High-resolution banana row maps for the characterization of spatial variability in the field

ABSTRACT

Precision agriculture (PA) practices in banana production chains have received limited attention. Based on the literature, the investigation of spatial and temporal variability in banana orchards should be customized according to the characteristics of the crop. This study aimed to develop and evaluate methods for mapping the spatial variability in soil properties at row- and clump-resolutions in a banana orchard, and to generate row and clump maps with high-spatial-resolution soil property information. A banana orchard was investigated, and georeferenced soil sampling was conducted with calibration and validation points. Methods for reconstructing banana rows and clumps were proposed, called Methods 1 and 2 and Alternative Methods 1 and 2. Surface and line maps at row- and clump-resolutions for soil chemical and physical properties were generated using ordinary kriging and Voronoi polygons. Subsequently, the discrepancies between the data obtained from the validation points and the predictions devised from the surfaces generated by the proposed approaches were calculated, and the RMSE was used as a performance parameter. Methods 1 and 2 were appropriate and reliable approaches for site-specific management and allow for specific and optimized crop management in banana cultivation, offering greater accuracy in cultivation operations such as fertilization.

Keywords
Kriging; line maps; precision agriculture; soil sampling; Voronoi polygon

INTRODUCTION

Adoption of small-scale Precision Agriculture (PA) practices for fruit growth is promising. Several studies have analyzed and characterized the spatial and temporal variability in soil and plant characteristics in orchards to enable more accurate agronomic decision-making and optimize the efficient distribution of resources in the field (Bassoi et al., 2014; Colaço et al., 2019a,b; Uribeetxebarria et al., 2019; Selvaraj et al., 2020; Gatti et al., 2022; Kasimati et al., 2023). However, the use of PA, specifically for banana production, has received limited attention.

Few studies have presented approaches to implement PA practices in banana orchards for site-specific management. These include Stoorvogel and Orlich (2000), Fu et al. (2019), Neupane et al. (2019), Calou et al. (2020), Chen et al. (2020), Lamour et al. (2020, 2021), Wu et al. (2022) and Aeberli et al. (2023). As these authors have stated in their studies, the low adoption of PA in banana growth is believed to be related to: i) the low availability of data from the orchard; ii) the fact that the crop is labor-dependent and is yet to exploit advantages of sensor-embedded agricultural machinery; iii) small size of the production area and high cost of adopting PA; iv) absence of support policies; and v) the fact the methods for investigating spatial and temporal variability must be specific to the characteristics of bananas, that is, the extant approaches are not particularized to the banana crop.

Agricultural operations in banana orchards, such as fertilizing, occur manually and in clump resolution and are managed by plant units (Lamour et al., 2021). However, agronomic decisions generally consider broad average values for the growing area. Localized management at the plant level is a future goal envisaged by the PA. Therefore, banana cultivation-specific techniques and approaches are desirable for the PA of fruit. This advancement is necessary owing to the prominent position of banana as the most consumed fruit worldwide and its impact on local and global trade (Neupane et al., 2019). In terms of world trade, banana is the fifth most important agricultural product (Raja and Rajendran, 2023), representing more than 113 million tons of bananas produced worldwide per year, with India, the People’s Republic of China, Indonesia, and Brazil being the largest producers (Embrapa, 2023).

Banana cultivation is considered semi-perennial because each plant completes its cycle after the bunch is harvested, and orchards produce bananas for several years before the plot is renovated (Lamour et al., 2021). Individual banana plants develop at their own pace and are not synchronized in their phenology (Leroux et al., 2018). This asynchrony in phenology between plants within the same field highlights the importance of adopting PA practices that enable site-specific management at plant-resolution.

Studies have been conducted in banana fields to investigate the spatial and temporal variability in soil chemical and physical properties (Zucoloto et al., 2011a,b; Freitas et al., 2016); plant-related variables (Lamour et al., 2017, 2020, 2021; Leroux et al., 2018); and disease management, such as yellow and black Sigatokas and Panama disease (Zucoloto et al., 2009; Uchôa et al., 2011; Freitas et al., 2016; Gómez-Correa et al., 2017), and generate surface maps of the banana orchard. The surface map displaying the spatial distribution of a specific attribute is a tool developed to visualize this variability aimed at guiding decision-makers in crop management. These maps are generated using geometric methods, such as Voronoi polygons (Reem, 2023), interpolations via deterministic methods like inverse distance weighted (IDW), or geostatistical techniques such as ordinary kriging (Hilal et al., 2024). Generally, the performance of these methods, in terms of their fidelity to reality, can vary due to the complexity of spatially estimating variables in the field and because each method operates on different principles of surface generation. However, as banana management is conducted plant-by-plant, the way the surface maps are presented may not be as useful for the specific management of banana plantations.

Therefore, other approaches to mapping the spatial variability in soil and plant properties should be customized, such as the generation of high-spatial-resolution maps of rows and clumps in banana orchards obtained from these surface maps. This type of specific management has been seen as promising for other crops that are managed by rows, such as sugarcane, and has been adopted in high-spatial-resolution row-mapping studies in sugarcane plantations (Maldaner and Molin, 2020). These tools and approaches are the first step towards understanding the variability in variables within each banana row and adapting local management practices, such as fertilization. The optimization and conscious use of agricultural resources are strategies linked to the Sustainable Development Goals (SDGs) established in 2015 by the United Nations Member States, especially Goal 2, which promotes sustainable agriculture, increasing agricultural yield, enhancing the income of small-scale food producers and promoting land and soil health (UN, 2024). Furthermore, they have the potential to empower small and medium-sized farmers to become more active in the context of PA by guiding their decision-making.

This study proposes mapping approaches to understand the spatial variability in variables in banana orchards. We hypothesized that mapping the soil properties spatial variability at row- and clump-resolutions would provide accurate information that can support the decision-making of banana growers in the field. This study aimed to i) develop and evaluate methods for mapping the spatial variability in soil properties at row- and clump-resolution in a banana orchard; and ii) create row and clump maps with soil property information at high spatial resolution.

MATERIALS AND METHODS

The steps followed in this study are summarized in the flowchart shown in figure 1. These steps are explained in detail below.

Figure 1
Flowchart of the steps taken in this study.

Study area and soil sampling

The study area is located in Seropédica, state of Rio de Janeiro (22° 45′ 08.62″ S and 43° 40′ 28.50″ W). The climate of the study region is tropical-rainy, with a dry winter, Aw. The average annual rainfall in the municipality is 1,354 mm, with an average annual temperature of 23.50 °C. The studied banana orchard was established in December 2016 with the seedlings of ‘BRS Princesa’ variety planted in single rows with average row spacings of 2.50 and 2.00 m between clumps. The average planting density was approximately 2,000 plants ha^-1. The soil in the area was classified as Argissolo Amarelo (Santos et al., 2018), which corresponds to Typic Hapludult (Soil Survey Staff, 2014). The altitude in the area ranges from 20.41 to 21.17 m.

Banana orchard was cultivated following traditional local cultivation practices, with conventional “Mother-Daughter-Granddaughter” clump management (Alves and Oliveira, 1999). A sprinkler irrigation system, top-dressing using mineral fertilizer N-P-K (Freire, 2013), and organic top-dressing with tanned bovine manure, with 10 L applied per clump, were employed. Weeds were controlled by weeding around the base of the plants and monthly mowing. The recommended cultivation practices for banana crops were conducted in accordance with Borges and Souza (2004), specifically including the removal of inactive leaves and cutting the pseudostem after harvest. Banana orchard was in its fourth cultivation cycle and had an average yield of 18.42 Mg ha^-1 from previous cycles.

The plot covers 0.185 ha, in which 42 sample points were georeferenced and spaced in a semi-regular grid of approximately 6.80 × 6.00 m². Geographical coordinates of each point were calculated and determined using a Leica Total Station, TPS300 Basic Series, with a distance error of 2.00–5.00 mm + 2.00 ppm (accuracy). Soil sampling points were defined by identifying the first row of banana trees as the reference line, and sample collection was performed between the clumps in this row, followed by soil sampling elsewhere in the orchard (Figure 2). Soil sampling was conducted in July 2019. Forty-two soil samples, representing each point on the sampling grid, were collected using a Dutch auger at layers of 0.00-0.05, 0.05-0.10, 0.10-0.20, and 0.20-0.40 m, totaling 168 samples. Samples were sent to the laboratory to determine soil chemical and physical properties.

Figure 2
Sampling grid of points and reference line for row generation in the banana orchard in Seropédica – Rio de Janeiro, Brazil.

Five soil properties were selected owing to their agronomic importance in banana cultivation. Soil properties considered in this study were assimilable phosphorus (P; g kg^-1), cation exchange capacity (T; cmol_c kg^-1), total organic carbon (TOC; g kg^-1), base saturation (V; %), and clay (Clay; g kg^-1). The value T was obtained by calculating T = sum of bases (SB) + H⁺ + Al³⁺, where SB included the sum of the exchangeable bases (Ca²⁺ + Mg²⁺ + K⁺ + Na⁺) and total acidity (H+Al). Total acidity was extracted from the soil using 1 mol L^-1 of calcium acetate at pH 7.0, removing Al³⁺ and H⁺ that were retained on the surface of the colloids by electrostatic forces. Base saturation (V) was calculated as the percentage of the ratio between SB and T. All soil chemical properties were obtained according to the method described by Teixeira et al. (2017). The TOC was determined according to the method described by Yeomans and Bremner (1988). Granulometric fractions of the total sand, clay, and silt were obtained according to the modifications proposed by Rezende (1979) and the pipette method (Day, 1965).

Spatial analysis - Semivariogram and ordinary kriging (OK)

The spatial structure of the soil properties was evaluated using isotropic variogram modeling in SmartMap, an open-source QGIS plugin for digital mapping (Pereira et al., 2022). The semivariograms for each soil property were tested for exponential, spherical, and Gaussian models, and the variogram parameters of the nugget effect, plateau, and range were obtained. The leave-one-out cross-validation (LOOCV) method was used to measure the performance of the fitted models, with the best model being that with the lowest root-mean-square error (RMSE; Equation 1) (Pang et al., 2023). The RMSE behaves like the standard deviation of the residuals and indicates how the residuals are spread. It is always non-negative; the closer it is to zero, the higher the quality of the measured or estimated values, and it has the same unit as the measurements (Tyagi et al., 2022; Pellikka et al., 2023).

$\text{ RMSE }=\sqrt{\frac{1}{n}{\Sigma }_{ i=1}^{ n} \left\{Z\right(xi)-\stackrel{ ^}{Z}(xi){\}}^2}$ Eq. 1

in which: Ẑ(xi) denotes the predicted value; Ẑ(xi) is the observed (known) value; and n is the number of values in the dataset.

Surface maps of the soil properties were generated by OK using QGIS software v. 3.22.10. The X- and Y-pixel dimensions of the generated surfaces are detailed in the following section.

Row mapping approaches in the banana orchard

Banana tree-rows were reconstructed based on the defined reference line (first row of banana trees) and the known spacing between the rows in the banana grove. The QGIS software vectorization tools were used to create the geometry of the lines and make them parallel. Mapping of the spatial variability in soil properties at banana row- and clump-resolution was performed using four approaches: Method 1, Alternative Method 1, Method 2, and Alternative Method 2.

Method 1: to estimate the values of the soil properties in banana rows, Voronoi diagrams were generated using the “Voronoi Polygons” tool in QGIS software. This algorithm segments the area into smaller polygons, each corresponding to the coverage area of each point in the processed dataset (Reem, 2023). Subsequently, the Voronoi polygon values were joined to the line geometries corresponding to the banana rows using the QGIS “Intersection” tool. This tool overlaps geometries (Voronoi diagram and lines) and generates a line geometry output with the values where both layers intersect (Maldaner and Molin, 2020), that is, the lines inherit the polygon values. These steps generated soil attribute maps at banana row resolution (Figure 3).

Figure 3
Methods for obtaining maps in row-resolution (Method 1) to characterize the spatial variability in soil properties.

Method 2: The line geometries were broken into 2.00-m nodes (average spacing between clumps in a row), and a grid of points was generated in the lines with a spacing of 2.00 × 2.50 m² (Figure 4). The QGIS “Join attributes by nearest” vectorization tool was used to assign values corresponding to the soil sampling points (red dots) to the grid of points generated in the rows (blue dots). Therefore, the soil properties of each point in the grid are the average of the soil property data of its neighbors.

Figure 4
Methods for obtaining maps at banana clump-resolution (Method 2) to characterize the spatial variability in soil properties.

From the grid of points generated on the lines, the values of the soil properties were estimated for the entire length and surface of the banana plantation using the “Voronoi Polygons” tool in QGIS software. Simultaneously, a buffer was generated for the layer of points corresponding to the clumps using the grid of points generated in the rows. Next, the Voronoi polygon values were joined to the banana clump-corresponding buffer geometries using the QGIS “Intersection” tool. These steps generated maps of soil properties at the banana plantation clump-level.

Alternative Methods 1 and 2: This study presents two alternative approaches for analyzing rows and clumps in banana orchards. The proposed Alternative Methods 1 and 2 follow the same steps as Methods 1 and 2, respectively. However, in the steps where the geometries of the rows intersected with the clump buffer with spatialized surface, instead of using the Voronoi polygon-generated surface, the surface obtained using the semivariogram and ordinary kriging was used. Thus, the OK-generated surface in Alternative Method 1 had pixels with dimensions of 6.00 × 6.00 m², whereas the OK-generated surface in Alternative Method 2 had pixels of 2.50 × 2.40 m². These pixel dimensions correspond to the average pixel resolutions achieved using the Voronoi polygons in Methods 1 and 2.

Analyses and validation

Comparison between the soil properties at the different layers was assessed using an analysis of variance (ANOVA) and, when significant differences were found, the average values were compared by the Tukey at 5.00 % significance test (p<0.05).

A validation dataset was sampled to assess the predictive estimates of the proposed approaches. Eighteen points were georeferenced on a sampling grid and soil samples were collected from the layers, 0.00-0.05, 0.05-0.10, 0.10-0.20, and 0.20-0.40 m (Figure 5). The discrepancy between the data obtained from the validation points and the predictions obtained from the surfaces generated by the proposed approaches was calculated using Method 1, Alternative Method 1, Method 2, and Alternative Method 2.

Figure 5
Validation points sampled in the banana orchard.

The RMSE is the performance parameter used (Equation 1). To validate the proposed line mapping methods, a comparison was made between the line and clump maps generated from the surface maps interpolated using the Voronoi and OK methods. The JupyterLab software (Kluyver et al., 2016; Jupyter, 2022) was used to process the data.

RESULTS AND DISCUSSION

The observed frequency distribution of the soil property values indicate the trend in the data and its behavior at each layer. Exploratory and descriptive data for the dataset used in this study are presented in table 1 and figure 6.

Table 1
Descriptive statistics of the properties of the Argissolo Amarelo at different layers in the banana orchard

Variable	Layer	Min	Max	Mean	STD	CV	Skewness
	m
P (g kg-1)	0.00-0.05	6.92	49.47	25.10	11.26	0.45	0.48
	0.05-0.10	6.42	36.75	14.53	7.77	0.53	1.57
	0.10-0.20	7.04	45.83	15.60	9.60	0.62	1.76
	0.20-0.40	3.45	16.64	6.63	3.15	0.48	1.53
T (cmolc kg-1)	0.00-0.05	6.14	12.72	8.60	1.37	0.16	0.45
	0.05-0.10	6.87	12.66	8.66	1.25	0.14	1.36
	0.10-0.20	5.05	12.68	6.80	1.38	0.20	1.96
	0.20-0.40	5.91	10.41	7.95	1.15	0.14	0.38
V (%)	0.00-0.05	67.03	92.32	81.66	5.96	0.07	-0.47
	0.05-0.10	59.70	87.60	74.81	6.09	0.08	-0.33
	0.10-0.20	37.07	85.41	61.09	10.72	0.18	0.54
	0.20-0.40	43.16	74.54	58.87	7.63	0.13	-0.23
TOC (g kg-1)	0.00-0.05	9.92	32.00	16.36	4.13	0.25	1.27
	0.05-0.10	8.15	17.41	11.32	2.17	0.19	0.72
	0.10-0.20	6.72	25.66	10.37	3.21	0.31	2.77
	0.20-0.40	3.15	8.70	5.45	1.49	0.27	0.41
Clay (g kg-1)	0.00-0.05	14.00	299.00	231.12	56.34	0.24	-1.90
	0.05-0.10	122.00	401.00	268.24	55.04	0.21	0.01
	0.10-0.20	122.00	395.00	239.17	56.92	0.24	0.55
	0.20-0.40	87.00	388.00	266.95	60.89	0.23	-0.86

• Min: minimum; Max: maximum; STD: standard deviation; CV: coefficient of variation; P: phosphorus; T: cation exchange capacity; TOC: total organic carbon; and V: base saturation.

Figure 6
Histograms of the soil properties evaluated at the different soil layers along the depth of 0.00-0.40 m. The x-axis indicates the values obtained for each variable and the y-axis represents their densities or frequencies. P: phosphorus; T: cation exchange capacity; TOC: total organic carbon; V: base saturation; clay. The dashed vertical line indicates the average value of the soil property. Horizontally, different lowercase letters indicate a significant difference between the soil property at different layers (Tukey’s test; p<0.05).

Histograms of the soil properties suggest: (i) the average values were not similar, with a significant difference for most propertties at the different soil layers, using the Tukey test at 95.00 % confidence level (p<0.05); (ii) despite the skewness of the data observed for most of the soil properties, the data had a tendency towards normality; and (iii) the variance was different between the soil layers but had a pattern of behavior.

Soil chemical properties were higher at the surface and decreased as soil depth increased. This behavior was due to the application of inputs, deposition of plant material from the banana plant, and decomposition of organic matter, making nutrients in the soil more available on the surface. Even in a small plot (0.185 ha), there was a variation in the values of soil properties and heterogeneity were evident. In addition, extreme values were noticeable, manifested by elongated distribution tails to the right and left in the histograms. This heterogeneity was evidenced by the mean values and standard deviations of the soil properties.

Figure 6 also shows that the “T” values follow a normal distribution, implying that approximately 95.00 % of the data is expected to be within two standard deviations of the mean. This finding is fundamental in the context of statistical inference, given the error rates are restricted to 5.00 % (Siegel and Wagner, 2022).

Except for clay, a reduction in the standard deviation of the data for the surface soil properties were observed (0.00-0.10 m), with a sharp increase at 0.10-0.20 m and a subsequent reduction at 0.20-0.40 m. This indicates the heterogeneous pattern of soil properties on the surface (0.00-0.05 m) is presented again at 0.10-0.20 m, which is related to the greater presence of banana roots on the surface, concentrated at around 62.00 % at 0.00-0.30 m soil layer, according to the literature (Embrapa, 2023). The standard deviation is a number that summarizes how far away from the mean the data values are normally; that is, it measures the extent of randomness of individuals in relation to the mean (Siegel and Wagner, 2022). Because the standard deviation is the square root of the variance, these trends were expected to be reflected in the plateaus of the semivariograms (semivariance on the x-axis of the semivariogram). However, histograms do not provide spatial information, which prevents statements about the area of influence or the periodicity of segregation in space (Benito et al., 2023).

The study of variograms facilitated the modelling of the spatial correlations between the values of soil properties at the sampling points in the banana orchard. Variogram analysis (Table 2) indicated the existence of distinct spatial patterns for the soil properties in relation to the studied layers.

Table 2
Variographic parameters of soil properties at the layers of 0.00–0.40 m obtained from geostatistical analysis

Variable	Layer	Model	Nugget	sill	Range	RMSE – LOOCV
	m				m
P (g kg-1)	0.00-0.05	esf	34.00	173.34	19.50	13.00
T (cmolc kg-1)		esf	1.35	2.33	20.87	1.40
V (%)		exp	0.00	37.50	17.21	4.81
TOC (g kg-1)		gau	15.48	20.37	29.32	4.06
Clay (g kg-1)		esf	10.00	3916.15	35.41	63.45
P (g kg-1)	0.05-0.10	esf	35.47	68.12	35.41	7.51
T (cmolc kg-1)		exp	0.00	1.74	16.82	1.18
V (%)		exp	21.97	36.81	18.60	5.65
TOC (g kg-1)		gau	3.13	5.48	34.31	1.96
Clay (g kg-1)		exp	5.00	3392.78	27.10	44.20
P (g kg-1)	0.10-0.20	esf	21.91	86.25	13.39	11.09
T (cmolc kg-1)		esf	1.31	2.39	35.19	1.23
V (%)		exp	18.44	137.09	26.22	10.26
TOC (g kg-1)		esf	6.77	12.18	25.26	3.31
Clay (g kg-1)		esf	20.00	3939.46	28.15	46.38
P (g kg-1)	0.20-0.40	esf	5.39	9.60	15.99	3.19
T (cmolc kg-1)		exp	0.00	1.65	35.41	0.87
V (%)		gau	13.79	75.83	21.74	6.17
TOC (g kg-1)		esf	0.22	2.86	35.41	1.05
Clay (g kg-1)		exp	10.00	2599.51	26.86	51.35

• RMSE – LOOCV: Root Mean Squared Error, obtained from leave-one-out validation; Model: Sph - spherical, Exp – exponential, Gau – Gaussian; P: phosphorus; T: cation exchange capacity; TOC: total organic carbon; V: base saturation.

The variables P, T, TOC, V, and Clay had different semivariogram adjustment models (spherical, exponential, and Gaussian), which suggests each variable has a different spatial behavior in banana orchard. Except for T and V, all nugget-effect values were greater than zero at all layers, indicating unstructured variability, which is related to the discontinuity in the semivariogram and natural variability of the soil and plants (Lamour et al., 2020).

Phosphorus and clay had the highest cross-validation RMSE values, indicating the complexity of modeling soil P and physical properties. The RMSE-CV provides a measure of how well the fitted semivariogram models explain the observed data, in which low values indicate a good model fit (Tyagi et al., 2022). This highlights the importance of considering other approaches or factors that may influence the complex variations in soil properties.

Phosphorus was the property with the highest nugget-effect value. This characteristic is associated with the study region, which naturally has soils with low P content; P is easily absorbed by the high levels of iron and aluminum oxides present in Argissolos (Santos et al., 2018). Furthermore, to obtain accurate estimates and calculate an appropriate variogram, it is essential to have a substantial number of samples (Maroufpoor et al., 2020).

Therefore, it is important to consider that the magnitude of the nugget-effect is related to the spatial resolution of the analyses (Benito et al., 2023). In this study, 42 soil samples were obtained from a field of less than 1.00 ha, making it a dense sampling. This indicates the complexity of spatially modeling P in the soil (Baio et al., 2023), which may contribute to the observed high nugget-effect values. Thus, despite the successful spatial modeling of P from high-resolution sampling in banana orchard, the nugget-effect values were substantially higher than those identified in the variogram.

Phosphorus spatial characterization, as well as the other essential nutrients for banana plants at the different stages of production, is necessary to verify the availability of nutrients and the nutritional balance of the soil, in accordance with Liebig’s Law of the Minimum (Wallace, 1993). Based on this, together with leaf analysis and the orchard’s yield history, it is possible to understand what has been exported by the plants over the cycles and generate more assertive input recommendations, for example.

Even in small cultivated plots, soil properties were observed to exhibit different ranges of values at the same depth, indicating that variations in content can be identified with greater spatial resolution. Therefore, obtaining a substantial number of observations is crucial. This higher density of data improves the understanding of spatial variations in soil properties and is fundamental for the accurate calculation of experimental variograms (Lamour et al., 2020). With greater spatial resolution, it is possible to identify more detailed variations in soil composition, which allows for a more accurate analysis of the plot and optimized decision-making (Heydari et al., 2023).

Detection of patches with different soil properties suggests that uniform management throughout the orchard may not be the most appropriate. The spatial variation of soil properties in this study, such as T value and V, indicates fertilizations can be optimized according to local nutritional needs. For example, the soil in a region with higher V values, such as in the south of the study area, does not have the same nutritional needs as the eastern region, with lower V values. Therefore, these spatial differences affect the fertilization recommendation for the crop and ensure sustainability in the use of agricultural resources.

In the validation of the proposed approaches, the row and clump maps generated by the OK, Alternative Methods 1 and 2 had the lowest RMSE values compared to the Voronoi polygon estimates in Methods 1 and 2. This pattern of behavior was observed for all soil properties at all layers, except for T at 0.00-0.05 and 0.10-0.20 m (Table 3).

Table 3
Root-mean-square error (RMSE) values obtained by the validation and prediction points for soil properties at 0.00-0.40 m from the proposed approaches: Method 1, Alternative Method 1, Method 2, and Alternative Method 2

Variable	Layer	Method 1	Alternative Method 1	Method 2	Alternative Method 2
	m
P (g kg-1)	0.00-0.05	15.11	11.89	15.25	12.91
	0.05-0.10	9.49	5.77	10.99	5.53
	0.10-0.20	10.40	6.97	11.94	6.89
	0.20-0.40	5.78	4.79	6.62	3.75
T (cmolc kg-1)	0.00-0.05	1.72	1.75	1.86	1.61
	0.05-0.10	1.55	1.18	1.08	1.33
	0.10-0.20	0.94	1.01	0.98	1.05
	0.20-0.40	1.00	0.82	1.07	0.73
V (%)	0.00-0.05	9.15	7.15	7.95	8.62
	0.05-0.10	8.18	5.90	9.05	6.56
	0.10-0.20	8.98	5.18	9.73	5.11
	0.20-0.40	7.53	6.35	7.42	5.91
TOC (g kg-1)	0.00-0.05	5.31	4.63	5.32	4.88
	0.05-0.10	3.54	2.89	3.51	3.09
	0.10-0.20	2.98	2.22	2.92	2.32
	0.20-0.40	1.89	1.49	1.79	1.45
Clay (g kg-1)	0.00-0.05	73.43	65.97	74.36	55.82
	0.05-0.10	54.89	50.01	65.95	45.20
	0.10-0.20	57.08	48.59	69.45	46.91
	0.20-0.40	102.67	81.68	98.81	67.49

• P: phosphorus; T: cation exchange capacity; TOC: total organic carbon; V: base saturation.

Individual rows and clumps with soil property values showed variations over short distances were preserved for approaches using the Voronoi polygon (Methods 1 and 2). Data smoothing occurred in all maps generated by OK (Figure 7). Data smoothing was also identified by Maldaner and Molin (2020) in a study examining information processing in sugarcane rows. This study explored yield data using OK surface maps and sugarcane row maps using Voronoi polygons. They observed the OK method was widely known to operate as a low-pass filter, resulting in a smoothing effect on spatial data.

Figure 7
Spatial variability maps of soil properties, V and clay at 0.05-0.10 m, at row and individual clump-resolution in a banana orchard. It is possible to verify data smoothing by OK-using alternative methods. V: base saturation.

Therefore, the lower RMSE values obtained by the alternative methods (OK) may have been influenced by the smoothing of the data compared to Methods 1 and 2. This is because RMSE is sensitive to extreme values, as these high-squared values can amplify the impact of forecast errors (Tyagi et al., 2022). Furthermore, the density of points in this study (42 sampling points + 18 sampling points for validation) is believed to cause higher RMSE values for Methods 1 and 2 when used with Voronoi polygon. This is because studies with higher point densities, which have observed lower RMSE values for methods using the Voronoi polygon compared with the OK method, as exemplified in a study by Maldaner and Molin (2020).

In general, for soil properties, Methods 1 and 2 showed less data smoothing than the alternative methods because the Voronoi diagram is a purely geometric concept. It is a mathematical structure that divides a space into regions based on the proximity of specific points called “generators” or “vertices.” Each region of the Voronoi polygon is defined as the area closer to a specific generator than to any other generator in space (Reem, 2023).

For instance, the V contents at 0.05-0.10 m, which were 59.70 and 87.60 %, identified as extreme values in Methods 1 and 2, were not detected by the alternative methods. Instead, these values were smoothed to the minimum and maximum levels of 69.65 and 82.57%, respectively, in both alternative methods. This is because regions with low values of a given soil property are not sufficiently large to influence the pixel values in OK. Therefore, the Voronoi polygon preserved certain values obtained from banana orchards (Figure 8).

Figure 8
Spatial characterization of soil properties, T and P at 0.00-0.05 m, at row- and clump-resolution in banana orchard. The average pixel size in Method 1 and Alternative Method 1 was 6.00 × 6.00 m², whereas in Method 2 and Alternative Method 2, it was 2.50 × 2.40 m². P: phosphorus; T: cation exchange capacity.

The pixel size used to generate the surface maps using the OK and Voronoi methods directly influenced the resulting values in the maps representing the banana rows and clumps. This influence, in turn, was reflected in the RMSE values. The average pixel size values in Method 1 and Alternative Method 1 were 6.00 × 6.00 m², whereas in Method 2 and Alternative Method 2, they were 2.50 × 2.40 m² (Figure 8). Because of the higher spatial resolution of the aforementioned methods, the soil properties in the banana orchards were represented more accurately.

The proposed approaches at all studied soil layers had low RMSE values for the soil property T, which generated maps of banana rows and clumps with a low number of variation errors, providing high reliability for its application in localized management and individualized management approaches per row.

The TOC was the second soil property with the lowest RMSE (Figure 9), followed by V. The complex natural variability of TOC poses a substantial challenge for monitoring specific management practices in the field (Pellikka et al., 2023). Modeling soil C and mapping it at row-resolution in a banana orchard represents a breakthrough, as it provides details of TOC variations along the banana rows. This approach is especially relevant at a local scale, given the need for further research on the modeling and mapping soil properties in tropical regions.

Figure 9
Line and surface maps of TOC at the studied layers along the depth of 0.00-0.40 m. Surface maps generated by the Voronoi polygon and OK represent the banana row maps by Method 1 and Alternative Method 1, respectively. OK: ordinary kriging, TOC: total organic carbon.

Additionally, the soil properties revealed a greater predictive capacity at the subsurface layers, especially at 0.20-0.40 m. This trend was evidenced by the lower RMSE values observed for all soil properties, except for P and Clay, which was possibly associated with the lower heterogeneity of the data compared to the surface soil values.

For banana orchards, row and clump maps provided details on the levels of soil properties in a banana row. This allows the cultivation of this crop to be dealt with in a specific and optimized manner, allowing cultivation operations such as fertilization to be performed with greater precision. Therefore, for banana-growing areas, surface maps may not be the best way to visualize spatial variability in the field.

In the current scenario and at a local scale, it was found that for almost all soil properties, the RMSE values were lower for Alternative Methods 1 and 2 than for Methods 1 and 2; however, the alternative methods exhibited data smoothing at all soil layers. Notably, the RMSE values for Methods 1 and 2 were very close to those for the alternative methods, with the row and clump mapping methods preserving the extreme values obtained for the banana orchard. Therefore, Methods 1 and 2 seem appropriate and reliable approaches for site-specific management, especially for the T, in which the highest quality estimated values were obtained.

In addition, for all the approaches proposed, the RMSE values for the clay were too high to generate reliable maps of banana rows and clumps, demonstrating greater accuracy in predicting chemical properties in relation to the physical properties of the soil in the banana orchard. Despite this, as previously mentioned, high RMSE values were observed for P, especially on the surface, due to the complexity of modeling its spatial structure. Therefore, in future studies, other approaches to generating surface maps or different sampling strategies should be explored to reduce errors in data estimates.

In agricultural practice, banana growers usually manually perform field operations in row- and clump-resolutions. Therefore, the main contribution of our study is the development of a method capable of generating detailed maps of banana rows and clumps and providing a solid basis for farmers to make more accurate decisions. These maps provide crucial information on the spatial distribution of soil and can be generated using the proposed approaches for plant-related attributes and other relevant variables. Using these data, banana growers can make more assertive and strategic decisions regarding management and agricultural practices.

Generating detailed maps of rows and clumps helps optimize the allocation of resources, such as labor and inputs, resulting in more efficient and economical operations. Optimizing decision-making can also positively impact the yield and quality of agricultural products because banana production is continuous over time, and various phenological stages coexist in the field (Aeberli et al., 2023).

This study addresses proposals for characterizing the spatial variability of soil properties based on maps of rows and clumps in banana plantations, seeking strategies for localized management in orchards. Certainly, the existing PA practices and strategies described in the literature should be adapted and customized for banana cultivation, considering its particularities and demands, such as the approaches proposed in this study.

In this study, the clay and TOC contents showed specific spatial characteristics that should be evaluated in future studies, as shown in figures 7 and 9, respectively. Soil texture, for example, is a characteristic that does not change in a short time in the field. Therefore, other paths for implementing PA practices in this context could be followed, such as targeted sampling and spatial and temporal investigations based on the delimitation of management zones (Uribeetxebarria et al., 2019) or, in the absence of high-resolution data, conducting studies considering a cell-size approach in the banana orchard (cell sampling) (Molin et al., 2015). Targeting zones with homogeneous characteristics seems promising for managing rows and clumps in banana orchards. In addition, targeted sampling demonstrates economic viability because the ability of farmers to employ dense sampling density for the entire field is limited.

CONCLUSIONS

This study showed the spatial heterogeneity of the properties of an Argissolo Amarelo at fine-scale resolution in a banana orchard in the fourth production cycle. Methods were proposed to generate maps of the variability of spatial soil properties at the resolutions of banana rows and clumps. For all the proposed methods, the soil properties T, V, and TOC were found to have the lowest RMSE values in validation, indicating reliability in the specific management of these properties by the proposed approaches. However, the P and clay properties, despite the spatial heterogeneity observed in the plot, showed higher RMSE values due to the complexity of modeling their spatial structures. The approaches proposed as Alternative Methods 1 and 2 showed greater accuracy and performance in estimating soil properties than Methods 1 and 2, respectively. However, the alternative methods smoothened the minimum and maximum values for estimating soil properties at all layers assessed, whereas Methods 1 and 2 reliably represented the ranges of the existing values for soil properties in the banana orchard. Notably, the RMSE values of Methods 1 and 2 were very close to those of the alternative methods, suggesting an equivalence between them. Therefore, Methods 1 and 2 appear to be appropriate and reliable approaches for site-specific management. This allows specific and optizimed crop management in banana cultivation, enabling cultivation operations, such as fertilization, to be conducted more accurately.

• How to cite: Silva ERO, Barros MM, Silva GO, Vaz AFS, Pereira MG. High-resolution banana row maps for the characterization of spatial variability in the field. Rev Bras Cienc Solo. 2024;48:e0240001. https://doi.org/10.36783/18069657rbcs20240001

DATA AVAILABILITY

All data was generated or analyzed in this study.

REFERENCES

• Aeberli A, Phinn S, Johansen K, Robson A, Lamb DW. Characterisation of banana plant growth using high-spatiotemporal-resolution multispectral UAV imagery. Remote Sens. 2023;15:679. https://doi.org/10.3390/rs15030679
  » https://doi.org/10.3390/rs15030679
• Alves EJ, Oliveira MA. Práticas culturais. In: Alves EJ, editor. A cultura da banana: Aspectos técnicos, socioeconômicos e agroindustriais. Brasília, DF: Embrapa Produção de Informação; 1999. p. 335-51.
• Baio FHR, Alixame D, Neves DC, Teodoro LPR, Silva Júnior CA, Shiratsuchi LS, Oliveira JT, Teodoro PE. Adding random points to sampling grids to improve the quality of soil fertility maps. Precis Agric. 2023;24:2081-97. https://doi.org/10.1007/s11119-023-10031-x
  » https://doi.org/10.1007/s11119-023-10031-x
• Bassoi LH, Miele A, Reisser Junior C, Gebler L, Flores CA, Alba JMF, Grego CR, Terra VSS, Timm LC, Nascimento PS. In: Bernardi ACC, Naime JM, Resende AV, Bassoi LH, Inamasu RY, editors. Agricultura de precisão: Resultados de um novo olhar. Brasília, DF: Embrapa Instrumentação; 2014. p. 350-60.
• Benito S, Egels G, Hartmaier A, Weber S. Statistical characterization of segregation-driven inhomogeneities in metallic microstructures employing fast first-order variograms. Mater Today Commun. 2023;34:105016. https://doi.org/10.1016/j.mtcomm.2022.105016
  » https://doi.org/10.1016/j.mtcomm.2022.105016
• Borges AL, Souza LS. O cultivo da bananeira. Cruz das Almas: Embrapa Mandioca e Fruticultura; 2004.
• Calou VBC, Teixeira AS, Moreira LCJ, Lima CS, Oliveira JB, Oliveira MRR. The use of UAVs in monitoring yellow sigatoka in banana. Biosyst Eng. 2020;193:115-25. https://doi.org/10.1016/j.biosystemseng.2020.02.016
  » https://doi.org/10.1016/j.biosystemseng.2020.02.016
• Chen M, Tang Y, Zou X, Huang K, Huang Z, Zhou H, Wang C, Lian G. Three-dimensional perception of orchard banana central stock enhanced by adaptive multi-vision technology. Comput Electron Agr. 2020;174:105508. https://doi.org/10.1016/j.compag.2020.105508
  » https://doi.org/10.1016/j.compag.2020.105508
• Colaço AF, Molin JP, Rosell-Polo JR, Escolà A. Spatial variability in commercial orange groves. Part 1: Canopy volume and height. Precis Agric. 2019a;20:788-804. https://doi.org/10.1007/s11119-018-9612-3
  » https://doi.org/10.1007/s11119-018-9612-3
• Colaço AF, Molin JP, Rosell-Polo JR, Escolà A. Spatial variability in commercial orange groves. Part 2: Relating canopy geometry to soil attributes and historical yield. Precis Agric. 2019b;20:805-22. https://doi.org/10.1007/s11119-018-9615-0
  » https://doi.org/10.1007/s11119-018-9615-0
• Day PR. Particle fractionation and particle-size analysis. In: Black CA, editor. Methods of soil analysis: Part 1 Physical and mineralogical properties, including statistics of measurement and sampling. Madison: American Society of Agronomy; 1965. p. 545-67. https://doi.org/10.2134/agronmonogr9.1.c43
  » https://doi.org/10.2134/agronmonogr9.1.c43
• Empresa Brasileira de Pesquisa Agropecuária – Embrapa. Banana. Cruz das Almas, BA: Embrapa Mandioca e Fruticultura; 2023 [cited 2023 Dec 10]. Available from: https://www.embrapa.br/mandioca-e-fruticultura/cultivos/banana
  » https://www.embrapa.br/mandioca-e-fruticultura/cultivos/banana
• Freire LR. Manual de calagem e adubação do estado do Rio de Janeiro. Brasília, DF: Embrapa; Seropédica: Editora Universidade Rural; 2013.
• Freitas AS, Pozza EA, Alves MC, Coelho G, Rocha HS, Pozza AAA. Spatial distribution of Yellow Sigatoka Leaf Spot correlated with soil fertility and plant nutrition. Precis Agric. 2016;17:93-107. https://doi.org/10.1007/s11119-015-9409-6
  » https://doi.org/10.1007/s11119-015-9409-6
• Fu L, Duan J, Zou X, Lin G, Song S, Ji B, Yang Z. Banana detection based on color and texture features in the natural environment. Comput Electron Agr. 2019;167:105057. https://doi.org/10.1016/j.compag.2019.105057
  » https://doi.org/10.1016/j.compag.2019.105057
• Gatti M, Garavani A, Squeri C, Diti I, De Monte A, Scotti C, Poni S. Effects of intra-vineyard variability and soil heterogeneity on vine performance, dry matter and nutrient partitioning. Precis Agric. 2022;23:150-77. https://doi.org/10.1007/s11119-021-09831-w
  » https://doi.org/10.1007/s11119-021-09831-w
• Gómez-Correa JC, Torres-Aponte WS, Cayón-Salinas DG, Hoyos-Carvajal LM, Castañeda-Sánchez DA. Modelación espacial de la Sigatoka negra (Mycosphaerella fijiensis M. Morelet) en banano cv. Gran Enano. Rev Ceres. 2017;64:47-54. https://doi.org/10.1590/0034-737X201764010007
  » https://doi.org/10.1590/0034-737X201764010007
• Heydari L, Bayat H, Castrignanò A. Scale-dependent geostatistical modelling of crop-soil relationships in view of Precision Agriculture. Precis Agric. 2023;24:1261-87. https://doi.org/10.1007/s11119-023-09989-5
  » https://doi.org/10.1007/s11119-023-09989-5
• Hilal A, Bangroo SA, Kirmani NA, Wani JA, Biswas A, Bhat MI, Farooq K, Bashir O, Shah TI. Geostatistical modeling - a tool for predictive soil mapping. In: Lamine S, Srivastava PK, Kayad A, Muñoz-Arriola F, Pandey PC, editors. Earth observation: remote sensing in precision agriculture. Massachusetts: Academic Press; 2024. p. 389-418. https://doi.org/10.1016/B978-0-323-91068-2.00011-4
  » https://doi.org/10.1016/B978-0-323-91068-2.00011-4
• Jupyter. JupyterLab: A Next-Generation Notebook Interface [internet]. 2022 [cited 2023 Nov 12] Available from: https://jupyter.org/
  » https://jupyter.org/
• Kasimati A, Psiroukis V, Darra N, Kalogrias A, Kalivas D, Taylor JA, Fountas S. Investigation of the similarities between NDVI maps from different proximal and remote sensing platforms in explaining vineyard variability. Precis Agric. 2023;24:1220-40. https://doi.org/10.1007/s11119-022-09984-2
  » https://doi.org/10.1007/s11119-022-09984-2
• Kluyver T, Ragan-Kelley B, Pérez F, Granger B, Bussonnier M, Frederic J, Kelley K, Hamrick J, Grout J, Corlay S, Ivanov P, Avila D, Abdalla S, Willing C, Team JP. Jupyter Notebooks – A publishing format for reproducible computational workflows. In: Loizides F, Schmidt B, editors. Positioning and power in academic publishing: Players, agents and agendas. Amsterdam: IOS Press; 2016. p. 87-90. https://doi.org/10.3233/978-1-61499-649-1-87
  » https://doi.org/10.3233/978-1-61499-649-1-87
• Lamour J, Le Moguédec G, Naud O, Lechaudel M, Taylor J, Tisseyre B. Evaluating the drivers of banana flowering cycle duration using a stochastic model and on farm production data. Precis Agric. 2021;22:873-96. https://doi.org/10.1007/s11119-020-09762-y
  » https://doi.org/10.1007/s11119-020-09762-y
• Lamour J, Naud O, Lechaudel M, Le Moguédec G, Taylor J, Tisseyre B. Spatial analysis and mapping of banana crop properties: issues of the asynchronicity of the banana production and proposition of a statistical method to take it into account. Precis Agric. 2020;21:897-921. https://doi.org/10.1007/s11119-019-09700-7
  » https://doi.org/10.1007/s11119-019-09700-7
• Lamour J, Naud O, Lechaudel M, Tisseyre B. Mapping properties of an asynchronous crop: The example of time interval between flowering and maturity of banana. Adv Anim Biosci. 2017;8:481-6. https://doi.org/10.1017/S2040470017000449
  » https://doi.org/10.1017/S2040470017000449
• Leroux C, Jones H, Pichon L, Guillaume S, Lamour J, Taylor J, Naud O, Crestey T, Lablee JL, Tisseyre B. GeoFIS: An open source, decision-support tool for precision agriculture data. Agriculture. 2018;8:73. https://doi.org/10.3390/agriculture8060073
  » https://doi.org/10.3390/agriculture8060073
• Maldaner LF, Molin JP. Data processing within rows for sugarcane yield mapping. Sci Agric. 2020;77:e20180391. https://doi.org/10.1590/1678-992X-2018-0391
  » https://doi.org/10.1590/1678-992X-2018-0391
• Maroufpoor S, Bozorg-Haddad O, Chu X. Geostatistics: Principles and methods. In: Samui P, Bui DT, Chakraborty S, Deo RC, editors. Handbook of probabilistic models. Oxford: Butterworth-Heinemann; 2020. p. 229-42. https://doi.org/10.1016/B978-0-12-816514-0.00009-6
  » https://doi.org/10.1016/B978-0-12-816514-0.00009-6
• Molin JP, Amaral LR, Colaço AF. Agricultura de precisão. São Paulo: Oficina de Textos; 2015.
• Neupane B, Horanont T, Hung ND. Deep learning based banana plant detection and counting using high-resolution red-green-blue (RGB) images collected from unmanned aerial vehicle (UAV). PLoS ONE. 2019;14:e0223906. https://doi.org/10.1371/journal.pone.0223906
  » https://doi.org/10.1371/journal.pone.0223906
• Pang Y, Wang Y, Lai X, Zhang S, Liang P, Song X. Enhanced Kriging leave-one-out cross-validation in improving model estimation and optimization. Comput Methods Appl Mech Eng. 2023;414:116194. https://doi.org/10.1016/j.cma.2023.116194
  » https://doi.org/10.1016/j.cma.2023.116194
• Pellikka P, Luotamo M, Sädekoski N, Hietanen J, Vuorinne I, Räsänen M, Heiskanen J, Siljander M, Karhu K, Klami A. Tropical altitudinal gradient soil organic carbon and nitrogen estimation using Specim IQ portable imaging spectrometer. Sci Total Environ. 2023;883:163677. https://doi.org/10.1016/j.scitotenv.2023.163677
  » https://doi.org/10.1016/j.scitotenv.2023.163677
• Pereira GW, Valente DSM, Queiroz DM, Coelho ALF, Costa MM, Grift T. Smart-Map: An open-source QGIS plugin for digital mapping using machine learning techniques and ordinary kriging. Agronomy. 2022;12:1350. https://doi.org/10.3390/agronomy12061350
  » https://doi.org/10.3390/agronomy12061350
• Raja NB, Rajendran PS. A novel fuzzy-based modified GAN and faster RCNN for classification of banana leaf disease. J Inst Eng India Ser A. 2023;104:529-40. https://doi.org/10.1007/s40030-023-00743-8
  » https://doi.org/10.1007/s40030-023-00743-8
• Reem D. The projector algorithm: A simple parallel algorithm for computing Voronoi diagrams and Delaunay graphs. Theor Comput Sci. 2023;970:114054. https://doi.org/10.1016/j.tcs.2023.114054
  » https://doi.org/10.1016/j.tcs.2023.114054
• Rezende JO. Consequências da aplicação de vinhaça sobre algumas propriedades físicas de um solo Aluvial (estudo de um caso) [thesis]. Piracicaba: Universidade de São Paulo, Escola Superior de Agricultura “Luiz de Queiroz”; 1979.
• Santos HG, Jacomine PKT, Anjos LHC, Oliveira VA, Lumbreras JF, Coelho MR, Almeida JA, Araújo Filho JC, Oliveira JB, Cunha TJF. Sistema brasileiro de classificação de solos. 5. ed. rev. ampl. Brasília, DF: Embrapa; 2018.
• Selvaraj MG, Vergara A, Montenegro F, Ruiz HA, Safari N, Raymaekers D, Ocimati W, Ntamwira J, Tits L, Omondi AB, Blomme G. Detection of banana plants and their major diseases through aerial images and machine learning methods: A case study in DR Congo and Republic of Benin. Journal of Photogrammetry and Remote Sensing. 2020;169:110-24. https://doi.org/10.1016/j.isprsjprs.2020.08.025
  » https://doi.org/10.1016/j.isprsjprs.2020.08.025
• Siegel AF, Wagner MR. Variability: Dealing with Diversity. In: Siegel AF, Wagner MR, editors. Practical business statistics. Cambridge: Academic Press; 2022. p. 105-34 https://doi.org/10.1016/B978-0-12-820025-4.00005-1
  » https://doi.org/10.1016/B978-0-12-820025-4.00005-1
• Soil Survey Staff. Keys to Soil Taxonomy. USDA-Natural Resources Conservation Service; 2014.
• Stoorvogel JJ, Orlich RA. An integrated system for precision agriculture in bananas. In: Proceedings of the 5th International Conference on Precision Agriculture. Madison: American Society of Agronomy; 2000. p. 1-14.
• Teixeira PC, Donagemma GK, Fontana A, Teixeira WG. Manual de métodos de análise de solo. 3. ed. rev e ampl. Brasília, DF: Embrapa; 2017.
• Tyagi K, Rane C, Harshvardhan, Manry M. Regression analysis. In: Pandey R, Khatri SK, Singh NK, Verma P, editors. Artificial intelligence and machine learning for EDGE Computing. Cambridge: Academic Press; 2022. p. 53-63. https://doi.org/10.1016/B978-0-12-824054-0.00007-1
  » https://doi.org/10.1016/B978-0-12-824054-0.00007-1
• Uchôa CN, Pozza EA, Pozza AAA, Moraes WS. Modelagem geoestatística da sigatoka-negra e sua relação com a fertilidade do solo. Biosci J. 2011;27:357-62.
• United Nations – UN. Sustainable Development Goals [internet]. 2024 [cited 2024 Apr 10]. Available from: https://brasil.un.org/pt-br/sdgs
  » https://brasil.un.org/pt-br/sdgs
• Uribeetxebarria A, Martínez-Casasnovas JA, Escolà A, Rosell-Polo JR, Arnó J. Stratified sampling in fruit orchards using cluster-based ancillary information maps: a comparative analysis to improve yield and quality estimates. Precis Agric. 2019;20:179-92. https://doi.org/10.1007/s11119-018-9619-9
  » https://doi.org/10.1007/s11119-018-9619-9
• Wallace A. The law of the maximum. Better Crops. 1993;77:20-2.
• Wu F, Duan J, Ai P, Chen Z, Yang Z, Zou X. Rachis detection and three-dimensional localization of cut off point for vision-based banana robot. Comput Electron Agr. 2022;198:107079. https://doi.org/10.1016/j.compag.2022.107079
  » https://doi.org/10.1016/j.compag.2022.107079
• Yeomans JC, Bremner JM. A rapid and precise method for routine determination of organic carbon in soil. Commun Soil Sci Plant Anal. 1988;19:1467-76. https://doi.org/10.1080/00103628809368027
  » https://doi.org/10.1080/00103628809368027
• Zucoloto M, Lima JSS, Coelho RI. Spatial variability of chemical attributes in the soil and yield of banana ‘Prata Ana’. Rev Bras Frutic. 2011a;33:479-84. https://doi.org/10.1590/S0100-29452011000500064
  » https://doi.org/10.1590/S0100-29452011000500064
• Zucoloto M, Lima JSS, Coelho RI, Quartezani WZ. Spatial variability of the granulometric fractions and the production of banana cultivar ‘Prata Anã’. Idesia. 2011b;29:47-52. https://doi.org/10.4067/S0718-34292011000200006
  » https://doi.org/10.4067/S0718-34292011000200006
• Zucoloto M, Souza JS, Coelho RI. Uso da geoestatística na probabilidade de ocorrência do Mal-do-Panamá em bananeira Prata Anã. Rev Fac Nac Agron. 2009;62:4793-6.

Edited by

• Editors:
  José Miguel Reichert https://orcid.org/0000-0002-1163-2893 and Quirijn de Jong Van Lier https://orcid.org/0000-0002-1163-2893

Publication Dates

• Publication in this collection
  16 Dec 2024
• Date of issue
  2024

History

• Received
  03 Jan 2024
• Accepted
  21 May 2024