving the different spectral bands of the images. These calculations produce a new image that highlights specific characteristics related to the physiological functioning of the plants [27]. The VI defined so far have in common the use of reflectance values in the R and NIR spectral bands, which are dedicated to the spectral behavior in this region of the ES. The reflectance of the vegetation goes from a relative minimum in the R corresponding to the absorption band of chlorophyll to an absolute maximum in the NIR, which is a consequence of the multiple scattering of the radiation within the cellular structure [25].

One of the most commonly used indices in vegetation assessment is the normalized difference vegetation index (NDVI), which quantifies the relationship between the energy absorbed and emitted by terrestrial objects [28]. In addition to NDVI, other indices relate to biomass and leaf area index. These include the green normalized difference vegetation index (GNDVI), which uses the G band instead of the R-band of the spectrum. There is also the RATIO index, which relates the high NIR reflectance of vegetation to low R reflectance. The enhanced vegetation index (EVI) is an enhancement of NDVI designed to perform better in areas of dense vegetation and to reduce atmospheric effects by being sensitive to variations in vegetation cover and leaf area. Similarly, the adjusted ground vegetation index (SAVI) is an enhancement of NDVI that attempts to compensate for the effects of ground brightness by relating the nutrients and R bands [29, 30]. Another index used in grasses is the normalized pigmented chlorophyll ratio index (NPCI) [31]. The VI is useful in determining productive characteristics in grasses, such as height, presence of disease, or available quantity.

Table 1 shows the most commonly used vegetation indices used to evaluate biomass production and nutritional content of grasses, and the results obtained when applying these indices. They are mainly used to determine the productive properties that can directly influence the quality of the plants and their use in animal feed.

Vegetation indexFormulaCharacteristics evaluatedAuthor
NDVINIR − R/NIR + RType of grass and amount[32]
RATIONIR/RBiomass and nitrogen[33]
GNDVINIR − G/NIR + GDifferentiation between soil and grasses[34]
NDRENIR − RE/NIR + REAmount and chlorophyll[35]
SAVINIR − R/NIR + R + 0.5Amount and cover[36, 37]
OSAVI1.6 * (NIR − R)/(NIR + R + 0.16)Biomass[38]
EVI2.5 * (NIR−R)/(NIR + 6 * R − 7.5 * B) + 1Biomass and cover[39]

Table 1.

Vegetation index, formulas, constants and characteristics were evaluated to predict the quality of pastures through spectroscopic techniques.

Vegetation index: NDVI, normalized difference; RATIO, simple ratio indices; GNDVI, green normalized difference; NDRE, normalized difference red edge; SAVI, soil adjusted; OSAVI, optimized soil adjusted; EVI, enhanced; R, red; G, green; B, blue; NIR, near infrared; RE, red edge.

Among the results found to evaluate the quality and nutritional content of forages by spectroscopy, the work of [40] stands out. The reason is that this study has found a significant correlation between forage yield and RGB bands based on VI. This positive relationship demonstrates the usefulness of VI derived from color imagery for estimating forage quality and nutrient content. Soil chemical characteristics were also evaluated as an indirect measure of grass quality. One study found that soil nutrient content, especially potassium and phosphorus, was strongly related to electrical conductivity and NDVI [41]. These results demonstrate the relationships between soil and crop characteristics and allow the optimization of variable rate fertilization.

Likewise, Fava et al. [33] evaluated biomass and nitrogen status in grasses at three different growth stages and in grazed and ungrazed plots, finding good results for assessing nitrogen content, green biomass, and leaf area index using the NIR (775–820 nm) and longer wavelengths of the red edge (740–770 nm). Another report shows the feasibility of using NDVI to assess moisture content in forages.

2.2 Quality characteristics in grasses determined through spectroscopic techniques

In general, the quality of grasses is measured using various characteristics such as dry matter, digestibility, energy, organic matter, protein, and carbohydrates [42, 43]. Also, by many VI, factors affecting the quality of grasses, such as plant diseases, amount of biomass, or type of fertilization, are directly related to the nutritional characteristics (Figure 4). In addition, the quality of the grasses can also be measured indirectly, through the content of soil nutrients. In the soil matrix, the content of sand, silt, and clay can be monitored by spectroscopy [44, 45], the content of organic carbon and organic matter, which are synonymous with soil fertility and therefore, of plant growth [46, 47]. However, several studies have managed to determine using spectroscopy many characteristics focused on the productivity of grasses.

Figure 4.

Main nutritional and production quality characteristics evaluated in grasses.

Results of previous studies agree that the use of spectroscopy in forages can be useful to determine factors such as mass and growth, light interception, turf heterogeneity, nitrogen deficiency, and drought stress, using non-destructive methods [48]. The determination of the concentration of nutrients such as phosphorus, potassium, sulfur, calcium, and magnesium, some minor elements. and the sugar content of grasses has also been studied using spectroscopic techniques [49]. In addition, spectroscopy can work in other areas indirectly related to the quality of grasses and plants, for example, [50] they worked on forest dynamics and land use; in addition [51] applied spectroscopic techniques to determine the quality of the milk from cattle fed with fresh grasses and [52] evaluated carcasses of lambs that were fed with fresh grasses. This is due to the fact that there are many characteristics, both direct and indirect, that can be evaluated in grasses to determine their quality. This can open new avenues for future research and can clarify that these techniques, belonging to the new PA, are widely used in the investigative field of vegetation.

2.3 Investigations carried out in grasses using spectroscopic techniques

A study conducted by [53] demonstrates the implementation of portable spectroscopic sensors to determine the nutrient content in grasses, finding good results in determining height and protein content, which can lead to improved productivity for farmers through access to this type of technology. In this review, it has been shown that NIR instruments are the most widely used in spectroscopy and many of the studies carried out in grasses focus on the use of this method, with good results. For example, Restaino et al. [54] used 120 grass samples analyzed by NIR to determine nutrients in grasses, demonstrating great potential in this technique; Danelli et al. [55] predicted grass quality parameters using NIR, using 1615 samples, managed to find optimal settings, and considered this technique as an aid in grass management; Catunda et al. [56] developed calibration strategies in grasses using NIR to predict their nutritional composition using 2622 samples, they found that characteristics such as ash, protein, and fiber can be easily determined using this methodology; Parrini et al. [57] evaluated the quality of fresh grasses using NIR, finding the estimation of some chemical parameters of grasses associated with quality feasible.

The NIR methodology has the advantage that it can be applied in the laboratory and in the field, for this reason, Serrano et al. [5] used grasses for continuous grazing to test the technique in the field, demonstrating that the combination of spectroscopic and ultrasonic methods to determine pasture quality factors can be accurate even in very heterogeneous grasses due to grazing animal. In this review, 95% of the studies carried out have included NIR analysis in the laboratory or through UAVs, where methodologies that include the use of HSI and multispectral images can also be observed, while the other percentage have used techniques that include only RGB bands. This leads to the idea that studies of grasses have evolved to include not only the visible part of the spectrum but also using the NIR region, which provides more information due to its greater number of spectral bands.

Table 2 summarizes the information found by different authors on the use of spectroscopy in grasses to determine their quality, including parameters such as the method used, the spectral characteristics of the method, and the results obtained in this review.

StudySpectral rangeResults
Techniques in real time and monitor the changes in the nutrients of the different types of grasses with NIR portable [53]950–1750 nmHigh correlation between grass height and some nutrients
Potential data for grass quality assessment using UAV with HSI in field [7]350–2500 nmModels with good predictive potential for CP and fibers
Spectroscopy for predicting the nutritional value of forage silages with NIR [54]400–2500 nmPotential to predict DM, CP and fibers
Calibrations for predicting fresh grass quality parameters with NIR [55]1100–2500 nmOptimal predictions for fibers and ash variables
The best calibrations for a grassland sample collection with NIR [5]1100–2500 nmAdjustments greater than 80% for pasture nutritional variables
Predictive models for estimating the chemical composition of Brachiaria with NIR [58]400–2500 nmNutrient variables were estimated with higher prediction in dry grasses
Use canopy spectra with NIR portable to predict parameters for individual ryegrass plants [59]300–2500 nmIt was possible to perform a classification of the plants using of VI
Predictive models to estimate different concentrations at the field level, using spectroscopy data from grass images using UAV with HSI [60]450–998 nmGeneration of predictive models for the CP and fibers variables and their variations in space
Performance of DL-based segmentation algorithms in the context of crop segmentation approaches using UAV with RGB [61]It was possible to determine different compositions of plants using models
Design and development of a portable using UAV with MSI for grassland vegetation growth monitoring system [6]450, 550, 650, 750, 850, and 960 nmThe growth of different pastures was predicted using VI

Table 2.

Compilation of different spectral methodologies and characteristics of the equipment used in spectroscopy applied to grasses.

SR, spectral range; HSI, hyperspectral; CP, crude protein; DM, dry matter; DL, deep learning; MSI, multispectral; UAV, unmanned aerial vehicles.

2.4 Machine learning models and predictive metrics applied to grass quality

In the area of measuring the intensities of reflected light in narrow spectral bands, particularly in the NIR range, advances in optical methods have enabled the identification of chemical bonds between hydrogen and carbon, hydrogen and nitrogen, and hydrogen and oxygen. These chemical bonds have proven to be key indicators for determining the nutritional characteristics of grass, including the presence of crude protein, fibers, and other plant constituents [57, 59, 62]. These new technological alternatives, along with the use of machine learning statistical models, can produce good results in predicting various nutrient variables in soils and grasses. According to [63], machine learning algorithms have great potential for analyzing spectral data and analyzing attributes in grasses. Statistical models perform the function of developing, evaluating, and improving the equations that calibrate and predict the presence of elements or nutrients from reflectance spectra, replacing traditional chemical techniques [64].

Knowing the relationship between the spectral bands and the grass variables requires the use of multivariate statistics [65], so several statistical models have been developed. The most commonly used is the partial least squares regression (PLSR), which explores the relationships that may exist between all the variables [66]. The cubist model (CUB) is also used to predict nutrients using spectroscopy. This algorithm performs the construction of an unconventional regression tree [67], the prediction is made based on intermediate linear models step by step, and it creates subsets of data and rules to select only some predictor variables [68]. On the other hand, the random forest (RF) model works with many decision trees that are known as classifiers; these use representative variables of the sample, forming classifier nodes for the set of variables that have multivariate characteristics, in the case of the spectral bands of each sample [69].

In addition, the support vector machine (SVM) algorithm, using the hyperplane, recognizing different categories, and maximum margins, manages to separate the data into different categories by selecting the most appropriate vectors for the prediction [70]. However, there are many algorithms that can be applied to spectroscopic data. Others widely used are principal component regression (PCA), multiple linear regression (MLR), and artificial neural networks (ANN) [71, 72]. To evaluate the predictive performance of the above machine learning models, [73] considers that some of the most commonly used metrics in these models are the coefficient of determination (R2), the ratio of performance to deviation (RPD), and the root-mean-square error (RMSEP). It’s important to obtain models that have R2 values close to 1, low RMSEP values since the difference between the predicted and observed values of the model should be small, and RPD values greater than 2 [74]. These model selection parameters will indicate that the selected model can efficiently predict grass quality properties using spectroscopic methods.

In reviewing the selected articles, it was found that 84% of the articles belonging to this review used the PLSR machine learning model for data analysis. The remaining percentage includes PCA, SVM, and CUB models. The widespread use of the PLSR model for predicting quality variables using spectroscopy may be due to its ease of execution in various statistical software, in addition to the fact that it is a model that reduces the dimensionality of the data, and can reduce memory requirements.

For this reason, the PLSR models used in the various studies collected in this review presented optimal fits in predicting different characteristics in the grasses, providing high reliability for use in future studies. For example, [75] used this model to predict crude protein (CP) in grasses and found an R2 of 0.98 and an RPD of 4.12, indicating a high fit of the model found. Results of a R2 were also found to predict the CP in the grass of 0.69 and 0.73 respectively, but the RPD found was 1.95, which means that the performance of the model for this feature is regular [76, 77]. Another similar result was found by [78], where the RPD found was 1.84, the R2 was 0.77 and the RMSEP was 2.05. Finally, for the CP variable, [5] found an R2 of 0.84 and an RPD of 4. According to all the results related to this nutritional variable of the grass, the CP can be predicted with PLSR models with a high fit, but many samples must be evaluated to make the model more and more fit. The fibers variable was also been predicted using PLSR models, [75] finding an R2 of 0.94, an error of 2.94, and an RPD of 2.83. For the same variable, R2 of 0.67 and 0.76 and RMSEP of 5.77 and 4 were obtained [7, 56]. This variable was predicted using RS and found a fit of 0.66 and an RPD of 1.71 [79]. This variable resulted in minor adjustments to the CP variable but has great potential to be predicted using spectroscopy and machine learning models.

Regarding other variables that have been predicted in grasses, [55] managed to find an R2 for the ash characteristic in the grass of 0.75 and an error of 1.01, which means that this technique allows one to know the amount of minerals in the grass. Also, [80] found a good fit for the nutrient variable EE in the grass, finding an R2 of 0.73, an RDP of 1.69, and an RMSEP of 0.23. The CP, fibers, and dry matter variables are the most analyzed by researchers and the CP variable is the one that shows the highest adjustment and prediction results in the different evaluation criteria of the models. However, the other nutritional variables also present good adjustment results.

Regarding the productive variables of grasses, VI has also been widely used in spectroscopy to determine nutritional and productive variables, mainly in remote sensing and moving cameras. In general, the variables related to the quality of the pasture, evaluated using spectroscopy and statistical analysis, have generated optimal results in terms of their adjustment, which indicates that the use of these techniques is a great tool for knowing the quality of the grass, before being offered as food to cattle. Regarding the statistical models, although SVM and CUB have been used in this type of study, they are not preferred by researchers, and it is then, the PLSR model is the most used. The NIR methodology is the most used to carry out research on grass quality due to its wide range in the electromagnetic spectrum, NIR the interaction between plant compounds can be observed much easier, while equipment that only RGB cameras or multispectral may be limited in predicting plant properties due to few bands and little information available. This is due to the great simplicity of programming and interpretation of the results, as it reduces the dimensionality of the spectral data. Table 3 shows a summary of the studies carried out on the grasses, the statistical models applied and the fitting values found in each one of them.

ParameterStatistical modelR2RMSERPDAuthor
Crude proteinPLSR0.792.48—[7]
0.90—2.37[81]
0.882.142.26[57]
0.841.624.00[5]
0.901.483.1[82]
0.830.642.4[80]
0.772.051.84[78]
PCA0.82—1.88[58]
RF0.702.95—[7]
Dry matterPLSR0.96—2.60[81]
0.872.702.75[57]
0.941.972.7[82]
0.57—1.51[58]
FibersPLSR0.606.34—[7]
0.764.001.54[57]
0.892.472.90[82]
0.303.321.1[82]
0.553.231.50[80]
0.616.981.90[5]
RF0.526.98—[7]
PCA0.82—2.23[61]
AshPLSR0.641.051.87[58]
0.751.011.2[82]
0.710.631.87[80]
Ether extractPLSR0.450.281.18[57]
0.650.241.4[82]
0.730.231.69[80]
LigninePLSR0.731.831.61[57]

Table 3.

Nutritional characteristics evaluated in grasses using spectroscopy.

PLSR, partial least squares; PCA, principal component analysis; RF, random rorest; R2, determination coefficient; RMSEP, root-mean-square error; RPD, ratio of performance to deviation.

Advertisement

3. Conclusions

Spectroscopy has been widely used in recent years to evaluate the quality of grasses. This is confirmed by the large amount of research supporting its use and demonstrating its effectiveness in predicting productive and nutritional variables in grasses. Both laboratory and field methods can be easily applied, and the use of machine learning statistical models to predict variables related to quality has promoted its use in agricultural areas. In line with the rapid advances that have occurred in the field of precision agriculture, it is expected that the development and use of spectroscopic techniques to determine grass quality will increase in the coming years due to the need to limit the use of chemical analysis and promote non-destructive methods in the field. The development and widespread use of more accessible machine learning statistical models and portable equipment is expected to enable the collection and analysis of real-time information in the field to facilitate the work of graziers.

Advertisement

Acknowledgments

This publication and the financial support of the MSc students was possible thanks to the project “Design and validation of predictive models to determine Cation Exchange Capacity (CEC), Organic Matter (OM) and Nitrogen (N) in soils from hyperspectral images” through the agreement 2022-7204, financed by the University of Antioquia Foundation.

Advertisement

Conflict of interest

The authors declare no conflict of interest.

Advertisement

List of abbreviations

NIR

near infrared

RS

remote sensing

ES

electromagnetic spectrum

VIS

visible spectrum

IR

infrared

R

red

G

green

B

blue

RE

red edge

SWIR

short wave infrared

MWIR

medium wave infrared

LWIR

long wave infrared

UAVs

unmanned aerial vehicles

HSI

hyperspectral

MSI

multispectral

VI

vegetation index

NDVI

normalized difference vegetation index

GNDVI

green normalized difference vegetation index

RATIO

simple ratio index

NDRE

normalized difference red edge

EVI

enhanced vegetation index

SAVI

adjusted ground vegetation index

NPCI

chlorophyll ratio index

PLSR

partial least square regression

RF

random forest

SVM

support vector machine

CUB

cubist model

R2

coefficient of determination

RMSEP

root-mean-square error

RPD

ratio of performance to deviation

CP

crude protein

DM

dry matter

References

  1. 1. Silveira ML, Kohmann MM. Maintaining soil fertility and health for sustainable pastures. In: Management Strategies for Sustainable Cattle Production in Southern Pastures. Cambridge, Massachusetts: Academic Press; 2020. pp. 35-58. DOI: 10.1016/b978-0-12-814474-9.00003-7
  2. 2. Westerman RL. Soil Testing and Plant Analysis. 3rd ed. Vol. 152, No. 2. Madison: Soil Science; 1991. p. 137
  3. 3. Molano ML. Caracterización nutricional de forrajes tropicales usando espectroscopía de infrarrojo cercano (nirs). Bogotá, Colombia: Universidad Nacional de Colombia; 2013
  4. 4. Teye E, Amuah CL, Atiah K, Darko RO, Abindaw T, Amoah KK, et al. Quick determination of soil quality using portable spectroscopy and efficient multivariate techniques. Journal of Spectroscopy. 2023;2023:11. DOI: 10.1155/2023/2024318
  5. 5. Serrano J, Shahidian S, Carreira E, Nogales-Bueno J, Rato AE. Predicting the evolution of pasture quality by NIRS: Perspectives for real-time pasture and grazing management. In: Online AgEng2021 Conference, 5–8 July 2021. Évora, Portugal; 2021. pp. 1-8
  6. 6. Zhang A, Hu S, Zhang X, Zhang T, Li M, Tao H, et al. Ahandheld grassland vegetation monitoring system based on multispectral imaging. Agriculture. 2021;11(12):12-62. DOI: 10.3390/agriculture11121262
  7. 7. Fernández-Habas J, Cañada MC, Moreno AM, Leal-Murillo JR, González-Dugo MP, Oar BA, et al. Estimating pasture quality of Mediterranean grasslands using hyperspectral narrow bands from field spectroscopy by random forest and PLS regressions. Computers and Electronics in Agriculture. 2022;192:106614. DOI: 10.1016/j.compag.2021.106614
  8. 8. Akhloufi M, Bendada A. Fusion of active and passive infrared images for face recognition. Proceedings of SPIE - The International Society for Optical Engineering. 2013;8705:84-93. DOI: 10.1117/12.2017942
  9. 9. Huck CW, Bec KB. Advances in near-infrared spectroscopy and related computational methods. Molecules. 2019;24(23):43-70. DOI: 10.3390/books978-3-03928-053-7
  10. 10. Czarnecki MA, Morisawa Y, Futami Y, Ozaki Y. Advances in molecular structure and interaction studies using near-infrared spectroscopy. Chemical Reviews. 2015;115(18):9707-9744. DOI: 10.1021/cr500013u
  11. 11. Bec KB, Grabska J, Huck CW. Near-infrared spectroscopy in bio-applications. Molecules. 2020;25(12):29-48. DOI: 10.3390/molecules25122948
  12. 12. Coates J. Interpretation of infrared spectra, a practical approach. In: Meyers RA, McKelvy ML, editors. Encyclopedia of Analytical Chemistry: Applications, Theory and Instrumentation. Chichester, England; 2000. DOI: 10.1002/9780470027318.a5606
  13. 13. Amirvaresi A, Nikounezhad N, Amirahmadi M, Daraei B, Parastar H. Comparison of near-infrared (NIR) and mid-infrared (MIR) spectroscopy based on chemometrics for saffron authentication and adulteration detection. Food Chemistry. 2021;344:128647. DOI: 10.1016/j.foodchem.2020.128647
  14. 14. Randelović P, Dorević V, Milić S, Balesević-Tubić S, Petrović K, Miladinović J, et al. Prediction of soybean plant density using a machine learning model and vegetation indices extracted from RGB images taken with a UAV. Agronomy. 2020;10(8):1108. DOI: 10.3390/agronomy10081108
  15. 15. Lu B, Dao PD, Liu J, He Y, Shang J. Recent advances of hyperspectral imaging technology and applications in agriculture. Remote Sensing. 2020;12(16):26-59. DOI: 10.3390/rs12162659
  16. 16. Darvishzadeh R, Matkan AA, Ahangar AD. Inversion of a radiative transfer model for estimation of rice canopy chlorophyll content using a lookup-table approach. Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2012;5(4):1222-1230. DOI: 10.1109/JSTARS.2012.2186118
  17. 17. Lodhi V, Chakravarty D, Mitra P. Hyperspectral imaging system: Development aspects and recent trends. Sensing and Imaging. 2019;20(35):1-24. DOI: 10.1007/s11220-019-0257-8
  18. 18. Chuvieco E. Fundamentos de teledetección espacial. Estudios Geográficos. 1991;52(203):371
  19. 19. Mena FC. Tecnologías geoespaciales: experiencias aplicadas al estudio y gestión del territorio. 2009
  20. 20. Paoletti ME, Haut JM, Plaza J, Plaza A. Estudio comparativo de técnicas de clasificación de imágenes hiperespectrales. Revista Iberoamericana de Automática e Informática Industrial. 2019;16(2):129-137. DOI: 10.4995/riai.2019.11078
  21. 21. Mantripragada K, Dao PD, He Y, Qureshi FZ. The effects of spectral dimensionality reduction on hyperspectral pixel classification: A case study. PLoS One. 2022;17(7):0269174. DOI: 10.1371/journal.pone.0269174
  22. 22. Su WH. Advanced machine learning in point spectroscopy, RGB-and hyperspectral-imaging for automatic discriminations of crops and weeds: A review. Smart Cities. 2020;3(3):767-792. DOI: 10.3390/smartcities3030039
  23. 23. Houborg R, Fisher JB, Skidmore AK. Advances in remote sensing of vegetation function and traits. International Journal of Applied Earth Observation and Geoinformation. 2015;43:1-6. DOI: 10.1016/j.jag.2015.06.001
  24. 24. Gausman HW. The leaf mesophylls of twenty crops, their light spectra, and optical and geometrical parameters. 1972. 10.22004/ag.econ.160035.
  25. 25. Kharuf-Gutierrez S, Hernández S, Orozco R, Aday OC, Delgado I. Análisis de imágenes multiespectrales adquiridas con vehículos aéreos no tripulados. Ingeniería Electrónica, Automática y Comunicaciones. 2018;39(2):79-91
  26. 26. Tucker CJ. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment. 1979;8(2):127-150. DOI: 10.1016/0034-4257(79)90013-0
  27. 27. Gilabert MA, González-Piqueras J, García-Haro J. Acerca de los índices de vegetación. Revista de Teledetección. 1997;8:1-10
  28. 28. Pettorelli N. The Normalized Difference Vegetation Index. USA: Oxford University Press; 2014. p. 224. DOI: 10.1093/acprof:osobl/9780199693160.001.0001
  29. 29. Bannari A, Morin D, Bonn F, Huete AR. A review of vegetation indices. Remote Sensing Reviews. 1995;13(1–2):95-120. DOI: 10.1080/02757259509532298
  30. 30. Rondeaux G, Michael S, Baret F. Optimization of soil-adjusted vegetation indices. Remote Sensing of Environment. 1996;55(2):95-107. DOI: 10.1016/0034-4257(95)00186-7
  31. 31. Tao H, Feng H, Xu L, Miao M, Largo H, Yue J, et al. Estimation of crop growth parameters using UAV-based hyperspectral remote sensing data. Sensors. 2020;20(5):1296. DOI: 10.3390/s20051296
  32. 32. Rouse J, Haas R, Schell J, Deering D. Monitoring vegetation systems in the Great Plains with ERTS. NASA Special Publication. 1974;351(1):309-317
  33. 33. Fava F, Colombo R, Bocchi S, Meroni M, Sitzia M, Fois N, et al. Identification of hyperspectral vegetation indices for Mediterranean pasture characterization. International Journal of Applied Earth Observation and Geoinformation. 2009;11(4):233-243. DOI: 10.1016/j.jag.2009.02.003
  34. 34. Huete AR. A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment. 1988;25(3):295-309
  35. 35. Barnes E, Clarke T, Richards S, Colaizzi P, Haberland J, Kostrzewski M, et al. Coincident detection of crop water stress, nitrogen status and canopy density using ground based multispectral data. In: Proceedings of the Fifth International Conference on Precision Agriculture; Bloomington, USA. 2000
  36. 36. Daughtry C, Walthall C, Kim M, De Colstoun EB, McMurtrey IJ. Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance. Remote Sensing of Environment. 2000;74(2):229-239. DOI: 10.1016/S0034-4257(00)00113-9
  37. 37. Qi J, Chehbouni A, Huete AR, Kerr YH, Sorooshian S. A modified soil adjusted vegetation index. Remote Sensing of Environment. 1994;48(2):119-126. DOI: 10.1016/0034-4257(94)90134-1
  38. 38. Boschetti M, Bocchi S, Brivio PA. Assessment of pasture production in the Italian Alps using spectrometric and remote sensing information. Agriculture, Ecosystems and Environment. 2007;118(1–4):267-272. DOI: 10.1016/j.agee.2006.05.024
  39. 39. Jiang Z, Huete AR, Didan K, Miura T. Development of a two-band enhanced vegetation index without a blue band. Remote Sensing of Environment. 2008;112(10):3833-3845. DOI: 10.1016/j.rse.2008.06.006
  40. 40. Lussem U, Bolten A, Gnyp ML, Jasper J, Bareth G. Evaluation of rgb-based vegetation indices from UAV imagery to estimate forage yield in grassland. Developments, Technologies and Applications in Remote Sensing. 2018;42:1215-1219. DOI: 10.5194/isprs-archives-XLII-3-1215-2018
  41. 41. Mazur P, Gozdowski D, Wójcik-Gront E. Electrical conductivity of the soil and vegetation indices derived from satellites for the evaluation of the content of phosphorus, potassium and magnesium, pH and delimitation of management zones within the field. Agriculture. 2022;12(2):354. DOI: 10.3390/agricultura12060883
  42. 42. O’Donovan M, Lewis E, O’ Kiely P. Requirements of future grass-based ruminant production systems in Ireland. Irish Journal of Agricultural and Food Research. 2011;50(1):1-21
  43. 43. Burns GA, Gilliland TJ, Grogan D, Watson S, O’ Kiely P. Assessment of herbage yield and quality traits of perennial ryegrasses from a national variety evaluation scheme. The Journal of Agricultural Science. 2013;151(3):331-346. DOI: 10.1017/S0021859612000251
  44. 44. Jaconi A, Cora V, Axel D. Near-infrared spectroscopy as an easy and precise method to estimate soil texture. Geoderma. 2019;337:906-913. DOI: 10.1016/j.geoderma.2018.10.038
  45. 45. Coblinski JA, Giasson É, Demattê JA, Dotto AC, Costa JJF, Vašát R. Prediction of soil texture classes through different wavelength regions of reflectance spectroscopy at various soil depths. Catena. 2020;189:104485. DOI: 10.1016/j.catena.2020.104485
  46. 46. Crucil G, Castaldi F, Aldana-Jague E, van Wesemael B, Macdonald A, Van Oost K. Assessing the performance of UAS-compatible multispectral and hyperspectral sensors for soil organic carbon prediction. Sustainability. 2019;11(7):1889. DOI: 10.3390/su11071889
  47. 47. Meng X, Bao Y, Liu J, Liu H, Zhang X, Zhang Y, et al. Regional soil organic carbon prediction model based on a discrete wavelet analysis of hyperspectral satellite data. International Journal of Applied Earth Observation and Geoinformation. 2020;89(1):102-111. DOI: 10.1016/j.jag.2020.102111
  48. 48. Zhou W, Yang H, Xie L, Li H, Huang L, Zhao Y, et al. Hyperspectral inversion of soil heavy metals in three-river source region based on random forest model. Catena. 2021;202:105222. DOI: 10.1016/j.catena.2021.105222
  49. 49. Schut AGT, Ketelaars JJMH. Monitoring growth of grass swards using imaging spectroscopy. Grass and Forage Science. 2003;58(3):276-286. DOI: 10.1046/j.1365-2494.2003.00379.x
  50. 50. Schut AGT, Lokhorst C, Hendriks MMWB, Kornet JG, Kasper G. Potential of imaging spectroscopy as tool for pasture management. Grass and Forage Science. 2005;60(1):34-45. DOI: 10.1111/j.1365-2494.2005.00449.x
  51. 51. Herpin U, Cerri CC, Carvalho MCS, Markert B, Enzweiler J, Friese K, et al. Biogeochemical dynamics following land use change from forest to pasture in a humid tropical area (Rondonia, Brazil): A multi-element approach by means of XRF-spectroscopy. Science of the Total Environment. 2002;286(1–3):97-109. DOI: 10.1016/S0048-9697(01)00967-6
  52. 52. Capuano E, Rademaker J, van den Bijgaart H, van Ruth SM. Verification of fresh grass feeding, pasture grazing and organic farming by FTIR spectroscopy analysis of bovine milk. Food Research International. 2014;60:234-241. DOI: 10.1016/j.foodres.2013.12.024
  53. 53. Bell MJ, Mereu L, Davis J. The use of mobile near-infrared spectroscopy for real-time pasture management. Frontiers in Sustainable Food Systems. 2018;2:76. DOI: 10.3389/fsufs.2018.00076
  54. 54. Restaino EA, Fernández EG, La Manna A, Cozzolino D. Prediction of the nutritive value of pasture silage by near infrared spectroscopy (NIRS). Chilean Journal of Agricultural Research. 2009;69(4):560-566
  55. 55. Danelli PP, Carlini P, Bernabucci U, Ronchi B. Quality evaluation of regional forage resources by means of near infrared reflectance spectroscopy. Italian Journal of Animal Science. 2016;3(4):363-376. DOI: 10.4081/ijas.2004.363
  56. 56. Catunda KL, Churchill AC, Power SA, Moore BD. Near infrared spectroscopy calibration strategies to predict multiple nutritional parameters of pasture species from different functional groups. Journal of Near Infrared Spectroscopy. 2022;30(5):254-263. DOI: 10.1101/2021.07.31.454175
  57. 57. Parrini S, Acciaioli A, Crovetti A, Bozzi R. Use of FT-NIRS for determination of chemical components and nutritional value of natural pasture. Italian Journal of Animal Science. 2019;17(1):87-91. DOI: 10.1080/1828051X.2017.1345659
  58. 58. Ribeiro MCA, Guerra GL, Cano C, Nóbrega de Carvalho L, Galbeiro S, Siqueira PR, et al. Prediction models of the nutritional quality of fresh and dry Brachiaria brizantha cv. Piatã grass by near-infrared spectroscopy. Journal of Applied Animal Research. 2023;51(1):193-203. DOI: 10.1080/09712119.2023.2172022
  59. 59. Smith C, Cogan N, Badenhorsh P, Spangenberg G, Smith K. Field spectroscopy to determine nutritive value parameters of individual ryegrass plants. Agronomy. 2019;9(6):1-14. DOI: 10.3390/agronomy9060293
  60. 60. Wijesingha J, Astor T, Schulze-Brüninghoff D, Wengert M, Wachendorf M. Predicting forage quality of grasslands using UAV-borne imaging spectroscopy. Remote Sensing. 2020;12(1):126. DOI: 10.3390/rs12010126
  61. 61. Kartal S. Comparison of semantic segmentation algorithms for the estimation of botanical composition of clover-grass pastures from RGB images. Ecological Informatics. 2021;66:101467. DOI: 10.1016/j.ecoinf.2021.101467
  62. 62. Foley WJ, McElwee A, Lawler I, Aragonés L, Woolnough AP, Berding N. Ecological applications of near-infrared reflectance spectroscopy a tool for rapid, cost-effective prediction of the composition of plant and animal tissues and aspects of animal performance. Oecologia. 1998;116(3):293-305. DOI: 10.1007/s004420050591
  63. 63. Verrelst J, Camps-Valls G, Muñoz J, Rivera JP, Veroustraete F, Clevers JG, et al. Optical remote sensing and the retrieval of terrestrial vegetation bio-geophysical properties- a review. Journal of Photogrammetry and Remote Sensing. 2015;108:273-290. DOI: 10.1016/j.isprsjprs.2015.05.005
  64. 64. Calle A, Salvador P. Revisando el concepto de resolución en teledetección. Revista de Teledetección. 2012;37:74-79
  65. 65. Geladi P. Chemometrics in spectroscopy. Part 1, classical chemometrics. Spectrochimica Acta Part B: Atomic Spectroscopy. 2003;58(5):767-782. DOI: 10.1016/S0584-8547(03)00037-5
  66. 66. Peng X, Shi T, Song A, Chen Y, Gao W. Estimating soil organic carbon using VIS/NIR spectroscopy with SVMR and SPA methods. Remote Sensing. 2014;6(4):2699-2717. DOI: 10.3390/rs6042699
  67. 67. Minasny B, McBratney AB, Stockmann U, Hong SY. Cubist a regression rule approach for use in calibration of NIR spectra. In: 16th International Conference on Near Infrared Spectroscopy 2–7 June 2013; France.
  68. 68. Morellos A, Pantazi XE, Moshou D, Alexandridis T, Whetton R, Tziotzios G, et al. Machine learning based prediction of soil total nitrogen, organic carbon and moisture content by using VIS-NIR spectroscopy. Biosystems Engineering. 2016;152:104-116. DOI: 10.1016/j.biosystemseng.2016.04.018
  69. 69. Zhang D, Zhang H, Zhao Y, Chen Y, Ke C, Xu T, et al. A brief review of new data analysis methods of laser-induced breakdown spectroscopy: Machine learning. Applied Spectroscopy Reviews. 2020;57(2):1-23. DOI: 10.1080/05704928.2020.1843175
  70. 70. Qi Y, Hu D, Jiang Y, Wu Z, Zheng M, Chen EX, et al. Recent progresses in machine learning assisted Raman spectroscopy. Advanced Optical Materials. 2023;11(14):1-22. DOI: 10.1002/adom.202203104
  71. 71. Shahin MA, Symons SJ. Detection of Fusarium damaged kernels in Canada Western Red Spring wheat using visible/near-infrared hyperspectral imaging and principal component analysis. Computers and Electronics in Agriculture. 2011;75(1):107-112. DOI: 10.1016/j.compag.2010.10.004
  72. 72. Su WH, Bakalis S, Sun DW. Potato hierarchical clustering and doneness degree determination by near-infrared (NIR) and attenuated total reflectance mid-infrared (ATR-MIR) spectroscopy. Journal of Food Measurement and Characterization. 2019;13:1218-1231. DOI: 10.1007/s11694-019-00037-3
  73. 73. Ahmadi A, Emami M, Daccache A, He L. Soil properties prediction for precision agriculture using visible and near-infrared spectroscopy: A systematic review and meta-analysis. Agronomy. 2021;11(433):1-14. DOI: 10.3390/agronomy11030433
  74. 74. Chang CW, Laird DA, Mausbach MJ, Hurburgh CR. Near-infrared reflectance spectroscopy–principal components regression analyses of soil properties. Soil Science Society of America Journal. 2001;65(2):480. DOI: 10.2136/sssaj2001.652480x
  75. 75. Massignani C, Vandresen BB, Marques JV, Kazama R, Osmari MP, Silva-Kazama DC. A single calibration of near-infrared spectroscopy to determine the quality of forage for multiple species. Research, Society and Development. 2021;10(10):1-10. DOI: 0.33448/rsd-v10i10.18990
  76. 76. Geipel J, Bakken AK, Jørgensen M, Korsaeth A. Forage yield and quality estimation by means of UAV and hyperspectral imaging. Precision Agriculture. 2021;22:1437-1463. DOI: 10.1007/s11119-021-09790-2
  77. 77. Thomson AL, Karunaratne SB, Copland A, Stayches D, Morse E, Jacobs J. Use of traditional, modern, and hybrid modeling approaches for in situ prediction of dry matter yield and nutritive characteristics of pasture using hyperspectral datasets. Animal Feed Science and Technology. 2020;269:114670. DOI: 10.1016/j.anifeedsci.2020.114670
  78. 78. Duranovich FN, Yule IJ, Lopez-Villalobos N, Shadbolt NM, Draganova I, Morris S. Using proximal hyperspectral sensing to predict herbage nutritive value for dairy farming. Agronomy. 2020;10(11):18-26. DOI: 10.3390/agronomy10111826
  79. 79. Safari H, Fricke T, Wachendorf M. Determination of fiber and protein content in heterogeneous pastures using field spectroscopy and ultrasonic sward height measurements. Computers and Electronics in Agriculture. 2016;123:256-263. DOI: 10.1016/j.compag.2016.03.002
  80. 80. Berauer BJ, Wilfahrt PA, Reu B, Schuchardt MA, Garcia-Franco N, Zistl-Schlingmann M, et al. Predicting forage quality of species-rich pasture grasslands using vis-NIRS to reveal effects of management intensity and climate change. Agriculture, Ecosystems and Environment. 2020;296:106929. DOI: 10.1016/j.agee.2020.106929
  81. 81. Murphy DJ, O’Brien B, O’Donovan M, Condon T, Murphy MD. A near-infrared spectroscopy calibration for the prediction of fresh grass quality on Irish pastures. Information Processing in Agriculture. 2022;9(2):243-253. DOI: 10.1016/j.ecoinf.2021.101467
  82. 82. Parrini S, Staglianò N, Bozzi R, Argenti G. Can grassland chemical quality be quantified using transform near-infrared spectroscopy? Animals. 2021;12(1):86. DOI: 10.3390/ani12010086

Written By

Manuela Ortega Monsalve, Tatiana Rodríguez Monroy, Luis Fernando Galeano-Vasco, Marisol Medina-Sierra and Mario Fernando Ceron-Munoz

Reviewed: 23 August 2023 Published: 14 November 2023

© 2023 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.