SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2022, 221-230 SAINS TANAH Ae Journal of Soil Science and Agroclimatology Journal homepage: http://jurnal.
id/tanah The reliability of Unmanned Aerial Vehicles (UAV.
equipped with multispectral cameras for estimating chlorophyll content, plant height, canopy area, and fruit total number of Lemons (Citrus limo.
Buyung Al Fanshuri1,2.
Cahyo Prayogo3*.
Soemarno3.
Sugeng Prijono3.
Novi Arfarita4 Doctoral Program of Agriculture Science.
Faculty of Agriculture.
University of Brawijaya.
Malang.
Indonesia Indonesia Institute for Testing Citrus and Subtropical Standard Instrument.
Indonesia Department of Soil Science.
Faculty of Agriculture.
University of Brawijaya.
Malang.
Indonesia Faculty of Agriculture.
University Islam Malang.
Malang.
Indonesia
ARTICLE INFO
ABSTRACT
Keywords:
Multispectral Unmanned Aerial Vehicle (UAV) Image Field Measurements Nondestructive Vegetation Indices Monitoring lemon production requires appropriate and efficient technology.
The use of UAVs can addressed these challenges.
The purpose of this study was to determine the best vegetation indices (VI.
for estimating chlorophyll content, plant height (PH), canopy area (CA), and fruit total numberas (FTN).
CCM 200 was used as a tool to measure the chlorophyll content index (CCI), the number of fruits was measured by hand-counter, and other variables were recorded in meters.
The UAV used was a Phantom 4 with a multispectral camera capable of capturing five different bands.
The VIs was obtained via analysis of digital numbers generated by the multispectral camera.
Then, the VIs was correlated with the CCI.
PH.
CA and FTN.
VIs tested included the following: the normalized difference vegetation index (NDVI), the normalized difference vegetation index-green (NDVI.
, the normalized different index (NDI), green minus red (GMR), simple ratio (SR), the Visible Atmospherically Resistant Index (VARI), normalized difference red edge (NDRE), simple ratio red-edge (SRRE), the simple ratio vegetation index (SRVI), and the Canopy Chlorophyll Content Index .
I).
The best model for predicting CCI was obtained using the NDVIg (R2=0.
RMSE=6.
1665 and RRMSE=0.
Meanwhile.
SR turned out to be the best model for predicting PH (R2=0.
RMSE=15.
6432 and RRMSE=0.
(R2=0.
RMSE= 0.
8826 and RRMSE=0.
, and FTN (R2=0.
RMSE=24.
5574 and RRMSE=0.
The implication of these results for future activities includes establishing early monitoring and evaluation systems for lemon yield and production.
This model was developed and tested in this specific location and under these environmental conditions.
Article history Submitted: 2023-03-24 Accepted: 2023-10-24 Available online: 2023-12-03 Published regularly:
December 2023 * Corresponding Author Email address:
cahyoprayogo@yahoo.
How to Cite: Fanshuri.
Prayogo.
Soemarno.
Prijono.
Arfarita.
The reliability of Unmanned Aerial Vehicles (UAV.
equipped with multispectral cameras for estimating chlorophyll content, plant height, canopy area, and fruit total number of Lemons (Citrus limo.
Sains Tanah Journal of Soil Science and Agroclimatology, 20.
: 221-230.
https://doi.
org/10.
20961/stjssa.
INTRODUCTION
One of the most consumed fruits during the COVID-19 pandemic were lemons (Citrus limo.
(Kutyauripo et al.
, 2.
Lemons contain a wide range of phytochemicals .
itamins and secondary metabolite.
, which are very beneficial for the body, including maintaining health and supporting the immune system (Ke et al.
, 2.
Lemons are not only consumed as fresh fruit but are also used as a main ingredient in beverages, medicines, and cosmetics.
Thus, the consumption of this type of citrus is continuously growing (Lv et al.
, 2.
The amount of citrus fruit production must be maintained at a sufficient level.
Therefore, citrus production monitoring technology is needed.
In many citrus producing countries, monitoring plant health through chlorophyll content is still manual (Zhang et al.
, 2.
Chlorophyll content of leaves and canopy of citrus plants closely related to the level of photosynthetic capacity and synthesis of plant carbohydrates to produce high quality production (Wang et al.
, 2.
Chlorophyll plays an important role in the photosynthesis process as a catalyst, this molecule is found in the chloroplasts of green plants (Myers, 2.
Leaf chlorophyll is traditionally measured by laboratory analysis following a destructive method developed by Wintermans and De Mots .
This STJSSA, p-ISSN 1412-3606 e-ISSN 2356-1424 http://dx.
org/10.
20961/stjssa.
Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 method is expensive and laborious and requires experienced laboratory staff.
It is also time consuming, especially on large tracts of land.
Remote sensing technology can be used to determine plant growth and production without measuring directly .
, nondestructivel.
as an alternative to conventional methods (Sishodia et al.
, 2.
Agronomic factors also play a role in crop production, including plant height, canopy area, and number of fruits.
Measurements of these factors in the field are still manual.
The number of fruits is counted by hand counting and the crown area and plant height are measured with a roll meter.
A nondestructive method for measuring chlorophyll content long in development is ground remote sensing, including using the chlorophyll content meter CCM 200.
Typical CCMs operate by differential absorption of light at two One is in the near-infrared range, passing through leaf pigments relatively unimpeded and serving as a reference beam, and the other is tuned to the peak absorbance of chlorophyll.
The transmission of beam energy, expressed as the ratio of absorbance beam to reference beam, yields a unitless value called the chlorophyll content index (CCI).
The CCM-200 is also an effective tool to estimate the chlorophyll concentration in date palm leaves quickly and non-destructively (Almansoori et al.
, 2.
Another measurement technique that is being developed is remote sensing with UAVs .
nmanned aerial vehicle.
This is a nondestructive technology that has been used to predict the value of leaf chlorophyll content without involving laboratory A multispectral camera can be mounted to the UAV apparatus (Yuan, 2.
Although the UAV methods collect data at lower altitudes than satellites, this technology is timesaving compared to direct measurement.
UAV is able to nondestructively predict pigments in leaves and plant canopies accurately (Tahir et al.
, 2018.
Yuan, 2.
The UAV method is faster and more effective than nondestructive analysis that employs satellite imagery, which is now widely used in agriculture.
Satellite methods have the advantage of being able to cover large areas of land and save time (Zaigham Abbas Naqvi et al.
, 2.
Generally, remote sensing measurements use NDVI which has a close correlation with plant growth and yield (Huang & Han, 2.
This method posits that the vegetation spectral index's estimate of photosynthetic capability is directly related to crop yield (Peroni Venancio et al.
, 2.
This assumption is based on research that imply that spectral measurements like as the NDVI can capture many of the factors that affect crop growth, development, and ultimately yield (Benincasa et al.
, 2.
Target objects containing green living plants or not are assessed using a graphic indicator called NDVI (Gitelson.
Other vegetation indices have been developed to compare with NDVI approaches, such as the Normalized Difference Vegetation Index-green (NDVI.
(Peroni Venancio et al.
, 2.
, the Normalized Different Index (NDI) (Widjaja Putra & Soni, 2.
Green Minus Red (GMR) (Wang et al.
Simple Ratio (SR) (Wang et al.
, 2.
, the Visible Atmospherically Resistant Index (VARI) (Widjaja Putra & Soni, 2.
Nnormalized Difference Red Edge (NDRE) (Widjaja Putra & Soni, 2.
Simple Ratio red edge (SRRE) (Gitelson et , 1.
, the Simple Ratio vegetation index (SRVI) (Widjaja Putra & Soni, 2.
, and the Canopy Chlorophyll Content Index .
I) (Widjaja Putra & Soni, 2.
A previous study reported that UAV imagery can be used to distinguish between healthy and diseased citrus plants (Fanshuri & Yunimar, 2.
Zaigham Abbas Naqvi et al.
used the DVI (Difference Vegetation Inde.
, the RDVI (Renormalized Difference Vegetation Inde.
, the MTVI2 (Modified Triangular Vegetation Index .
, the SARVI (Soil and Atmospherically Resistant Vegetation Inde.
and the Iron Oxide index to determine regression models between Vegetation Indices (VI.
and citrus leaf chlorophyll content.
However, the implementation of UAV techniques using various VIs to predict citrus chlorophyll content, plant height, canopy area, and fruit total number is not well investigated.
This study aims to develop a vegetation index that can be used to predict chlorophyll content, plant height, canopy area, and number of fruits in lemon citrus plants.
MATERIAL AND METHODS
Location and soil conditions This research was conducted at the Tlekung Experimental Garden.
Indonesian Citrus and Subtropical Fruit Research Institute (ICSFRI), in Batu City.
East Java.
Indonesia (-7.
, which has an altitude of 950 m above sea level.
Five-year-old citrus trees were used under rain-fed irrigation (Figure .
The lemon variety used in this experiment was Cai kahuripan, which is the most popular lemon variety in Indonesia.
Soil at the research site was classified as sandy loam .
% sand, 28% silt, and 40% cla.
Soil analysis showed that total N content was 0.
C-organic was 1.
available was 82.
4 ppm.
K available was 0.
67 cmol.
Ca was 76 cmol.
Mg was 0.
3 cmol.
kg-1, and CEC was 14.
Experiment design The lemon (Citrus limo.
trees utilized in this experiment were cultivated in a sloping area, and different positions along the slope .
op, middle, and bottom position.
were The middle position had the steepest incline.
Fifteen plants from each slope position were collected, resulting in a total of 45 plants for sampling.
The sample plants were measured for plant height, canopy area, and total fruit number, so there were 45 points of data for each.
The chlorophyll content of each plant from different slope positions was collected from five different leaf positions .
orth, west, south, east, and middle of the canop.
, for a total of 225 leaf samples.
The leaves selected were the second or third from the tip of the twig.
UAV high-resolution multispectral imaging and image In general, farmers in Indonesia have citrus orchards that are narrow and scattered.
A tool for remote sensing that is more suitable under these conditions than satellite imagery is the UAV.
Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 Figure 1.
Location of study The lower flying altitude of the UAV results in higher photo resolution, time saved, and low atmospheric interception (Yuan, 2.
For this reason.
UAVs were used in this research.
The current investigation, in particular, made use of high-resolution multispectral imagery captured by the DJI Phantom 4 (P.
The P4 Multispectral is a highprecision drone capable of multispectral imaging functions.
The imaging system contains six cameras with 1/2.
9-inch CMOS sensors, including an RGB camera and a multispectral camera array containing five cameras for multispectral imaging, covering the following bands: blue (B): 450 nm A 16 green (G): 560 nm A 16 nm.
red (R): 650 nm A 16 nm.
edge (RE): 730 nm A 16 nm.
and near-infrared (NIR): 840 nm A 26 nm.
The DJI Phantom 4 Pro aircraft weighs 1487g, contains a 6000mAh LiPo2S battery, and has a maximum flight speed of 6m.
s-1 .
utomatic fligh.
or 5m.
s-1 .
anual The UAV was flown at a height of 30 m at 09.
00 local time (UTC 07:.
Radiometric corrections increased the radiometric quality of the data by correcting image reflectance while taking scene illumination and sensor effect into account.
Multispectral image radiometric corrections and calibrations were carried out in three steps: .
orthomosaicking, .
digital surface map (DSM), and .
index computation (McCluney, 2.
This process was carried out using DJI TERRA software and by extracting digital numbers using ArcGIS software.
On-site chlorophyll, plant height, canopy area, and fruit total number data acquisition In June 2022, fieldwork was conducted to obtain in-situ chlorophyll data from citrus tree leaves.
The chlorophyll content of 45 citrus plants was measured using a chlorophyll meter (CCM .
CCM 200 is a nondestructive tool for measuring chlorophyll content in leaves without harming Five leaves from each tree were collected and measured with a CCM 200 plus chlorophyll meter (Opti- Sciences Inc.
Tyngsboro.
MA.
USA), and the mean value for each plant sample (CCI) was obtained.
All measurements were carried out in the morning .
to avoid variations caused by chloroplast movement throughout the day (Pereyra et al.
, 2.
The plant agronomic variables measured in this study were plant height, canopy area, and fruit total number.
Plant height was measured from the ground surface to the tip of the tallest plant using a tape Canopy area was calculated using the circle formula .
14 x r1 x r.
, where r1 is the north-south canopy width and r2 is the east-west canopy width.
The number of fruits, from the smallest to the largest, was counted for each plant using a hand counter.
Plant height, canopy area, and total number of fruit are important agronomic parameters in determining citrus production in Indonesia.
Lemons in Indonesia .
tropical climat.
are different from those in the subtropics.
The color of the fruit produced is not yellow as in subtropical This is because the temperature is higher than in subtropical regions.
Local markets prioritize the content of the juice produced, not the color of the fruit, so the harvest is carried out when the fruit is green.
The rainy and dry seasons in the tropics also result in less leaf loss than in the subtropics.
Vegetation indices Ten vegetation indices were selected based on calculations from the five bands of the UAV multispectral The digital number of each band was obtained by extracting the red (R), green (G), blue (B), near infrared (NIR), and red edge (RE) values.
The extracted values were used to calculate the 10 selected vegetation indices, which are the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Vegetation Index-green (NDVI.
, the Normalized Different Index (NDI).
Green Minus Red (GMR).
Simple Ratio (SR), the Visible Atmospherically Resistant Index (VARI).
Normalized Difference Red Edge (NDRE).
Simple Ratio red edge (SRRE), the Simple Ratio vegetation index (SRVI), and Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 the Canopy Chlorophyll Content Index .
I).
The purpose of calculating the vegetation index is to test the proximity of the vegetation index to the greenish character of the leaves.
The greenness of the leaves was nondestructively measured using a Chlorophyll Content Meter (CCM) as described above.
The formula for obtaining the vegetation index is presented in Table 1.
Data analysis The data was analyzed using Microsoft Excel.
Descriptive analysis aims to describe the data generally.
A correlation was conducted to discover the relationships between the research variables.
Variables that had both positive and negative correlations were analyzed using regression In regression analysis, the outcome variables (Y) are CCI, plant height, canopy area, and fruit total number, while the input variables (X) are NDVI.
NDVIg.
NDI.
GMR.
SR.
VARI.
NDRE.
SRRE.
SRVI , and cI.
Table 1.
Selected broadband vegetation indices for UAV multipectral cameras Vegetation indices 1 Normalized Difference Vegetation Index (NDVI) Normalized Difference Vegetation Index-green (NDVI.
Normalized Different Index (NDI) Green Minus Red (GMR) Simple Ratio (SR) Visible Atmospherically Resistant Index (VARI) Normalized Difference Red Edge (NDRE) Simple Ratio red edge (SRRE) Simple Ratio vegetation index (SRVI) Canopy Chlorophyll Content Index .
I) Formula ycAyaycI Oe ycI ycAyaycI ycI ya OeycI ya ycI yaOeycI ya ycI 0.
G-R
ya ycI ya OeycI ya ycIOeyaA ycAyaycI Oe ycIyceyccyayccyciyce ycAyaycI ycIyceyccyceyccyciyce ycAyaycI ycIyceyccyayccyciyce ycAyaycI ycI ycAyaycIya ycAyaycOya Figure 2.
Results of UAV image processing on citrus orchards Reference (Berger et al.
, 2.
(Peroni Venancio et al.
, 2.
(Widjaja Putra & Soni, 2.
(Wang et al.
, 2.
(Wang et al.
, 2.
(Widjaja Putra & Soni, 2.
(Widjaja Putra & Soni, 2.
(Gitelson et al.
, 1.
(Widjaja Putra & Soni, 2.
(Widjaja Putra & Soni, 2.
Fanshuri et al.
Table 2.
Descriptive analysis data for five bands Variable Mean
NIR
Red edge SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023
Maximum
RESULTS
Image mapping Image processing produced an integrated map of the citrus groves.
The resulting map is presented in Figure 2.
The map is presented based on each band (R.
RE, and NIR).
The maximum and minimum numbers on the map show the digital number values of all objects contained in the image.
Meanwhile, a description of the digital numbers of the selected leaf samples is presented in Table 2.
In principle, the camera captures the wavelength reflections of sunlight.
the visible wavelength (RGB), the highest average value was obtained in band G .
, and the lowest was recorded in band B .
, which is shown in Table 2.
The deviation value of R .
is higher than that of G .
and B .
The range of values was 55.
21Ae116.
27 for R, 72.
87Ae127.
23 for G, 16.
34Ae69.
44 for B, 4879.
37Ae7765.
29 for NIR, and 2887.
08Ae 36 for RE.
Ground confirmed clorophyll content, plant height, canopy area, fruit total number, and vegetation indices Ground confirmed clorophyll content, plant height, canopy area, and fruit total number values were acquired from field collection data.
The value of the vegetation indices was obtained by entering the digital number values into the appropriate formula according to Table 1.
Plant samples were taken at the same point.
Table 3 shows the descriptive analysis of CCI, plant height, canopy area, fruit total number, and vegetation indices.
From Table 3, it can be concluded that the range of CCI was 44.
2658Ae87.
6104, while PH values were 1219Ae229.
5599 cm.
CA values were 3.
0759Ae7.
0963 m2.
FTN values were 47.
6767Ae162.
NDVI values were 7410Ae0.
NDVIg values were 0.
0398Ae0.
NDI
values were 0.
0398Ae0.
GMR values were 9.
3816Ae 2043.
SR values were 1.
0796Ae1.
VARI values were 0576Ae0.
NDRE values were 0.
0127Ae0.
SRRE values 0206Ae2.
SRVI values were 50.
5909Ae111.
and cI values ranged from 0.
0166Ae0.
Correlation among vegetation indices and ground confirmed chlorophyll content, plant height, canopy area, fruit total number The value of r indicates the level of correlation between Table 4 shows that the vegetation indices have strong and very strong relationships to CCI and plant height, respectively, while the canopy area and fruit total number have moderate to very strong relationships.
NDRE.
SRRE, and cI were negatively correlated with crop variables, while the other vegetation indices were positively correlated.
As shown Minimum Standard Deviation Table 3.
Descriptive analysis data for ground-truthed chloropyll content (CCI), plant height (PH), canopy area (CA), fruit total number (FTN), and selected vegetation indices Variable CCI PH.
CA.
FTN
NDVI
NDVIg
NDI
GMR
VARI
NDRE
SRAeRE SRAeVI cI Mean Maximum Minimum Standard Deviation in Table 4, the highest r value was obtained between NDVIg and all parameters (CCI=0.
9205, plant height = 0.
canopy area = 0.
8211, fruit total number = 0.
SRRE had the lowest correlation with CCI (-0.
, plant height (-0.
, and canopy area (-0.
, while the lowest correlation to fruit total number was recorded by NDRE (-0.
Chlorophyll content, plant height, canopy area, and fruit total number modeling using vegetation indices The results of regression modeling between vegetation indices and crop parameters are shown in Table 5.
Generally, three types of regression curves were found: exponential, polynomial, and linear.
For chlorophyll content modeling.
NDVI and GMR showed exponential relationships.
SRRE and SRVI generated polynomial regressions, and the other indices showed linear relationships.
Plant height showed a linear relationship with GMR and polynomial relationships with all other calculated indices.
For canopy area modeling.
VARI.
NDRE, and Ic showed linear relationships, while other indices had polynomial models.
Fruit total number modeling resulted in polynomial regression models for NDVI.
SR, and SRVI, while other indices showed linear relationships.
SR has the highest R2 value and lowest RMSE for plant height modeling (R2=0.
8266 and RMSE=15.
with the equation PH = -695.
69SR2 2051.
7SRAe1268.
Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 Table 4.
Correlation coefficient and significance between vegetation indices and ground-truthed chlorophyll content (CCI), plant height, canopy area, and fruit total number
CCI
Plant height
Canopy area
Fruit total number
Vegetation NDVI
NDVIg
NDI
GMR
VARI
NDRE
SRAeRE SRAeVI cI Remarks: r = Correlation coefficient, ** = Highly significant .
<0.
Table 5.
Model, coefficient determinant (R.
, root mean square error (RMSE), and relative root mean square error (RRMSE) between ground-truthed chlorophyll content (CCI) and vegetation indices Vegetation indices Equation
RMSE
RRMSE
NDVI
CCI = 0.
9530 exp .
2310 NDVI)
NDVIg CCI = 397.
1135 NDVIg 30.
NDI
CCI = 397.
3294 NDI 30.
GMR
CCI = 28.
9328 exp .
0539 GMR)
CCI = 158.
2573 SRAe124.
VARI
CCI = 378.
8461 VARI 24.
NDRE
CCI = -92.
6414 NDRE 82.
SRAeRE CCI = 12.
4679 SR RE2Ae70.
9672 SR RE 142.
SRAeVI CCI = -0.
0051 SR VI2 1.
5924 SR VIAe25.
cI CCI = -74.
6055 CCI 83.
SR also has the highest R2 and the lowest RMSE for canopy area modeling (R2=0.
6886 and RMSE=0.
with the equation CA = -23.
291SR2 70.
794SRAe46.
04, and for fruit total number modeling (R2=0.
6850 and RMSE=24.
with the equation FTN = -794.
71SR2 2337.
5SRAe1545.
This can be seen in Tables 6, 7, and 8.
Meanwhile.
NDVIg has the highest R2 .
and the lowest RMSE .
for chlorophyll content modeling with the linear regression equation CCI = 397.
1135 NDVIg 30.
The most accurate model is that which has the highest R2 value with the lowest RMSE.
Therefore, the least accurate models were found for NDRE in all modeling (Tables 5, 6, 7, and .
DISCUSSION
Nondestructive measurement is needed to save time and This research produced a formula for estimating chlorophyll content, plant height, canopy area, and number of fruits.
The use of a multispectral camera in a UAV can produce a vegetation index to estimate chlorophyll content and plant agronomy.
Differences in chlorophyll contentand plant agronomy are caused by the position of the land.
The best agronomic variables were at the lowest land position, followed by the top and middle.
This is because of the slope of the land.
the most sloping land has the lowest growth.
Remote sensing data in this study was obtained from a multispectral camera installed on a UAV.
The results of image processing were used to calculate vegetation indices, which were then related to CCI, plant height, canopy area, and fruit total number.
CCI measurement uses an absorption approach (Parry et al.
, 2014.
Vesali et al.
, 2.
, while the UAV camera utilizes a reflectance measure (Samseemoung et al.
, 2.
Ten vegetation indices have varying correlations with CCI, plant height, canopy area, and fruit total number value.
Moderate correlations were found in SRRE for the canopy area and total fruit number, and in NDRE for the total fruit Other vegetation indices had strong or very strong correlations to the plant variables.
Meanwhile, for CCI and plant height modeling, all vegetation models had strong and very strong correlations, respectively.
The results of the regression analysis also show that there are variations in the coefficient of determination of the modeling sought.
The coefficient of determination in CCI modeling ranges from 0.
8, for plant height it ranges from 0.
4Ae0.
8, and for canopy area and total fruit number it ranges from 0.
3 to 0.
This shows that the highest model accuracy is in CCI modeling, followed by plant height, canopy area, and fruit total number.
SR obtained the highest R2 and the lowest RMSE and RRMSE in plant height, canopy area, and fruit total number modeling.
Meanwhile.
NDVIg has the highest R2 and the lowest RMSE in CCI modeling.
However.
R2.
RMSE, and RRMSE value gaps between NDVI.
NDVIg.
NDI.
SR, and VARI are only slightly different, with CCI under 0.
02 (R.
, 1.
1 (RMSE), and 0.
(RRMSE).
PH under 0.
15 (R.
, 6.
5 (RMSE), and 0.
04 (RRMSE).
CA under 0.
04 (R.
, 0.
08 (RMSE), and 0.
02 (RRMSE).
and FTN 11 (R.
, 3.
83 (RMSE), and 0.
02 (RRMSE).
This is in accordance with previous research that chlorophyll content Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 has strong correlations with the NDVI from UAVs (R > 0.
in maize (Marcial-Pablo et al.
, 2.
and rice (Ban et al.
, 2.
This shows that UAVs with multispectral cameras can be used to measure chlorophyll content (Benincasa et al.
, 2.
and other agronomic factors effectively.
This means that measurement will be more time-efficient and cover more land area than ground remote sensing (Zaigham Abbas Naqvi et al.
, 2.
or agronomic measurements in the field.
The lower R values on measurements of plant height, canopy area, and fruit total number are due to the interference from the reflections of the surrounding objects.
This study's findings revealed that the regression coefficient between the vegetation index and leaf chlorophyll content ranged from 5339 to 0.
This result is better than satellite remote sensing based on existing research.
Satellites are able to cover a wider capture area than UAVs.
However, atmospheric disturbances caused by clouds and reflections of other objects result in lower regression coefficients (Benincasa et , 2.
The resolution produced by a UAV image is also more detailed, where one pixel in the image represents 2.
cm in the field.
This is different from satellite imagery, where one pixel represents 0.
6-1 m in the field at high resolutions.
Table 6.
Model, coefficient determinant (R.
, root mean square error (RMSE), and relative root mean square error (RRMSE) between plant height (PH) and vegetation indices Vegetation indices Equation
RMSE
RRMSE
NDVI
PH = 2009.
7NDVI Ae2531.
9NDVI 915.
NDVIg PH = -2990.
6NDVIg2 1426.
2NDVIg 86.
NDI
PH = -2990.
9NDI2 1426.
3NDI 86.
GMR
PH = 7.
5943GMR 74.
PH = -695.
69SR2 2051.
7SRAe1268.
VARI
PH = -1856.
8VARI2 1219.
5VARI 76.
NDRE
PH = -139.
89NDRE2Ae137.
8NDRE 216.
SRAeRE PH = 21.
241SR RE2Ae127.
47SR RE 323.
SRAeVI PH = -0.
0053SR VI 2.
2595SR VI 38.
cI PH = -69.
721cI2Ae114.
47cI 217.
Table 7.
Model, coefficient determinant (R.
, root mean square error (RMSE), and relative root mean square error (RRMSE) between canopy area (CA) and vegetation indices RRMSE
Vegetation indices Equation
RMSE
NDVI
NDVIg
NDI
GMR
VARI
NDRE
SRAeRE SRAeVI Ic CA = 156.
25NDVI2Ae220.
36NDVI 80.
CA = -88.
546NDVIg2 50.
648NDVI 1.
CA = -88.
549NDI2 50.
651NDI 1.
CA = 0.
005GMR2 0.
1553GMR 1.
CA = -23.
291SR2 70.
794SR - 46.
CA = 31.
834 VARI 1.
CA = -7.
2142NDRE 6.
CA = 1.
1489SR RE2Ae6.
0926SR RE 11.
CA = -0.
0003SR VI2 0.
1047SR VI - 1.
CA = -5.
8126Ic 6.
Table 8.
Model, coefficient determinant (R.
, root mean square error (RMSE), and relative root mean square error (RRMSE) between fruit total number (FTN) and vegetation indices Vegetation indices Equation
RMSE
RRMSE
NDVI
NDVIg
NDI
GMR
VARI
NDRE
SRAeRE SRAeVI Ic FTN = 2790.
8NDVI2Ae3678.
3NDVI 1252.
FTN = 927.
49NDVIg 21.
FTN = 927.
58NDI 21.
FTN = 8.
269GMR - 13.
FTN = -794.
71SR2 2337.
5SR - 1545.
FTN = 888.
92VARI 7.
FTN = -191.
17NDRE 139.
FTN = -62.
929SR RE 202.
FTN = -0.
0102SR VI2 3.
3159SR VI - 87.
FTN = -156.
07Ic 141.
Fanshuri et al.
SAINS TANAH Ae Journal of Soil Science and Agroclimatology, 20.
, 2023 The results also show that the highest accuracy is achieved by modeling that uses three-band (RGB) data and the NDVIg in CCI modeling and SR in other modeling.
The order of modeling findings based on band data is an RGB-based formula followed by NIR and red edge.
This is due to the reflection of the observed object as the observed field variable is included in the visible wavelength .
Oe700 n.
This statement is supported by previous research on rice that variables related to canopy cover are influenced by green and red bands (Ban et al.
, 2.
However, red edge data-based modeling can be used to predict chlorophyll content (R2 = 0.
5Ae This is also supported by previous research on coffee plants, where the vegetation index using the red edge band produced a significant relationship to chlorophyll content in coffee leaves (Widjaja Putra & Soni, 2.
In general.
VI is an index of vegetation coverage, but vegetation is highly reflective in the near infrared and strongly absorbing in the red range .
ow reflectivit.
This harmful canopy background as well as other ground effects and noise have been masked by the appearance of blue channels (Senecal, 2.
Therefore.
VIs from red edge are only used as an alternative (Delegido et al.
, 2.
The best equations obtained for chlorophyll content, plant height, canopy area, and fruit total number are CCI = 1135 NDVIg 30.
PH = -695.
69SR2 2051.
7SRAe 4.
CA = -23.
291SR2 70.
794SRAe46.
04, and FTN = 71SR2 2337.
5SRAe1545.
This research also succeeded in developing vegetation indices that were used in previous According to previous studies, a UAV multispectral camera can be utilized to evaluate chlorophyll levels in citrus plants using the vegetation indices DVI (Vegetation Difference Inde.
RDVI (Renormalized Difference Vegetation Inde.
MTVI2 (Modified Triangle Vegetation Index .
SARVI (Soil and Atmosphere Resistant Vegetation Inde.
and Iron Oxide These estimations had R2 values of 0.
5Ae0.
8 (Zaigham Abbas Naqvi et al.
, 2.
Other research used the VIs Normalized Difference Vegetation Index (NDVI), the Transformed Normalized Difference Vegetation Index (TNDVI), the Modified Chlorophyll Absorbed Ratio Index (MCARI.
, the Soil Adjusted Vegetation Index (SAVI), and the Modified Soil Adjusted Vegetation Index (MSAVI.
and had R2 values of 0.
8 (Tahir et al.
, 2.
SR has been proven to be able to detect biomass.
N content, and LAI in rice plants.
The R value is higher than NGI.
NRI, and H (Wang et al.
, 2.
The NDVI is a vegetation index that is generally used in remote sensing.
This index only uses the red and NIR bands in its calculations.
It has been proven not only to estimate chlorophyll but also yield in wheat plants (Benincasa et al.
When compared to the NDVI, the SR index (R2 = 0.
predicted soybean grain yield better (Gcayi et al.
, 2.
CONCLUSION
This study proves that the vegetation index resulting from the analysis of UAV imagery results can be used to estimate the chlorophyll content, plant height, canopy area, and total fruit number of lemon (Citrus limo.
The relationship of the vegetation index to CCI and plant height is strong and very strong, respectively, while the crown area and number of fruits have a moderate to very strong relationship.
These results indicate that the use of UAVs can be used to replace manual measurementson the ground.
Accurate and fast forecasting of production components is needed in the future to assist the decision making of farmers and the government.
However, this research was conducted on green lemons with site-specific environmental conditions .
, dry land, a mountainous area with average rainfall of 1889 mm/year, and soil developed from Kawi volcanic material, which is classified as an inceptiso.
The results obtained may differ depending on climatic conditions, geographic locations, and soil characteristics.
Therefore, future studies will need to be carried out under different environmental conditions and with different varieties of lemons.
Acknowlegement Thanks to the Indonesian Agency for Agricultural Research and Development (IAARD) for the research funding provided.
Thanks also to Yossi Andika.
Sativandi Riza.
Iqbal Farel and Mradipta Panenggak who assisted in field data collection and image analysis.
Thank you also to Jacob Fettic who helped proof reading this paper.
Declaration of Competing Interest The authors declare that no competing financial or personal interests that may appear and influence the work reported in this paper.
References