Next Article in Journal
Balancing Accuracy and Efficiency: HWBENet for Water Body Extraction in Complex Rural Landscapes
Previous Article in Journal
A Real-Time Fusion of Two-Stage Point Cloud Clustering and Saliency Image for Water Surface Object Detection
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 17
Issue 22
10.3390/rs17223709
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Delineation of Management Zones Based on the Agricultural Potential Concept for Potato Production Using Optical Satellite Images

by
David A. Ramirez-Gonzalez
1,2,*,
Karem Chokmani
2,
Athyna N. Cambouris
1 and
Michelle L. D’Souza
3
1
Quebec Research and Development Centre, Agriculture and Agri-Food Canada, 2560 Hochelaga Blvd, Quebec City, QC G1V 2J3, Canada
2
National Institute for Scientific Research (INRS-ETE), 490 de la Couronne Street, Quebec City, QC G1K 9A9, Canada
3
McCain Foods, 8734 Main Street, Unit 1, Florenceville-Bristol, NB E7L 3G6, Canada
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(22), 3709; https://doi.org/10.3390/rs17223709
Submission received: 1 October 2025 / Revised: 7 November 2025 / Accepted: 9 November 2025 / Published: 14 November 2025
(This article belongs to the Section Remote Sensing in Agriculture and Vegetation)
Browse Figures
Review Reports Versions Notes

Highlights

What are the main findings?
  • NDVI composite metrics (mean, standard deviation, skewness) successfully delineated three agricultural potential (AP) zones (high, intermediate, low) that aligned with tuber yield and soil nutrient patterns (total N and C).
  • Multi-year NDVI composites from Sentinel-2 and Landsat 8 effectively captured spatial and temporal variability in crop productivity, demonstrating robustness and scalability across multiple fields.
What are the implications of the main findings?
  • This approach demonstrates that freely available satellite data (Sentinel-2, Landsat 8) can reduce reliance on expensive soil sampling and specialized equipment for MZ delineation, making precision agriculture more accessible to growers.
  • The methodology provides a scalable, cost-effective framework that supports the identification of stable productivity patterns and guiding sustainable management decisions.

Abstract

Management zones (MZs) are a key precision agriculture strategy for managing spatial variability in crops, but conventional delineation methods are costly, time-consuming, and rely on specialized equipment. Previous studies in potato production have primarily relied on single-year NDVI or proximal soil sensor data analyses, limiting their ability to capture temporal stability and variability across multiple fields. This study addresses this gap by applying multi-year, multi-source NDVI composites to characterize spatial and temporal patterns of agricultural potential across 17 commercial potato fields at McCain’s Farm of the Future, Florenceville-Bristol, New Brunswick. A total of 230 NDVI images from Sentinel-2 and Landsat 8 (2015–2023) were processed into composite metrics (mean, standard deviation, skewness) to delineate three agricultural potential (AP) MZs. Validation was conducted using 2023 potato tuber yield and soil physicochemical properties. The results showed statistically significant correlations between NDVI metrics and key soil nutrients (total carbon: |r| < 0.19; total nitrogen: |r| < 0.28), with tuber yield (|r| < 0.41). Spatial patterns of total carbon and nitrogen corresponded with delineated MZs, and tuber yield variability partially aligned with these zones. These findings demonstrate that multi-year NDVI composites provide a cost-effective and scalable approach for mapping agricultural potential, capturing both spatial and temporal variability, and supporting data-driven management decisions in potato production systems.

1. Introduction

Potato (Solanum tuberosum L.) is the most important non-cereal crop in the world, playing a crucial role in global food security as the crop that produces the highest nutrient content per unit area of cultivated land [1]. In Canada, potatoes are the largest vegetable crop, particularly important in Atlantic Canada where it represents 36 to 40% of total production [2]. Improving nutrient management is essential for the sustainability of potato production systems as the excessive use and inadequate application timing of inputs often leads to substantial losses, such as nitrogen leaching into groundwater, resulting in environmental risks and financial waste [3].
Precision agriculture (PA) offers promising solutions to these challenges. Defined as a management approach that integrates temporal, spatial, and plant-specific data with decision-support tools, PA aims to optimize resource use efficiency, productivity, profitability, and sustainability [4]. Since the early 1990s, PA has gained global attention for its ability to improve crop yields while minimizing input use by accounting for within-field variability [5,6]. One of the principal approaches used in precision agriculture is the delineation of management zones (MZs), several discrete, homogeneous areas within the field [7] that have different agricultural potential, such as fertility, water-use efficiency, and yield [8]. These zones can respond differently to inputs, and, consequently, require specific input management to ensure better input placement.
The delineation process uses spatially variable information linked to crop or soil properties in the field. The acquisition of this information is typically obtained through an intensive soil sampling program [9,10,11] including the use of proximal sensors [5,12]. Acquiring this highly effective data requires considerable time and specialized equipment, which is often inaccessible to most growers, thereby hindering the widespread adoption of this technique. In response to these challenges, remote sensing approaches have gained significant attention in recent years. Among them, optical satellite images have emerged as a promising solution for addressing data accessibility issues in PA. This technology enables the mapping of large areas at relatively low investment costs [13].
In potato production research, the delineation of MZs relied on proximal sensors such as the Geonics EM38 [14] and the VERIS 3100 models [12], which generate detailed georeferenced data to estimate indirect soil properties like apparent electrical conductivity (ECa). These measurements are used to characterize within-field variability [5,12,14]. ECa has been widely adopted due to its relative temporal stability [14] and shows strong relationships with several soil characteristics [15]. Recent studies have applied data fusion approaches combining soil ECa with complementary datasets such as topography, yield maps, and satellite imagery to enhance spatial variability characterization and improve the MZ delineation [16,17].
One popular technique is the application of vegetation indices (VIs). VIs highlight spectral features in images, allowing for the analysis of crop phenotypic traits by combining reflectance information from multiple image bands [18]. The Normalized Difference Vegetation Index (NDVI) is the most extensively studied index in agricultural applications [19], as it is closely correlated with biomass production, yield, and vegetative health. Additionally, NDVI should provide insights into the yield of potato crops by assessing on the functionality of detected leaves, essential for capturing light and carbon dioxide (CO2), the key components for the growth of storage organs, such as tubers [20], most NDVI studies focus on nutrient management and yield prediction, including yield prediction models, for cereals such as wheat [20], rice [21] and corn [22]. They demonstrate that VIs can be effectively used in the MZ delineation process, providing complementary data in productivity maps [13]. In the case of tuber crops like potatoes, as they accumulate most of their biomass underground, VIs such as NDVI, primarily detects above-ground traits, may have limited accuracy, potentially leading to mismatches in yield data and affecting crop characterization for PA. Recent studies have also investigated the adjustment of soil-related parameters in vegetation indices to mitigate saturation effects and improve the estimation of canopy biophysical variables, such as leaf area index, from multi-sensor satellite time series [23]. Despite these limitations, some studies have used NDVI for yield predictions [20,24,25] for MZ delineation in potato [16]. However, research on satellite-based vegetation indices for this MZ delineation remains scarce.
The aggregation of VI time series in addition to the merger of multiple sources of satellite data improve the delineation process by accounting for crop and yield dynamics [26]. For example, Řezník et al. [27] used the Enhanced Vegetation Index (EVI) time series derived from Landsat 8 and Sentinel-2 multispectral satellite to identify yield productivity zones for cereals, their study showed strong correlations between yield zones and actual productivity, demonstrating the potential of these zones for long-term yield forecasting and decision-making.
One of the techniques used to optimize VI time series analyses is image composition, primarily employed to enhance change detection at regional scales [28], this technique reduces the high-volume data required for time series analyses while conserving the essential information needed to identify characteristics in the field. Some also use metrics derived from composite vegetation index images as a robust alternative to reduce the noise in the classification process. For example, Maxwell and Sylvester [29] used the maximum and standard deviation of the NDVI temporal composite derived from Landsat imagery for the classification of ever-cropped land. Lanucara et al. [30] integrated composite VIs, including NDVI and SAVI2 (modified soil-adjusted vegetation index), climate data, soil and crop properties for MZ delineation at a regional level in a web-based spatial decision support system.
To our knowledge, no studies have yet applied a multi-source, multi-year NDVI composite approach for the delineating MZs in potato production, nor extended this analysis beyond individual field scales. This gap highlights the need for approaches that integrate temporal and spatial crop information over multiple seasons and fields to more effectively capture the variability associated with agricultural potential. Indeed, larger scales may enhance the delineation process by incorporating the inter-field variance, a factor not considered in the traditional field-level MZ approach. We hypothesize that crop development will adequately represent the key factors that affect agricultural potential (AP)—productivity per unit area per unit time [31], such as meteorological conditions, soil physicochemical characteristics, and agricultural management practices (e.g., fertilization). Therefore, VIs derived from optical satellite imagery offer an effective method for capturing AP variation within a given field and therefore delineating MZs, due to the connection between vegetation’s spectral properties and its biochemical and biophysical attributes [32]. Consequently, time series analysis of the vegetation indices in a field could effectively capture crop temporal dynamics at a regional level and provide a stronger delineation approach by considering factors not evaluated by traditional methods like soil sampling and proximal sensing.
The objective of this study is to develop a methodology for delineating MZs based on the concept of AP for commercial potato fields using a multisource diachronic analysis of optical satellite images. An eight-year (2015–2023) NDVI time series of Sentinel-2 and Landsat 8 images were used across 17 fields, and the delineation results were validated using potato tuber yield data and the soil physicochemical properties obtained from soil sampling.

2. Materials and Methods

2.1. Experimental Site

This study was conducted across 17 fields in Florenceville-Bristol, New Brunswick, (46°24′50″N, 67°36′22″W) (Figure 1). These fields are part of the Farm of the Future Canada (FOF CA), one of three Farms of the Future owned and operated by McCain Foods, aimed at advancing sustainable farming practices and exploring innovative agricultural technologies. A total of 170.6 ha was considered, with fields ranging in size from 3.8 and 26.2 ha. The study region is dominated by the production of potatoes (e.g., Caribou Russet and Russet Burbank cultivars), cereals (corn, barley, wheat, soybean), and cover crops (alfalfa, ryegrass, red clover) with a growing season lasting from May to early November. Since 2021, FOF CA has implemented regenerative agriculture practices, such as increasing crop diversity in the rotation, minimizing soil disturbance with low tillage, and protecting soils with cover crops. For confidentiality purposes, the fields used to validate tuber yield were anonymized and referred to as Fields X, Y, and Z.

2.2. Data Collection

2.2.1. Soil Sampling and Analysis

The soil sampling was conducted from 2020 to 2022 using a grid design of 50 m by 50 m with a sampling density of 4 samples ha−1. A total of 673 points were sampled at a sampling depth of 15 cm for soil chemical properties and 50% of points were considered for soil texture analysis. Soil samples were air-dried, ground, and passed through a 2 mm sieve before extraction using the Mehlich-3 solution at a soil–solution ratio of 1:10 [33]. The concentrations (mg kg−1) of phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), and aluminum (Al) in the extract were determined by inductively coupled plasma optical emission spectroscopy (ICP–OES; Model 4300DV, PerkinElmer, Shelton, CT, USA). Total nitrogen (total N, mg kg−1) and total carbon (total C, mg kg−1) were measured using an Elementar vario MAX CN analyzer (Elementar Analysensysteme GmbH, Hanau, Germany). Soil particle size distribution was determined using the pipette method [34]. Finally, soil pH was measured in distilled water with a 1:1 soil-to-water ratio [35].

2.2.2. Topography

Georeferenced elevation measurements for the study area were provided by FOF CA based on previous soil surveys of electrical conductivity, conducted using a GPS-enabled proximal sensor in 2022 and 2023. The data underwent a cleaning process, excluding outliers beyond three standard deviations from the mean. The cleaned data was then resampled to a 1 m resolution to produce a digital elevation map, and a slope map was subsequently generated using the Slope function in ArcGIS version 10.8.1 (ESRI, Redlands, CA, USA).

2.2.3. Tuber Yield

The spatial distribution of tuber yield of the potato Caribou Russet cultivar was obtained from three fields (X, Y, Z) (Figure 1) during the 2023 harvest season using a GreenSeeker (RiteYield system, Greentronics, Elmira, ON, Canada) yield monitor system. The raw tuber yield data underwent a cleaning process in which outliers were excluded (measurements exceeding the mean plus or minus three standard deviations were discarded). Yield maps were generated using ordinary kriging interpolation at a 1 m grid resolution using the Geostatistical wizard of ArcGIS (ESRI, Redlands, CA, USA) followed by a smoothing process using an average filter with a 30 m moving window.

2.2.4. Satellite Images

Freely accessible multispectral Sentinel-2 and Landsat 8 remote sensing data were obtained from the Harmonized Sentinel-2 MSI (MultiSpectral Instrument, Level-1C) and USGS Landsat 8 Collection 2 Tier 1 TOA (Top Of Atmosphere) Reflectance, available in Google Earth Engine (GEE) repositories (Table 1). The red (R) and near-infrared (NIR) bands were extracted for the NDVI calculation with spatial resolutions of 10 m and 30 m for Sentinel-2 and Landsat 8, respectively. NDVI monitoring was conducted for July (highest vegetative development for the potato crop, flowering growth stage) from 2015 to 2023.
July imagery was selected because, under the study conditions in Canada, this month corresponds to the stage of maximum ground cover and tuber initiation, when NDVI is most strongly correlated with yield and biomass [36,37,38]. For most potato varieties, this stage occurs approximately 60–80 days after planting [38], which corresponds to July under the study site’s growing season (May to November). A total of 230 images were processed and used to create composite images at a nominal 10 m spatial resolution. Due to crop rotation practices on the farm, the images included multiple crops in addition to potatoes, such as cereals and cover crops.

2.3. Methodological Workflow

The proposed methodological workflow (Figure 2) consists of four components: (1) preprocessing of satellite images, (2) image composition, (3) AP MZ delineation, and (4) validation.

2.3.1. Preprocessing Satellite Imagery

The image collections from Harmonized Sentinel-2 MSI (Level-2A) and the USGS Landsat 8 Collection 2 Tier 1 TOA Reflectance were filtered using general spatial and temporal filtering methods in GEE. These repositories contain orthorectified TOA images that were atmospherically corrected using the Sensor-Invariant Atmospheric Correction. This technique helps minimize the discrepancies in atmospheric caused by atmospheric effects on time-series satellite imagery [39]. The SIAC method was selected because it does not rely on sensor-specific optical band configurations, making it suitable for harmonizing imagery from different sensors. Applying the same correction method to both Sentinel-2 and Landsat 8 ensures better or greater radiometric consistency across datasets and reduces atmospheric discrepancies. This approach has been successfully validated for both sensors [40] and applied in multi-sensor harmonization for agricultural studies using Sentinel-2 and Landsat 8 imagery [39].

2.3.2. NDVI Image Composition

First, the corrected images were used to calculate the NDVI. NDVI was selected because it is the most used VI for potato-related studies as it can indicate plant health (especially of the crown), and compensate for changing illumination conditions, surface slope, and viewing angle [41]. NDVI was calculated using Equation (1), where the R and NIR bands of the electromagnetic spectrum, correspond to the wavelength 0.695 µm and 0.842 µm for Sentinel-2 and 0.631–0.692 µm and 0.772–0.898 µm for Landsat 8.
N D V I = N I R − R N I R + R
The resulting NDVI images were stacked by year, and an image composite was created from each set. The selected composition approach was the Maximum Value Composite (MVC). The MVC was applied separately for each year to reduce cloud contamination and capture peak vegetation conditions within each annual composite. The MVC procedure examines a series of multi-temporal satellite data during the compositing period and, by analyzing each value on a pixel-by-pixel basis, retains only the highest value for each pixel location [28,30,42], eliminating generally low NDVI values (NDVI < 0.2) caused by clouds and their shadows [29].
The R and NIR bands were extracted for the calculation of the NDVI with a spatial resolution of 10 m and 30 m for Sentinel-2 and Landsat 8, respectively. NDVI monitoring was conducted for July (the period of highest vegetative development for the potato crop, corresponding to the flowering growth stage) from 2015 to 2023. A total of 230 images were processed and used to create composite images at a nominal spatial resolution of 10 m. Due to crop rotation practices, the images included multiple crops in addition to potato crop, such as cereals and cover crops. The images were stacked into a single file. These composite images are cloud-free and generally correspond to conditions of maximum vegetative development, which could be highly valuable in assessing AP. The resulting annual NDVI MVC images from 2015 to 2023 were used to generate three maps from statistical metrics based on the annual NDVI MVC: mean value (central tendency), standard deviation (dispersion of MVC around the mean value), and skewness index (asymmetry of the MVC distribution). These parameters were obtained through pixel-by-pixel processing to capture the temporal behavior of vegetation in the study area. All NDVI image processing and stacking were conducted in GEE. To mitigate the border effect in the fields, a 10 m buffer was applied to eliminate the areas corresponding to pathways or zones of no production, which resulted in low NDVI values.

2.3.3. MZ Delineation

The unsupervised k-means clustering algorithm was used in the delineation process [43]. For each resulting image of the NDVI MVC-derived metrics (mean, standard deviation and skewness), the pixels from all fields were pooled into a single array, disregarding their geographic positions and passed through the k-means clustering algorithm as an input variable to classify them into three categories. This choice was informed by findings from previous field-scale studies in similar contexts rather than by optimization methods. Prior research on potato fields has identified optimal numbers of management zones ranging from two [12,14,16] to three [17,44]. At the farm scale, three zones were chosen as it is expected to provide a practical balance between capturing soil and yield variability while maintaining operational feasibility. Since this algorithm is sensitive to the range of the input variables [45], the Min Max normalization method was applied (Equation (2). This method adjusts the scale of the dataset’s minimum and maximum values to a range of 0 to 1. Classified pixels were then transformed from array format to map format, creating a single map consisting of three categories (classes). Following the preliminary delineation process, the zones underwent a smoothing procedure using the Sieve filter module available in QGIS to remove isolated pixel clusters and generate homogeneous zones.
Z = X − M i n ( X ) M a x X − M i n ( X )

2.3.4. Interpretation of the MZs Based on the AP Concept

The delineated MZs based on the NDVI MVC to represent the AP were interpreted using two criteria linked to the NDVI MVC-derived metrics: (1) vegetative development and (2) temporal stability. A stronger AP should not only be associated with better vegetative development, but it should also remain consistent over time within the zone. Vegetative development was assessed using the mean metric of the NDVI MVC, while temporal stability was evaluated through the standard deviation and skewness metrics. Specifically, the standard deviation measures the variability of vegetative development; a lower standard deviation indicates greater temporal stability in the field. NDVI MVC skewness metric, in turn, reflects the direction of this variability. A negative skewness suggests a prevalence of years with high production, whereas a positive skewness suggests more years of low production, both of which signal instability in vegetative development. Thus, a skewness value closer to zero implies better temporal stability, meaning that atypical data points are distributed more evenly around the average. This indicates more homogeneous vegetation over time within the selected region.

2.3.5. Validation of Management Zones (MZs)

The Pearson correlation coefficient (r) was analyzed to examine the relationships between NDVI MVC metrics and the corresponding soil physicochemical properties, slope, and tuber yield, evaluating the relationships between NDVI’s temporal variability and the validation variables. Data for the correlation evaluation was extracted from the NDVI composition maps and the tuber yield map using the coordinates of the soil samples as extraction points. The correlations were performed using the average tuber yield within a 3.4 m diameter around soil sampling locations. This diameter was chosen because it matches the width of the pass made by the potato planter, with each pass consisting of four rows in the FOF CA potato production system.
The average tuber yield within each zone for the fields with potatoes in 2023 (fields X, Y, and Z in Figure 1) was compared to validate the interpretation of the AP as defined by the study’s approach. Additionally, an intra-field comparison of mean tuber yield values was conducted on the fields with two or more AP MZs to further evaluate yield differences.
An analysis of variance was conducted using NDVI MVC metrics and soil physicochemical properties to evaluate the statistically significant differences among the AP MZs. For normally distributed parameters, an ANOVA combined with an LSD multiple comparison test (p-value < 0.05), while for non-normally distributed parameters, the Kruskal–Wallis test was used [44].

3. Results

3.1. Agronomic Interpretations of Obtained MZs Based on AP

For the entire farm, a map of three MZs was obtained. Applying this classification, fields were classified into three AP MZs using NDVI MVC metrics as criteria (Figure 3; Table 2) referred to as high, intermediate, and low AP based on the specific average value of the NDVI MVC mean metric (Table 2). All the delineation parameters (NDVI MVC metrics) showed statistically significant differences between zones (p < 0.05). The high AP MZ was characterized by a higher mean of the NDVI MVC metric, reflecting greater vegetative development. However, the NDVI MVC mean metric alone did not fully account for zonal variability, as the differences between the high and intermediate AP MZs were minimal. Notably, the intermediate AP MZ demonstrated greater temporal variability, with higher NDVI MVC standard deviation and skewness metrics values (NDVI MVC Mean = 0.87, NDVI MVC Standard Deviation = 0.05, NDVI MVC Skewness = −0.94) compared to the high AP MZ (NDVI MVC Mean = 0.88, NDVI MVC Standard Deviation = 0.03, NDVI MVC Skewness = −0.28). This suggests that the intermediate AP MZ experienced greater fluctuations in crop development over time. In contrast, the low AP MZ showed the lowest mean, the highest standard deviation, and a skewness value furthest from zero metrics (NDVI MVC Mean = 0.83, NDVI MVC Standard Deviation = 0.09, NDVI MVC Skewness = −1.24). This indicates weak and unstable vegetative development in these areas.

3.2. Validation of Methodology for AP MZs

3.2.1. Correlations Between Delineated Properties (NDVI MVC Metrics) and Validation Parameters

Several significant correlations (p ≤ 0.05) were identified between NDVI MVC metrics for each MZ and the corresponding soil properties (Table 3). While most correlations were weak (r < 0.25), they were statistically significant for multiple soil physicochemical properties, including sand content, total N, total C, pH, K, Mg, and slope. The strongest observed correlation was with tuber yield (r ≥ 0.25), which showed positive associations with NDVI MVC mean and NDVI MVC skewness metrics, and negative correlations with the NDVI MVC skewness metric. Total N and total C were significantly correlated with both the NDVI MVC standard deviation and NDVI MVC skewness metrics. Among the Mehlich-3 extractable elements, K was significantly correlated with NDVI MVC mean and NDVI MVC skewness metrics, while Mg showed significant correlations with both the NDVI MVC mean and NDVI MVC standard deviation metrics. Regarding soil texture, sand content was the only component that exhibited significant correlations with NDVI MVC mean and NDVI MVC standard deviation metrics.

3.2.2. Tuber Yield and AP MZs

Tuber yield maps for fields X, Y, and Z were created using ordinary kriging interpolation at a 1 m grid resolution. The exponential model was the best fit for modeling the spatial variability of the tuber yield for all the fields (Table 4). The nugget ratio varied from 23% to 38%, while the spatial range varied from 18 to 36 m across the fields. These results indicate strong (≤25%) and moderate spatial dependence (25–75%), in accordance with the classification proposed by Cambardella, Moorman [46]. The RMSES values (greater than 0.81) confirm the quality of the model fits for each field.
Overall, trends of tuber yield across fields aligned well with the AP MZs created (Figure 4). Higher yields generally occurred in the high AP MZ, as observed in field Z, where most of the area consistently had intermediate to high yields. In field Y, intermediate yields were concentrated in the west, while the highest yields were found in the east of the field. In field X, three MZs were identified, showing partial agreement with the yield map. For example, in the southwestern part of the field, a high yield aligns with the high AP MZ. Meanwhile, in the northeastern part, the yield showed intermediate to high production, which corresponded to the intermediate AP MZ. However, the low AP MZ, which corresponded to an intermediate yield in the field, did not show a clear visual correspondence.
Statistical analysis using Tukey and Kruskal–Wallis tests revealed that the high AP MZ produced a significantly greater mean tuber yield (60.1 Mg ha−1) compared to the intermediate (53.5 Mg ha−1) and low (45.7 Mg ha−1) AP MZs (p < 0.05) (Table 5). However, in field X—where all three AP MZs were present (high: 61.2 Mg ha−1, intermediate: 56.4 Mg ha−1, low: 45.7 Mg ha−1)—no significant difference was observed between the intermediate and low AP MZs (p > 0.05).
The absence of significant differences in field X may be attributed to the increase in tuber yield variability (characterized by the coefficient of variation, CV) within the zones, particularly in the high AP zone, where CV nearly doubled compared with the overall value (24.0% vs. 13.7%). A slight increase was also observed in the intermediate zone (23.0% vs. 21.5%), while the low zone remained stable (10.8%). These differences suggest that the high AP zone, which typically shows an intermediate level of variability, became much more heterogeneous in field X. This increase in CV likely reduced the statistical power of the post hoc test, thereby limiting the ability to detect differences among the AP MZs (Table 5).

3.2.3. AP MZs Soil Physicochemical Properties and AP MZs

The comparison of physicochemical properties across the AP MZs revealed several significant associations (Table 6), particularly for key nutrients related to potato tuber yield and soil fertility, particularly total N and total C. However, no significant differences (p > 0.05) in physicochemical properties were observed between the intermediate and low AP MZs, suggesting a degree of similarity in soil nutrient status and fertility between these two zones.
No consistent trend was observed for the Mehlich-3 extractable elements. For example, P and Mg were lowest in the high AP MZ and highest in the intermediate zone. Conversely, K was the lowest in the intermediate zone and higher across both the low and high zones. Higher total N (0.22%) and total C (2.45%) were associated with the high AP MZ (Table 2). Indeed, total C displayed similar spatial behavior to total N, with AP MZs averages ranging from 2.45 to 2.12% for total C, and from 0.22 to 0.19% for total N. The pH in the high MZ was slightly lower than the other two MZs, values ranging from 5.77 to 5.90.
Clay and sand content differed between AP MZs; however, while clay content increased from high to low AP MZ, no trend was observed for sand. Indeed, all three AP MZs are classified as loamy soils by the USDA’s soil taxonomy [47]. Slope did not differ significantly between zones, suggesting that topographic variability was not effectively captured.

4. Discussion

4.1. NDVI Derived Metrics Are a Valuable Proxy of Agricultural Potential as Evidenced by Correlations to Yield and Soil Properties

Higher tuber yields tended to occur in areas showing more vigorous and consistent crop development, as indicated by higher mean and lower temporal variability of NDVI MVC-derived metrics (Table 2). Although these correlations are weak, they suggest that NDVI metrics may serve as useful, though indirect, indicators of crop performance over time. Notably, the positive correlation with NDVI MVC skewness metric—interpreted here as a tendency toward left-skewed distributions—suggests that areas exhibiting earlier and more sustained vegetative growth tend to yield higher production.
The temporal patterns in NDVI MVC-derived metrics also provide insights into the stability of crop development. Zones with NDVI MVC skewness values near zero and lower NDVI MVC standard deviation metric values indicate more uniform and stable vegetative dynamics, reflecting favorable growing conditions that are likely influenced by a combination of nutrient availability, soil texture, and environmental factors.
The sensitivity of NDVI-based temporal metrics to underlying soil nutrient conditions was reflected in statistically significant, though modest, correlations with total N, total C, K, and Mg. These patterns indicate that NDVI MVC metrics capture general trends in soil fertility rather than precise quantitative relationships. In particular, high levels of total N and total C—recognized as key determinants of potato productivity [14,48,49]—likely support sustained canopy development and reduce temporal NDVI variability. In addition, the correlation between higher NDVI MVC mean metric values and higher total N content further suggests a link between vegetation vigor and nitrogen availability. This aligns with previous studies that emphasize the role of N as the most yield-determining macronutrient in potato production [41,50,51,52]. Total C trends followed those of total N, which is expected due to their shared origin in organic matter. Higher total C may indicate greater soil organic matter (SOM), a key factor in enhancing soil fertility through improved water-holding capacity, nitrogen mineralization, and aggregate stability [15]. Furthermore, the measured ranges for total N and total C closely matched values reported for potato fields in New Brunswick by Perron et al. [12], reinforcing the validity of the observations.
The pH remained remarkably stable, both within and across the AP MZs, corroborating earlier findings that pH generally exhibits low spatial variability due to its logarithmic nature [53,54]. While the observed pH values (5.8–5.9) are slightly below the optimal range for potato growth [55], their consistency may still provide a suitable environment for crop development. Finally, while statistically significant, the observed differences in clay and sand content across zones were not agronomically relevant, as all areas exhibited comparable textures within the loamy class. This texture is widely recognized as optimal for potato cultivation due to its favorable balance between drainage capacity and nutrient retention [56]. However, prior studies have noted that even small variations in clay content can impact water retention and drainage efficiency [12], which could subtly influence crop performance. Nevertheless, the observed relationship with sand content underscores the role of soil texture in shaping vegetation dynamics.
The observed yield trends aligned well with the MZ delineation, particularly in distinguishing high- from lower-yielding zones. Indeed, despite the lack of statistical significance, the intermediate AP MZ consistently exhibited higher average tuber yield than the low AP zone suggesting that the methodology was still capable of capturing most of the spatial and temporal yield variability in the studied fields. Given that the validation grid (50 m) exceeded the semivariogram range (36 m) (Table 4), the samples can be considered spatially independent, thereby supporting the validity of the statistical comparisons used for validation (Tukey/Kruskal–Wallis test). Notably, the field adjacent to Field Z demonstrated a strong spatial relationship with a subsection that differentiates between two-year and three-year crop rotations (FOF CA communication, Figure 3). Even though certain years included rotation crops instead of potatoes, the significant correlations between the AP MZs and tuber yield highlights the robustness of the methodology.
These results suggest that the NDVI-based temporal composite method is a potentially reliable tool for delineating MZs, as the NDVI MVC metrics captured key aspects of the underlying spatial variability and agricultural potential, even when the composite incorporated multiple crops in certain years, demonstrating its ability to reflect both short-term variability and longer-term patterns in crop performance. This approach enables data integration across years and crop types which may provide resilience against seasonal fluctuations (year to year variation) and may offer a practical solution for data-limited situations. Overall, the three delineated zones show a strong alignment with the yield patterns, supporting the practical utility of this simplified approach, that is, the use of NDVI MVC metrics as a valuable proxy of AP for identifying spatial zones of agronomic significance within a field.

4.2. Methodological Limitations and Areas of Improvement

A key limitation of this study is the use of only one year of potato tuber yield data for validation. Multi-year yield data, encompassing multiple crop types would have provided a more robust basis for evaluating the temporal stability and agronomic relevance of the delineated AP zones. Access to long-term, high-quality yield data remains a challenge as yield maps are often affected by errors related to field machinery operations, such as irregular combine harvester paths [57,58] and this issue is further exacerbated in the case of yield time series as highlighted by previous studies [59,60,61]. Also, the inclusion of rotational crops in NDVI composites likely introduced additional noise, partially explaining the moderate correlations observed between NDVI MVC metrics and yield or soil properties.
Temporal and crop rotation effects may have further influenced the correspondence between MZs and yield. Inter-annual climatic variability—including precipitation, temperature, and growing season length—can alter vegetation dynamics and NDVI signals. Similarly, rotational crops with differing canopy structures and rooting systems can modify soil properties and vegetation responses, potentially shifting the spatial boundaries of MZs over time. While the use of multi-year NDVI composites was used to integrate these effects and improve the characterization of agricultural potential patterns, temporal alignment between remote sensing imagery and yield observations remains critical for improving the robustness of management zone delineation.
Spatial resolution and smoothing also introduce limitations. The MZs were delineated at a 10 m resolution, while yield maps were generated at 1 m, leading to a loss of spatial detail and generalized intermediate values. The smoothing applied to create agronomically homogeneous zones further reduces sensitivity to small-scale variability. In addition, the methodology showed limited ability to capture differences, especially between the intermediate and low AP MZs for most soil properties may reflect the intrinsic limitations of the delineation approach in capturing subtle soil property gradients or the relatively homogeneous nature of these zones. Also, the lack of significant slope variation suggests that the spatial resolution and smoothing applied in the methodology may limit its sensitivity to topographic features. This may limit the methodology’s ability to capture such as water runoff and redistribution, which are influenced by terrain [62,63].
At the field level, discrepancies were observed. In Field Y, the intermediate AP MZ did not correspond to local yield patterns, likely due to resolution mismatches. In Field X, no statistically significant differences were detected among zones, even when accounting for in-field variability, and the low AP MZ occasionally corresponded to intermediate yields. This mismatch may be partly explained by residual effects from rotation crops, localized nutrient redistribution, or unaccounted for management practices influencing crop growth. Another contributing factor is that the classification was based on a broader dataset which included fields beyond those evaluated in the yield analysis. This wider spatial scope may have reduced the representativeness of the MZs for specific field-level conditions, making the yield maps less reflective of local variability across the study area. Finally, the constraint to three AP MZs may limit the resolution of spatial variability; using more classes (e.g., five zones) could improve precision.
Despite these limitations, the three delineated zones aligned well with yield patterns, and significant correlations between AP MZs and tuber yield (r ≥ 0.25, p ≤ 0.05) highlight the robustness of the methodology. NDVI MVC metrics partially capture much of the underlying spatial variability and agricultural potential patterns, even under multi-year, multi-crop scenarios. Notably, distinct patterns were detected in the main physicochemical properties, suggesting that further research is needed to determine the optimal conditions and temporal evaluation window for improving zone separation. Such refinements would enhance the methodology’s capacity to identify zones where vegetation develops most favorably and consistently, even under multi-crop scenarios such as crop rotation.

4.3. Practical Implications of the Methodology

The AP MZs identified in this study have strong practical value for precision and regenerative potato production. By integrating multi-year NDVI MVC metrics, producers can identify areas of high and low productivity potential and adjust management accordingly. High AP zones, showing stable vegetation and higher total N and C, can be managed with lower input intensities and focused nutrient monitoring. Intermediate AP zones benefit from adaptive management, such as variable-rate fertilization and targeted irrigation, while low AP zones—linked to weaker vegetation and greater variability—require soil structure improvement and organic matter restoration through practices like cover cropping. Ultimately, linking AP zones to regenerative practices can enhance sustainability outcomes.
Although the NDVI-based delineation enables spatially targeted management and optimized resource use, its implementation requires data processing and coding skills beyond most commercial farms. This is where McCain Foods and similar businesses play a crucial role, centralizing the processing of satellite data and generating ready-to-use AP maps for growers. By embedding this methodology into grower-facing digital solutions, McCain can translate complex geospatial analyses into actionable field recommendations that support variable-rate inputs, irrigation planning, and soil health improvement. The approach would also support seasonal decision-making: early-season NDVI for stand establishment, mid-season for nitrogen and irrigation adjustments, and post-harvest for assessing residue and cover crop performance.
Overall, this methodology offers a scalable, cost-effective, and science-based decision-support framework that bridges satellite monitoring with farm management. By leveraging McCain’s technical capacity and grower network, it can transform complex remote-sensing outputs into practical, regionally tailored guidance that improves both profitability and environmental performance in potato production.

5. Conclusions

This study presented a methodology utilizing NDVI composite time series as a tool for delineating AP MZs in potato production systems using multiple sources of satellite data. Despite tuber-yield-limiting factors and the lack of crop differentiation, the methodology successfully identified trends in key physicochemical soil properties associated with potato fertility, such as total C and total N. Additionally, the methodology detected spatial trends in tuber yield with moderate accuracy, though performance varied across fields especially at in-field scale.
These findings address a key research gap: the development of a scalable, cost-effective approach for assessing spatial patterns of agricultural potential without reliance on extensive soil sampling. By leveraging multitemporal, multi-source, and multi-field-scale satellite-derived vegetation indices, the methodology proposed a practical alternative to traditional, sampling-intensive approaches, enabling the MZ delineation in a more accessible and efficient manner.
Moreover, the method provides valuable tools for spatially assessing agricultural potential and supporting data-driven decision making in crop management. Further research into VI-composite-based approaches is highly encouraged, particularly to evaluate the effects of crop variability, the influence of delineation variables, and the temporal and spatial limitations of the methodology.

Author Contributions

Conceptualization, D.A.R.-G., K.C. and A.N.C.; methodology, D.A.R.-G., K.C., A.N.C., formal analysis and interpretation, D.A.R.-G., K.C. and A.N.C.; supervision, K.C. and A.N.C.; writing—original draft preparation, D.A.R.-G.; writing—review and editing, D.A.R.-G., K.C., A.N.C. and M.L.D. All authors have read and agreed to the published version of the manuscript.

Funding

This research was fully supported by Agriculture and Agri-Food Canada (ASP-175—Regenerative agriculture mitigating soil degradation and climate change challenges # J-002703), which funded all infrastructure and analytical components of the project, as well as provided financial support to the candidate throughout their master’s studies. Additional support for the candidate’s master’s work was provided by INRS through service revenues and the remaining funds from Professor Karem Chokmani’s research projects.

Data Availability Statement

Dataset available on request from the authors. The raw data supporting the conclusions of this article will be made available by the authors on request.

Acknowledgments

We would like to thank Andrée-Dominique Baillargeon, Camille Lambert-Beaudet, Bilal Javed, Jeff Daniel Steeven Nze Memiaghe, and Audrey-Kim Minville (all AAFC Quebec) for field assistance and technical support as well as the farm staff at the Farm of the Future Canada for their dedication in carrying out the day-to-day farming operations that made this research possible. We would also like to acknowledge the commitment of McCain Foods Limited and their collaborative efforts in understanding and demonstrating the benefits of regenerative agricultural practices through the Farms of the Future initiative.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
MZsManagement Zones
NDVINormalized Difference Vegetation Index
APAgricultural Potential
PAPrecision Agriculture
VisVegetation Indices
EVIEnhanced Vegetation Index
SAVI2Modified Soil-Adjusted Vegetation Index
FOF CAFarm of the Future Canada
PPhosphorus
KPotassium
CaCalcium
MgMagnesium
AlAluminum
Total NTotal Nitrogen
Total CTotal Carbon
GEEGoogle Earth Engine
RRed
NIRNear-Infrared
TOATop Of Atmosphere
MVCMaximum Value Composite

References

  1. FAO. FAO Statistical Yearbook 2013 World Food and Agriculture; FAO: Rome, Italy, 2013. [Google Scholar]
  2. Agriculture and Agri-Food Canada. Potato Market Information Review 2021–2022; Agriculture and Agri-Food Canada: Ottawa, ON, Canada, 2022. [Google Scholar]
  3. Davenport, J.R.; Milburn, P.H.; Rosen, C.J.; Thornton, R.E. Environmental impacts of potato nutrient management. Am. J. Potato Res. 2005, 82, 321–328. [Google Scholar] [CrossRef]
  4. International Society of Precision Agriculture. Precision Agriculture Definition. 2024. Available online: https://www.ispag.org/resources/definition (accessed on 1 May 2025).
  5. Cambouris, A.N.; Zebarth, B.J.; Ziadi, N.; Perron, I. Precision agriculture in potato production. Potato Res. 2014, 57, 249–262. [Google Scholar] [CrossRef]
  6. Faulin, G.D.; Molin, J.P.; Magalhães, R.P. Variabilidade Espacial do Teor de Água e sua Influência na Condutividade Elétrica do Solo. Master’s Thesis, Escola Superior de Agricultura Luiz de Queiroz, Piracicaba, Brazil, 2005. [Google Scholar]
  7. Ali, A.; Rondelli, V.; Martelli, R.; Falsone, G.; Lupia, F.; Barbanti, L. Management zones delineation through clustering techniques based on soils traits, NDVI data, and multiple year crop yields. Agriculture 2022, 12, 231. [Google Scholar] [CrossRef]
  8. Mulla, D.J. Soil spatial variability and methods of analysis. In Soil Crop and Water Management in the Sudano-Sahelian Zone; ICRISAT: Telangana, India, 1989; pp. 241–252. [Google Scholar]
  9. Basnyat, P.; McConkey, B.G.; Selles, F.; Meinert, L.B. Effectiveness of using vegetation index to delineate zones of different soil and crop grain production characteristics. Can. J. Soil Sci. 2005, 85, 319–328. [Google Scholar] [CrossRef]
  10. Karlen, D.; Sadler, E.; Busscher, W. Crop yield variation associated with Coastal Plain soil map units. Soil Sci. Soc. Am. J. 1990, 54, 859–865. [Google Scholar] [CrossRef]
  11. Long, D.; Carlson, G.; DeGloria, S. Quality of field management maps. In Site-specific Management for Agricultural Systems; Wiley Online Library: Hoboken, NJ, USA, 1995. [Google Scholar]
  12. Perron, I.; Cambouris, A.N.; Chokmani, K.; Gutierrez, M.F.V.; Zebarth, B.J.; Moreau, G.; Biswas, A.; Adamchuk, V. Delineating soil management zones using a proximal soil sensing system in two commercial potato fields in New Brunswick, Canada. Can. J. Soil Sci. 2018, 98, 724–737. [Google Scholar] [CrossRef]
  13. Damian, J.M.; Pias, O.H.D.C.; Cherubin, M.R.; Fonseca, A.Z.D.; Fornari, E.Z.; Santi, A.L. Applying the NDVI from satellite images in delimiting management zones for annual crops. Sci. Agric. 2019, 77, e20180055. [Google Scholar] [CrossRef]
  14. Cambouris, A.; Nolin, M.; Zebarth, B.; Laverdière, M. Soil management zones delineated by electrical conductivity to characterize spatial and temporal variations in potato yield and in soil properties. Am. J. Potato Res. 2006, 83, 381–395. [Google Scholar] [CrossRef]
  15. Khan, H.; Farooque, A.A.; Acharya, B.; Abbas, F.; Esau, T.J.; Zaman, Q.U. Delineation of management zones for site-specific information about soil fertility characteristics through proximal sensing of potato fields. Agronomy 2020, 10, 1854. [Google Scholar] [CrossRef]
  16. Hendawy, E.A.; Mohamed, O.H.; Abou-Hadid, A.F.; El-Shinawy, M.Z.; Belal, A. Delineation of Management Zones for Site-Specific Management of Potato Crop in Some Areas in Western Nile Delta, Egypt. Egypt. J. Soil Sci. 2024, 64, 1433–1448. [Google Scholar] [CrossRef]
  17. Javed, B.; Cambouris, A.N.; Duchemin, M.; Ziadi, N.; Karam, A. Management zone delineation: Utilizing multiple data sources to minimize soil spatial variability in commercial potato fields under Prince Edward Island pedoclimatic conditions. Can. J. Soil Sci. 2024, 105, 1–18. [Google Scholar] [CrossRef]
  18. Lin, Y.; Li, S.; Duan, S.; Ye, Y.; Li, B.; Li, G.; Lyv, D.; Jin, L.; Bian, C.; Liu, J. Methodological evolution of potato yield prediction: A comprehensive review. Front. Plant Sci. 2023, 14, 1214006. [Google Scholar] [CrossRef]
  19. Huang, S.; Tang, L.; Hupy, J.P.; Wang, Y.; Shao, G. A commentary review on the use of normalized difference vegetation index (NDVI) in the era of popular remote sensing. J. For. Res. 2021, 32, 1–6. [Google Scholar] [CrossRef]
  20. Vannoppen, A.; Gobin, A. Estimating farm wheat yields from NDVI and meteorological data. Agronomy 2021, 11, 946. [Google Scholar] [CrossRef]
  21. Huang, J.; Wang, X.; Li, X.; Tian, H.; Pan, Z. Remotely sensed rice yield prediction using multi-temporal NDVI data derived from NOAA’s-AVHRR. PLoS ONE 2013, 8, e70816. [Google Scholar] [CrossRef] [PubMed]
  22. Maresma, A.; Chamberlain, L.; Tagarakis, A.; Kharel, T.; Godwin, G.; Czymmek, K.J.; Shields, E.; Ketterings, Q.M. Accuracy of NDVI-derived corn yield predictions is impacted by time of sensing. Comput. Electron. Agric. 2020, 169, 105236. [Google Scholar] [CrossRef]
  23. Zhen, Z.; Chen, S.; Yin, T.; Chavanon, E.; Lauret, N.; Guilleux, J.; Henke, M.; Qin, W.; Cao, L.; Li, J. Using the negative soil adjustment factor of soil adjusted vegetation index (SAVI) to resist saturation effects and estimate leaf area index (LAI) in dense vegetation areas. Sensors 2021, 21, 2115. [Google Scholar] [CrossRef]
  24. Al-Gaadi, K.A.; Hassaballa, A.A.; Tola, E.; Kayad, A.G.; Madugundu, R.; Alblewi, B.; Assiri, F. Prediction of potato crop yield using precision agriculture techniques. PLoS ONE 2016, 11, e0162219. [Google Scholar] [CrossRef]
  25. Newton, I.H.; Islam, A.T.; Islam, A.S.; Islam, G.T.; Tahsin, A.; Razzaque, S. Yield prediction model for potato using landsat time series images driven vegetation indices. Remote Sens. Earth Syst. Sci. 2018, 1, 29–38. [Google Scholar] [CrossRef]
  26. Ali, A.; Martelli, R.; Scudiero, E.; Lupia, F.; Falsone, G.; Rondelli, V.; Barbanti, L. Soil and climate factors drive spatio-temporal variability of arable crop yields under uniform management in Northern Italy. Arch. Agron. Soil Sci. 2023, 69, 75–89. [Google Scholar] [CrossRef]
  27. Řezník, T.; Pavelka, T.; Herman, L.; Lukas, V.; Širůček, P.; Leitgeb, Š.; Leitner, F. Prediction of yield productivity zones from Landsat 8 and Sentinel-2A/B and their evaluation using farm machinery measurements. Remote Sens. 2020, 12, 1917. [Google Scholar] [CrossRef]
  28. Zhu, Z. Change detection using landsat time series: A review of frequencies, preprocessing, algorithms, and applications. ISPRS J. Photogramm. Remote Sens. 2017, 130, 370–384. [Google Scholar] [CrossRef]
  29. Maxwell, S.K.; Sylvester, K.M. Identification of “ever-cropped” land (1984–2010) using Landsat annual maximum NDVI image composites: Southwestern Kansas case study. Remote Sens. Environ. 2012, 121, 186–195. [Google Scholar] [CrossRef] [PubMed]
  30. Lanucara, S.; Praticò, S.; Pioggia, G.; Di Fazio, S.; Modica, G. Web-based spatial decision support system for precision agriculture: A tool for delineating dynamic management unit zones (MUZs). Smart Agric. Technol. 2024, 8, 100444. [Google Scholar] [CrossRef]
  31. MacVicar, C. Concerning the meaning of potential in agriculture. South Afr. J. Agric. Ext. 1974, 3, 1–4. [Google Scholar]
  32. Lillesand, T.; Kiefer, R.W.; Chipman, J. Remote Sensing and Image Interpretation; John Wiley & Sons: Hoboken, NJ, USA, 2015. [Google Scholar]
  33. Ziadi, N.; Tran, T.S. Mehlich 3-extractable elements. In Soil Sampling and Methods of Analysis; CRC Press: Boca Raton, FL, USA, 2007; pp. 81–88. [Google Scholar]
  34. Kroetsch, D.; Wang, C. Particle size distribution. Soil Sampl. Methods Anal. 2008, 2, 713–725. [Google Scholar]
  35. Hendershot, W.H.; Duquette, M. Soft reaction and exchangeable acidity. Soil Sampl. Methods Anal. 2008, 2, 173–178. [Google Scholar]
  36. Mukiibi, A.; Machakaire, A.; Franke, A.; Steyn, J. A systematic review of vegetation indices for potato growth monitoring and tuber yield prediction from remote sensing. Potato Res. 2025, 68, 409–448. [Google Scholar] [CrossRef]
  37. Biswal, P.; Faisal, A.; Swain, D.K.; Bhowmick, G.D.; Mohan, G. NDVI is the best parameter for yield prediction at the peak vegetative stage of potato (Solanum tuberosum L.). Clim. Smart Agric. 2025, 2, 100053. [Google Scholar] [CrossRef]
  38. Mhango, J.K.; Grove, I.G.; Hartley, W.; Harris, E.W.; Monaghan, J.M. Applying colour-based feature extraction and transfer learning to develop a high throughput inference system for potato (Solanum tuberosum L.) stems with images from unmanned aerial vehicles after canopy consolidation. Precis. Agric. 2022, 23, 643–669. [Google Scholar] [CrossRef]
  39. Liepa, A.; Thiel, M.; Taubenböck, H.; Steffan-Dewenter, I.; Abu, I.-O.; Dhillon, M.S.; Otte, I.; Otim, M.H.; Lutaakome, M.; Meinhof, D. Harmonized NDVI time-series from Landsat and Sentinel-2 reveal phenological patterns of diverse, small-scale cropping systems in East Africa. Remote Sens. Appl. Soc. Environ. 2024, 35, 101230. [Google Scholar] [CrossRef]
  40. Yin, F.; Lewis, P.E.; Gómez-Dans, J.L. Bayesian atmospheric correction over land: Sentinel-2/MSI and Landsat 8/OLI. Geosci. Model Dev. 2022, 15, 7933–7976. [Google Scholar] [CrossRef]
  41. Alkhaled, A.; Townsend, P.A.; Wang, Y. Remote sensing for monitoring potato nitrogen status. Am. J. Potato Res. 2023, 100, 1–14. [Google Scholar] [CrossRef]
  42. Holben, B.N. Characteristics of maximum-value composite images from temporal AVHRR data. Int. J. Remote Sens. 1986, 7, 1417–1434. [Google Scholar] [CrossRef]
  43. MacQueen, J. Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics; University of California Press: Oakland, CA, USA, 1965. [Google Scholar]
  44. Lajili, A.; Cambouris, A.N.; Chokmani, K.; Duchemin, M.; Perron, I.; Zebarth, B.J.; Biswas, A.; Adamchuk, V.I. Analysis of four delineation methods to identify potential management zones in a commercial potato field in eastern Canada. Agronomy 2021, 11, 432. [Google Scholar] [CrossRef]
  45. Mohamad, I.B.; Usman, D. Research article standardization and its effects on k-means clustering algorithm. Res. J. Appl. Sci. Eng. Technol. 2013, 6, 3299–3303. [Google Scholar] [CrossRef]
  46. Cambardella, C.A.; Moorman, T.B.; Novak, J.; Parkin, T.; Karlen, D.; Turco, R.; Konopka, A. Field-scale variability of soil properties in central Iowa soils. Soil Sci. Soc. Am. J. 1994, 58, 1501–1511. [Google Scholar] [CrossRef]
  47. Staff, S.S. Soil taxonomy: A basic system of soil classification for making and interpreting soil surveys. Agric. Handb. 1999, 436, 336–337. [Google Scholar]
  48. Muleta, H.D.; Aga, M.C. Role of nitrogen on potato production: A review. J. Plant Sci. 2019, 7, 36–42. [Google Scholar]
  49. Rawal, A.; Lankau, R.A.; Ruark, M.D. How does soil organic matter affect potato productivity on sandy soil? Soil Sci. Soc. Am. J. 2024, 88, 1748–1766. [Google Scholar] [CrossRef]
  50. Bucher, M.; Kossmann, J. Molecular physiology of the mineral nutrition of the potato. In Potato Biology and Biotechnology; Elsevier: Amsterdam, The Netherlands, 2007; pp. 311–329. [Google Scholar]
  51. Esteves, C.; Ribeiro, H.; Braga, R.P.; Fangueiro, D. Remote sensing (NDVI) and Apparent soil electrical conductivity (ECap) to delineate different zones in a vineyard. Biol. Life Sci. Forum 2021, 3, 42. [Google Scholar] [CrossRef]
  52. Zebarth, B.; Rosen, C. Research perspective on nitrogen BMP development for potato. Am. J. Potato Res. 2007, 84, 3–18. [Google Scholar] [CrossRef]
  53. Cox, M.; Gerard, P.D.; Wardlaw, M.; Abshire, M. Variability of selected soil properties and their relationships with soybean yield. Soil Sci. Soc. Am. J. 2003, 67, 1296–1302. [Google Scholar] [CrossRef]
  54. Farooque, A.A.; Zaman, Q.U.; Schumann, A.W.; Madani, A.; Percival, D.C. Delineating management zones for site specific fertilization in wild blueberry fields. Appl. Eng. Agric. 2012, 28, 57–70. [Google Scholar] [CrossRef]
  55. Waterer, D. Impact of high soil pH on potato yields and grade losses to common scab. Can. J. Plant Sci. 2002, 82, 583–586. [Google Scholar] [CrossRef]
  56. Goverment of New Brunswick. Soil Management. n.d. Available online: https://www2.gnb.ca/content/gnb/en/departments/10/agriculture/content/crops/potatoes/soil_management.html (accessed on 20 August 2024).
  57. Sudduth, K.; Drummond, S.; Birrell, S.; Kitchen, N. Spatial Modeling of Crop Yield Using Soil and Topographic Data; Stafford, J.V., Ed.; Bios Scientific Publishers Ltd.: Oxford, UK, 1997; pp. 439–447. [Google Scholar]
  58. Arslan, S.; Colvin, T.S. Grain yield mapping: Yield sensing, yield reconstruction, and errors. Precis. Agric. 2002, 3, 135–154. [Google Scholar] [CrossRef]
  59. Kitchen, N.; Sudduth, K.; Myers, D.; Drummond, S.; Hong, S. Delineating productivity zones on claypan soil fields using apparent soil electrical conductivity. Comput. Electron. Agric. 2005, 46, 285–308. [Google Scholar] [CrossRef]
  60. Leroux, C.; Jones, H.; Taylor, J.; Clenet, A.; Tisseyre, B. A zone-based approach for processing and interpreting variability in multi-temporal yield data sets. Comput. Electron. Agric. 2018, 148, 299–308. [Google Scholar] [CrossRef]
  61. Taylor, J.; McBratney, A.; Whelan, B. Establishing management classes for broadacre agricultural production. Agron. J. 2007, 99, 1366–1376. [Google Scholar] [CrossRef]
  62. Kiani, M.; Hernandez-Ramirez, G.; Quideau, S.A. Spatial variation of soil quality indicators as a function of land use and topography. Can. J. Soil Sci. 2020, 100, 463–478. [Google Scholar] [CrossRef]
  63. Li, X.; McCarty, G.W. Application of topographic analyses for mapping spatial patterns of soil properties. Geospat. Anal. Earth Obs. (EO) Data 2019, 9, 1–32. [Google Scholar]
Remotesensing 17 03709 g001
Figure 1. Study area: McCain’s Farm of the Future Canada in Florenceville—Bristol, New Brunswick. Fields labeled X, Y, Z were used for tuber yield validation.
Figure 1. Study area: McCain’s Farm of the Future Canada in Florenceville—Bristol, New Brunswick. Fields labeled X, Y, Z were used for tuber yield validation.
Remotesensing 17 03709 g001
Remotesensing 17 03709 g002
Figure 2. Methodological workflow (TOA: Top Of Atmosphere; AP, agricultural potential; MVC, Maximum Value Composite; MZs, management zones. NDVI, Normalized Difference Vegetation Index).
Figure 2. Methodological workflow (TOA: Top Of Atmosphere; AP, agricultural potential; MVC, Maximum Value Composite; MZs, management zones. NDVI, Normalized Difference Vegetation Index).
Remotesensing 17 03709 g002
Remotesensing 17 03709 g003
Figure 3. Management zones of high, intermediate and low agricultural potential created using NDVI MVC metrics.
Figure 3. Management zones of high, intermediate and low agricultural potential created using NDVI MVC metrics.
Remotesensing 17 03709 g003
Remotesensing 17 03709 g004
Figure 4. Maps of tuber yield (a) and MZ based on the AP for fields X, Y and Z (b).
Figure 4. Maps of tuber yield (a) and MZ based on the AP for fields X, Y and Z (b).
Remotesensing 17 03709 g004
Table 1. Characteristics of Sentinel-2 and Landsat 8 Satellites.
Table 1. Characteristics of Sentinel-2 and Landsat 8 Satellites.
SatelliteResolutionDataset Availability
TemporalSpatialStartEnd
Landsat 8 116 days30 m18 March 2013Present
Sentinel-2 25 days10 m27 June 2015Present
1 USGS Landsat 8 Collection 2 Tier 1 Top of Atmosphere Reflectance. 2 Harmonized Sentinel-2 MultiSpectral Instrument, Level-1C.
Table 2. Statistical comparison test of the average values of the NDVI MVC metrics (mean, standard deviation, skewness) for AP MZs at the FOF CA.
Table 2. Statistical comparison test of the average values of the NDVI MVC metrics (mean, standard deviation, skewness) for AP MZs at the FOF CA.
MZNDVI MVC Metrics
MeanStandard DeviationSkewness
High AP0.88 a0.03 c−0.28 a
Intermediate AP0.87 b0.05 b−0.94 b
Low AP0.83 c0.09 a−1.24 c
Note: MVC, maximum value composite; MZ, management zones; AP, agricultural potential; means followed by the same letter are not significantly different at the 5% significance level according to the Tukey/Kruskal–Wallis test.
Table 3. Pearson correlation coefficients (r) between NDVI MVC metrics and soil physicochemical properties, soil physiographic characteristics (topography) and tuber yield.
Table 3. Pearson correlation coefficients (r) between NDVI MVC metrics and soil physicochemical properties, soil physiographic characteristics (topography) and tuber yield.
Physicochemical PropertiesNDVI MVC Metrics
MeanStandard DeviationSkewness
Soil particle size
Clay−0.04 ns0.07 ns−0.05 ns
Silt−0.10 ns0.03 ns0.09 ns
Sand0.21 ***-0.12 *−0.05 ns
Total N0.02 ns−0.14 ***0.19 ***
Total C−0.01 ns−0.11 **0.28 ***
pH−0.02 ns0.00 ns−0.09 *
Mehlich-3 extractable elements
P0.00 ns−0.05 ns−0.07 ns
K−0.22 ***0.07 ns0.09 *
Ca0.03 ns−0.08 ns0.04 ns
Mg−0.14 ***0.14 ***0.03 ns
Slope−0.01 ns−0.04 ns0.10 *
Tuber yield0.35 ***−0.41 ***0.25 *
Note: MVC, maximum value composite; Total N, total nitrogen; Total C, total carbon; statistically significant at *, p ≤ 0.05; **, p ≤ 0.01; and ***, p ≤ 0.001; ns, not significant.
Table 4. Geostatistical parameters of the tuber yield 2023 of the fields X, Y, and Z.
Table 4. Geostatistical parameters of the tuber yield 2023 of the fields X, Y, and Z.
FieldModelNugget Ratio (%)Range (m)RMSES
XExponential27360.81
YExponential38270.85
ZExponential23180.92
Note: Sill ratio (%) = [C0/(C0 + C)] × 100 where C0 = Nugget: (random semi-variance) and C = Partial sill (difference between the sill (C0 + C) and nugget (C0) semi-variances); this ratio measures spatial dependence or structure as described by; RMSES, root mean square error standardized to describe goodness of fit of a model (optimal value = 1).
Table 5. Mean and coefficient of variation of 2023 tuber yield (Mg ha−1) in the MZs.
Table 5. Mean and coefficient of variation of 2023 tuber yield (Mg ha−1) in the MZs.
MZTuber Yield (Mg ha−1)
Fields X, Y and ZField X
FieldMeanCV (%)MeanCV (%)
High AP60.06 a13.6861.16 a24.03
Intermediate AP53.48 b21.4656.38 a22.98
Low AP45.73 b10.8145.73 a10.81
Note: MZ, management zones; AP, agricultural potential; CV, Coefficient of Variation; means followed by the same letter are not significantly different at the 5% significance level according to the Tukey/Kruskal–Wallis test.
Table 6. Comparison of physicochemical properties in the MZs.
Table 6. Comparison of physicochemical properties in the MZs.
Physicochemical PropertiesMZ
High APIntermediate APLow AP
nMeanCV (%)nMeanCV (%)nMeanCV (%)
Soil particle size
Clay (%)6413.0 b20.417614.1 ab64.65514.4 a18.5
Silt (%)6442.8 a 14.517641.4 a15.55542.5 a15.3
Sand (%)6444.2 ab12.917645.2 a11.65543.1 b11.7
Total N (%)1310.22 a19.33570.20 b20.11070.19 b20.6
Total C (%)1312.45 a20.43572.17 b24.11072.12 b22.5
pH1315.77 b8.23575.9 a7.31075.9 a6.4
Mehlich-3 extractable elements
P (mg kg−1)131167 b45.9357183 a40.5107172 ab37.3
K (mg kg−1)131232 a27.2357216 b27.6107252 a34.6
Ca (mg kg−1)1311354 a47.53571295 a45.31071214 a 33.6
Mg (mg kg−1)131101 b38.4357119 a37.4107118 a36.7
Topography
Slope (%)1319.6 a96.835810.5 a127.11078.2 a83.6
Note: MZ, management zones; AP, agricultural potential; n, counts; CV, coefficient of variation; Total N, total nitrogen; Total C, total carbon; means followed by the same letter are not significantly different at the 5% significance level according to the Tukey/Kruskal–Wallis test.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Ramirez-Gonzalez, D.A.; Chokmani, K.; Cambouris, A.N.; D’Souza, M.L. Delineation of Management Zones Based on the Agricultural Potential Concept for Potato Production Using Optical Satellite Images. Remote Sens. 2025, 17, 3709. https://doi.org/10.3390/rs17223709

AMA Style

Ramirez-Gonzalez DA, Chokmani K, Cambouris AN, D’Souza ML. Delineation of Management Zones Based on the Agricultural Potential Concept for Potato Production Using Optical Satellite Images. Remote Sensing. 2025; 17(22):3709. https://doi.org/10.3390/rs17223709

Chicago/Turabian Style

Ramirez-Gonzalez, David A., Karem Chokmani, Athyna N. Cambouris, and Michelle L. D’Souza. 2025. "Delineation of Management Zones Based on the Agricultural Potential Concept for Potato Production Using Optical Satellite Images" Remote Sensing 17, no. 22: 3709. https://doi.org/10.3390/rs17223709

APA Style

Ramirez-Gonzalez, D. A., Chokmani, K., Cambouris, A. N., & D’Souza, M. L. (2025). Delineation of Management Zones Based on the Agricultural Potential Concept for Potato Production Using Optical Satellite Images. Remote Sensing, 17(22), 3709. https://doi.org/10.3390/rs17223709

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop