A Leaf Chlorophyll Content Dataset for Crops: A Comparative Study Using Spectrophotometric and Multispectral Imagery Data

Abstract

This paper presents a dataset for a comparative analysis of direct (spectrophotometric) and indirect (multispectral imagery-based) methods for quantifying crop leaf chlorophyll content. The dataset originates from a study conducted in the Department of Cauca, Colombia, a region characterized by diverse agricultural production. Data collection focused on seven economically important crops, namely coffee (Coffea arabica), Hass avocado (Persea americana), potato (Solanum tuberosum), tomato (Solanum lycopersicum), sugar cane (Saccharum officinarum), corn (Zea mays) and banana (Musa paradisiaca). Sampling was conducted across various locations and phenological stages (healthy, wilted, senescent), with each leaf subdivided into six sections (A–F) to facilitate the analysis of intra-leaf chlorophyll distribution. Direct measurements of leaf chlorophyll content were obtained by laboratory spectrophotometry following the method of Jeffrey and Humphrey, allowing for the determination of chlorophyll A and B content. Simultaneously, indirect estimates of leaf chlorophyll content were obtained from multispectral images captured at the leaf level using a MicaSense Red-Edge camera under controlled illumination. A set of 32 vegetation indices was then calculated from these multispectral images using MATLAB. Both direct and indirect methods were applied to the same leaf samples to allow for direct comparison. The dataset, provided as an Excel (.xlsx) file, comprises raw data covering laboratory-measured chlorophyll A and B content and calculated values for the 32 vegetation indices. Each row of the tabular dataset represents an individual leaf sample, identified by plant species, leaf identifier, and phenological stage. The resulting dataset, containing 16,660 records, is structured to support research evaluating the direct relationship between spectrophotometric measurements and multispectral image-based vegetation indices for estimating leaf chlorophyll content. Spearman’s correlation coefficient reveals significant positive relationships between leaf chlorophyll content and several vegetation indices, highlighting its potential for a nondestructive assessment of this pigment. Therefore, this dataset offers significant potential for researchers in remote sensing, precision agriculture, and plant physiology to assess the accuracy and reliability of various vegetation indices in diverse crops and conditions, develop and refine chlorophyll estimation models, and execute meta-analyses or comparative studies on leaf chlorophyll quantification methodologies.

1. Dataset Specifications

In

Table 1. Summary of Dataset Characteristics.

Subject	Earth and Environmental Sciences
Specific subject area	Comparison of direct (spectrophotometric) and indirect (multispectral imagery-based vegetation indices) methods for quantifying leaf chlorophyll content in key crops.
Type of data	.xlsx file (dataset with numbers)
Data collection	Data were obtained by laboratory spectrophotometry and multispectral imaging. Leaves from seven crops were collected and analyzed. For spectrophotometry, chlorophyll was extracted from leaf tissue to determine chlorophyll A and B content. A MicaSense Red-Edge camera (Kansas, United States) captured images in five spectral bands under controlled illumination for multispectral imaging. Vegetation indices were calculated from these images. Both methods were applied to the same leaf samples to allow for a direct comparison of chlorophyll quantification approaches. Data were meticulously recorded and processed using MATLAB and statistical software.
Data source location	Data collection for this research was conducted in the department of Cauca, Colombia, a region known for its varied agricultural production. The study selected sites within Cauca to represent the varied growing conditions of different crops. Samples of coffee, Hass avocado, sugarcane, banana, tomato, and cassava were collected in Popayan, Cauca, at various locations in this area. For the maize crop, samples were collected in the rural area of Cajibío, also located in Cauca. Finally, potato samples were obtained in Paletará, which finalized the geographic scope of data collection within the department of Cauca, Colombia.
Data accessibility	Repository name: Zenodo Data identification number: https://doi.org/10.5281/zenodo.15002560 Direct URL to data: https://zenodo.org/records/15002561 (accessed on 8 September 2025) This archive is supported by the Corporación Universitaria Comfacauca—Unicomfacauca and hosted by Zenodo
Related research article	None

, the main characteristics of the dataset used in this research can be observed.

2. Value of Data

• This dataset provides an empirical framework for evaluating and validating indirect leaf chlorophyll content (LCC) estimations derived from multispectral imagery against direct spectrophotometric measurements.
• The dataset includes measurements of seven economically important crops in the department of Cauca, Colombia, such as coffee, avocado, potato, tomato, sugar cane, corn, and banana. Including this diverse range of plant species facilitates comparative analyses of their distinct physiological characteristics and supports the development of crop-specific and generalized LCC prediction models.
• The dataset covers various variables, including laboratory-measured chlorophyll A and B content and 32 vegetation indices derived from multispectral imagery. This dataset allows researchers to evaluate the performance of various vegetation indices in estimating LCC and to identify optimal indices for different crops or conditions.
• This resource constitutes an essential contribution to remote sensing and precision agriculture. It furnishes the necessary data for developing, refining, and validating remote sensing-based chlorophyll estimation algorithms while enabling data reuse for large-scale meta-analyses and cross-regional or inter-species comparative studies.

3. Summary

The Department of Cauca, Colombia’s agricultural sector, is characterized by diverse production, featuring crops of significant economic and regional/national food security importance. Among these crops, coffee (Coffea arabica) is the main crop grown in four agroecological zones and involves more than 93,000 peasant, Afro-descendant, and Indigenous families [1]. Hass avocado (Persea americana) is the third most important national fruit export, with a notable increase in 2023 [2]. In addition, crops such as potato (Solanum tuberosum), with an area of 177,000 hectares [3], and tomato (Solanum lycopersicum), promoted through protected production systems (greenhouses), contribute to agrobiodiversity and benefit numerous families [4]. Sugarcane (Saccharum officinarum), with 56,000 hectares for panela production, is a fundamental pillar of the national economy [5], while maize (Zea mays) generates a considerable number of jobs [6]. Finally, banana (Musa paradisiaca), with a cultivated area of more than 400,000 hectares, drives research and development strategies to improve its resistance to pests and diseases [7]. These crops are, therefore, of clear socioeconomic and productive relevance to the region.

Monitoring crop health is crucial for optimizing agricultural production’s quality and commercial value. Leaf chlorophyll content (LCC) serves as a well-established physiological indicator to assess crop health and the primary productivity of ecosystems, and crops produced in Cauca are no exception [8]. Determining the LCC allows for the early detection of possible phytosanitary or nutritional problems. Traditionally, LCC is determined using laboratory spectrophotometric techniques. While accurate, these methods require the chemical extraction of chlorophyll, making them destructive and limiting their application for in vivo and large-scale analyses [9].

Multispectral imaging offers an efficient, non-destructive alternative for the remote estimation of leaf chlorophyll content using vegetation indices [10]. Vegetation indices are mathematical combinations of spectral reflectance at specific wavelengths sensitive to variations in plant biophysical parameters, including chlorophyll content [11]. Using multispectral imaging, predictive systems have been developed to determine the LCC in different crops. However, the limitations of vegetation indices include the temporal frequency of satellite image acquisition and the low spatial resolution, especially for more minor crops [12]. These factors can lead to inconsistencies in vegetation index determination, partly due to their inherent spectral limitations [13].

A study correlating remote sensing-derived vegetation indices with laboratory-based chlorophyll measurements has been documented for Coffea arabica [1]. However, the applicability of these findings to other crops of regional and national relevance in Colombia is unknown. Therefore, it is unknown which are the optimal vegetation indices for LCC estimation in other plant species of agronomic interest in the region according to an accepted measurement in the domain, such as laboratory measurements.

This dataset was collected in research designed to evaluate and compare the effectiveness of direct spectrophotometric measurements versus indirect estimates using multispectral imaging to quantify LCC in key crops in the department of Cauca in Colombia. Multispectral images at the leaf level and corresponding direct chlorophyll measurements were acquired for seven representative crops of the region under controlled environmental conditions. The data comprises chlorophyll A and B content, measured by laboratory spectrophotometry, and 32 different vegetation indices derived from multispectral images of the same leaf samples, resulting in a dataset of approximately 16,000 records. The data acquisition and processing methodologies were adapted from established protocols in the domain.

Finally, this dataset provides a quantitative framework for evaluating the correlation between direct (destructive) and indirect (non-destructive) methods of chlorophyll quantification in crops. The complete raw and processed datasets are provided under an open access license to ensure transparency and extensibility. The resource is a benchmark for calibrating and validating remote sensing models engineered for chlorophyll content estimation. It facilitates an analysis of the relationships between vegetation spectral indices and absolute chlorophyll concentrations across diverse crop species and agroecological conditions. This structure supports subsequent applications, including meta-analyses, inter-regional and inter-crop comparative studies, and advancing non-destructive phytosanitary monitoring protocols for precision agriculture.

4. Data Description

This section describes the dataset, which is provided in the file ‘Segumiento_fenologico.xlsx’, an Excel workbook (.xlsx) conforming to ISO standards. This file contains data from a phenological and spectral study of various plant species, focusing on leaf characteristics through different stages of development. The dataset is structured to facilitate the analysis of relationships between direct laboratory measurements of chlorophyll and indirect estimates derived from vegetation indices across different plant species and leaf phenological stages.

The file Segumiento_fenologico.xlsx is organized in a tabular format, where each row represents an individual leaf sample and each column corresponds to a specific measured or calculated variable. The dataset includes chlorophyll content measurements, spectral reflectance indices, and leaf phenological stage classifications.

Table 2. Description of columns in the dataset.

Column Name	Description	Data Type
Plant species	Scientific name of the plant species	Text
Leaf	Identifier for the leaf sample, including leaf number and replicate (A–F)	Text
Chlorophyll A	Measurement of chlorophyll A content in the leaf sample	Numerical
Chlorophyll B	Measurement of chlorophyll B content in the leaf sample	Numerical
NDVI	Normalized Difference Vegetation Index	Numerical
GNDVI	Green Normalized Difference Vegetation Index	Numerical
NDRE	Normalized Difference Red Edge index	Numerical
OSAVI	Optimized Soil Adjusted Vegetation Index	Numerical
CVI	Chlorophyll Vegetation Index	Numerical
CCCI	Canopy Chlorophyll Content Index	Numerical
EVI	Enhanced Vegetation Index	Numerical
ARVI	Atmospherically Resistant Vegetation Index	Numerical
ARVI2	Atmospherically Resistant Vegetation Index 2	Numerical
GLI	Green Leaf Index	Numerical
ATSAVI	Adjusted Transformed Soil Adjusted Vegetation Index	Numerical
RBNDVI	Red-Blue Normalized Difference Vegetation Index	Numerical
NGRDI	Normalized Green Red Difference Index	Numerical
DVI	Difference Vegetation Index	Numerical
GARI	Green Atmospherically Resistant Index	Numerical
RDVI	Renormalized Difference Vegetation Index	Numerical
NLI	Non-linear Index	Numerical
MNLI	Modified Non-linear Index	Numerical
MSAVI2	Modified Soil Adjusted Vegetation Index 2	Numerical
TDVI	Transformed Difference Vegetation Index	Numerical
GEMI	Global Environment Monitoring Index	Numerical
CIGreen	Chlorophyll Index Green	Numerical
SR	Simple ratio	Numerical
IPVI	Infrared Percentage Vegetation Index	Numerical
GRNDVI	Green-Red Normalized Difference Vegetation Index	Numerical
GBNDVI	Green-Blue Normalized Difference Vegetation Index	Numerical
BNDVI	Blue Normalized Difference Vegetation Index	Numerical
CIrededge	Chlorophyll Index Red Edge	Numerical
NDWI	Normalized Difference Water Index	Numerical
RENDVI	Red Edge Normalized Difference Vegetation Index	Numerical
RENDVI2	Red Edge Normalized Difference Vegetation Index 2	Numerical
MTCI	MERIS Terrestrial Chlorophyll Index	Numerical
Leaf stage	Phenological stage of the leaf	Numerical

details the content of each column to clarify the dataset’s structure.

5. Methods

The objective of this study was to compare direct measurements of LCC using laboratory spectrophotometry with indirect measurements using vegetation indices. For this purpose, key references included studies by Solis et al. [1] comparing LCC with vegetation indices for chlorophyll A and B in coffee and another study [14] comparing chlorophyll A measurements with vegetation indices.

Laboratory chlorophyll measurements were conducted using the spectrophotometry method described by Jeffrey and Humphrey [15]. The method adopted the methodological framework employed by Solis et al. [1] for calculating spectral indices. Finally, statistical techniques were applied using an analytical pipeline that included descriptive and inferential statistics.

The primary materials used in this research included a MicaSense Red-Edge camera for multispectral imaging, a GENESYS™ 20 spectrophotometer (Massachusetts, United States) [16] for laboratory spectrophotometric analysis, Matlab software R2023b for multispectral image processing and analysis, and vegetation indices [1]. The PAST and R-4.5.1 programs were used for data analysis and correlation calculation. Figure 1 shows the research workflow.

6. Study Area and Plant Materials

Leaf samples of coffee, Hass avocado, sugarcane, banana, tomato, and cassava were collected in Popayán, Cauca, in a humid temperate climate, 1800 m.a.s.l., 2°29′35″ N, 76°34′33″ W, 25 °C, with fertile soils characteristic of the region. The maize samples were obtained in the rural area of Cajibío, Cauca, with a warm climate, 2°34′13″ N, 76°35′50″ W. Potato samples were collected in Paletará, Cauca, with a temperate climate, 2°12′33″ N, 76°30′28″ W, with a loose soil type.

To evaluate leaf chlorophyll in the seven representative crops of the department of Cauca, three types of leaves were selected from each one, including healthy, withered, and senescent leaves, analyzing a total of nine leaves per crop. Each leaf was divided into six sections called A, B, C, D, E, and F, as shown in Figure 2. This was performed to observe the distribution of chlorophyll in leaves of different species at localized points. For this purpose, three batches were randomly selected from each crop. The first batch of leaves was in the senescent or chlorotic state. The second batch of leaves was in an optimal state and presented a uniform green color. Finally, the third batch of wilted leaves presented their characteristic brown color.

7. Estimation of LCC by Laboratory Spectrophotometry

Photosynthetic pigment quantification in leaf samples (Figure 3) was performed using spectrophotometry with a GENESYS™ 20 spectrophotometer. The procedure followed the guidelines of Jeffrey and Humphrey [15], which recommend using leaf tissue, excluding the central vein, and minimizing secondary veins.

Immediately after collection, leaf samples were stored under refrigerated conditions (≤4 °C) to inhibit enzymatic activity and prevent pigment degradation during transport to the laboratory. To maintain turgor and minimize dehydration, each leaf sample was individually placed in hermetically sealed polyethylene bags containing water.

Subsequently, the samples were transported to the Corporación Universitaria Comfacauca—Unicomfacauca laboratory—where they were weighed, and multispectral images were captured for further processing. A total of 50 mg of leaf tissue from each sample was weighed for chlorophyll extraction, and 15 mL of 90% (v/v) acetone was added. The tissue was macerated, and the resulting extract was made up to a final volume of 6.5 mL with 90% (v/v) acetone, following the methodology described by Simon and Helliwell [17].

Finally, the liquid’s absorbance was measured at wavelengths 644 and 661, and the chlorophyll a and b content were estimated in mg/mL using the dichromatic Equations (1) and (2) of Jeffrey and Humphrey in [15].

$$Chla\left({\displaystyle \frac{\mathrm{m}\mathrm{g}}{\mathrm{m}\mathrm{l}}}\right)=11.24E661-2.04E644$$

(1)

$$Chlb\left({\displaystyle \frac{\mathrm{m}\mathrm{g}}{\mathrm{m}\mathrm{l}}}\right)=20.13E644-4.19E661$$

(2)

8. Multispectral Image Acquisition and Processing for Estimating Vegetation Indices

Multispectral images were acquired using a MicaSense Red-Edge camera, which captured spectral information in five discrete bands, including blue (λ = 475 nm), green (λ = 560 nm), red (λ = 668 nm), red edge (λ = 717 nm), and near-infrared (NIR, λ = 840 nm). The red-edge and near-infrared bands are highly sensitive to variations in LCC and plant biomass [18], making them relevant indicators of the vegetation’s physiological status. The image was captured in a controlled environment under uniform and semi-intensive illumination conditions for all leaf samples. The sensor was positioned at a nominal distance of 80 cm from the focal plane of the leaves. Before acquiring each set of images, a radiometric calibration was performed using the MicaSense reference panel at the same nominal distance of 80 cm. At least 10 images were taken per leaf for subsequent post-processing.

The vegetation indices were calculated using MATLAB software, a platform widely used in multispectral image processing and quantitative analyses of vegetation biophysical parameters. The MATLAB pipeline for processing and analyzing multispectral images and obtaining vegetation indices in regions of interest (ROIs) is detailed in Figure 4. In this sense, the post-processing applied comprised the conversion of pixel intensity to radiance and, subsequently, to reflectance through the application of a radiometric calibration panel, implemented through radiometric and geometric calibration algorithms, optimizing the spatial and spectral accuracy of the data.

The algorithm performs geometric alignment and spectral band integration, using band five as the reference image. First, it reads and registers the images by aligning bands 1 to 4 using a translation method based on an optimizer. Then, it combines the spectral bands into a multiband image and extracts the individual channels. From there, it generates an enhanced RGB image using a focus mask and gamma correction for more realistic visualization. It then calculates multiple vegetation indices and allows the user to select six regions of interest on the RGB image, where the indices are cropped and average values are calculated for each area. Finally, these values are displayed and exported to an Excel file for further analysis.

9. Statistical Pipeline of the Data

An analytical pipeline was implemented using the statistical tools PAST v5 (PAleontological STatistics) and R with packages such as dplyr and FactoInvestigate, which were recognized for their ability to analyze large volumes of data, calculate correlations, and describe relationships between variables from unidimensional data [1]. Initially, normality tests were performed to evaluate the data distribution obtained by direct and indirect measurements. The Shapiro–Wilk and Anderson–Darling tests [19] were used as statistical diagnostic methods, resulting in the majority of the data not following a normal distribution; it was decided to use Spearman’s correlation coefficient for the analysis, since this method does not require normality in the data and is suitable for ordinal variables or when a non-linear relationship between variables is presumed [20].

Descriptive statistics were used to summarize and characterize the central tendency and dispersion of chlorophyll measurements and vegetation indices, including the mean, median, and variance. In addition, a correlation matrix was generated that identified significant relationships between direct chlorophyll measurements and spectral indices, allowing for consistent patterns to be determined and providing insight into the association between direct and indirect methods for estimating leaf chlorophyll content.

10. Correlations Between Spectrophotometry Measurements and Vegetation Indices

In general, chlorophyll content (A and B) was positively correlated with most evaluated vegetation indices (Figure 5). That is, an increase in chlorophyll concentration tends to be associated with an increase in the values of these indices and vice versa. The magnitude of these correlations varies, ranging from values close to unity, indicating a strong positive association, to values close to zero, indicating a weak or null correlation. Notably, negative correlations were observed for the NDWI (Normalized Difference Water Index), which is expected, as this index is designed to detect water content, and increased chlorophyll concentration does not necessarily correlate with increased leaf water content.

The correlation between chlorophyll A and B is 0.95, indicating a strong positive correlation. This suggests that higher chlorophyll A content tends to be associated with higher chlorophyll B content, which is biologically consistent given that both pigments are related to photosynthesis.

Regarding correlations between chlorophyll A and vegetation indices, strong positive relationships (r > 0.7) are observed with the following indices: NDVI (r = 0.71), OSAVI (r = 0.70), EVI (r = 0.70), ARVI (r = 0.72), ARVI2 (r = 0.71), MNLI (r = 0.72), GEMI (r = 0.72), and IPVI (r = 0.71). There are also moderate positive correlations (between 0.5 and 0.7) with the indices GNDVI (r = 0.53), NDRE (r = 0.53), RBNDVI (r = 0.62), CIGreen (r = 0.53), CIrededge (r = 0.52), RENDVI (r = 0.53), and RENDVI2 (r = 0.51). Finally, weak correlations, both positive and negative, that are close to 0 are found for CVI (r = 0.14), CCCI (r = 0.005), GLI (r = 0.20), BNDVI (r = 0.33), and NDWI (r = −0.53).

For chlorophyll B, correlations were found with the vegetation indices NDVI (r = 0.66), OSAVI (r = 0.65), EVI (r = 0.65), ARVI (r = 0.67), and ARVI2 (r = 0.66). Some correlations are moderately strong, such as with ATSAVI (r = 0.64), RDVI (r = 0.64), MNLI (r = 0.67), MSAVI2 (r = 0.64), TDVI (r = 0.64), GEMI (r = 0.68), SR (r = 0.62), IPVI (r = 0.66), GRNDVI (r = 0.66), and DVI (r = 0.62). There are also moderate positive correlations (between r = 0.5 to 0.65) with the indices GNDVI (r = 0.55), NDRE (r = 0.56), RBNDVI (r = 0. 57), NGRDI (r = 0.35), CIGreen (r = 0.55), GBNDVI (r = 0.48), CIrededge (r = 0.55), RENDVI (r = 0.56), RENDVI2 (r = 0.54), and MTCI (r = 0.37). Finally, weak correlations, both positive and negative, that are close to 0 are observed for CVI (r = 0.22), CCCI (r = 0.05), GLI (r = 0.08), BNDVI (r = 0.28), and NDWI (r = −0.55).

11. Limitations

The dataset presented in this paper has some limitations that are important to recognize. First, the dataset is restricted to seven crops from a single department in Colombia, which limits the generalizability of the findings to other geographic regions, crop varieties, or agricultural practices with different climatic and edaphic conditions, which may restrict the generalizability of the results to other geographic locations with different environmental conditions, agricultural practices, or crop varieties. Second, the data were collected under controlled environmental conditions, which although beneficial for data standardization, may not fully reflect the variability and complexity of real agricultural scenarios. Third, the multispectral images were captured under controlled uniform illumination conditions, which may not represent natural variations in the field, potentially affecting the accuracy of indirect chlorophyll estimates. Finally, direct chlorophyll measurements, made by spectrophotometry, could have variability due to differences in pigment extraction or equipment calibration, introducing uncertainty in the data. These limitations should be considered when using the dataset in broader contexts.