Delineation of Management Zones for Site-Speciﬁc Information about Soil Fertility Characteristics through Proximal Sensing of Potato Fields

: The delineation of management zones (MZs) has been suggested as a solution to mitigate adverse impacts of soil variability on potato tuber yield. This study quantiﬁed the spatial patterns of variability in soil and crop properties to delineate MZs for site-speciﬁc soil fertility characterization of potato ﬁelds through proximal sensing of ﬁelds. Grid sampling strategy was adopted to collect soil and crop data from two potato ﬁelds in Prince Edward Island (PEI). DUALEM-2 sensor, Time Domain Reﬂectometry (TDR-300), GreenSeeker were used to collect soil ground conductivity parameter horizontal coplanar geometry (HCP), soil moisture content ( θ ), and normalized di ﬀ erence vegetative index (NDVI), respectively. Soil organic matter (SOM), soil pH, phosphorous (P), potash (K), iron (Fe), lime index (LI), and cation exchange capacity (CEC) were determined from soil samples collected from each grid. Stepwise regression shortlisted the major properties of soil and crop that explained 71 to 86% of within-ﬁeld variability. The cluster analysis grouped the soil and crop data into three zones, termed as excellent, medium, and poor at a 40% similarity level. The coe ﬃ cient of variation and the interpolated maps characterized least to moderate variability of soil fertility parameters, except for HCP and K that were highly variable. The results of multiple means comparison indicated that the tuber yield and HCP were signiﬁcantly di ﬀ erent in all MZs. The signiﬁcant relationship between HCP and yield suggested that the ground conductivity data could be used to develop MZs for site-speciﬁc fertilization in potato ﬁelds similar to those used in this study.


Introduction
The potato (Solanum tuberosum L.) is the world's third most consumed food, which is produced in all continents except Antarctica [1]. Potatoes are produced above 19 million hectares worldwide, Canada being ahead of many countries in its potato production. Two Atlantic Canadian provinces, including Prince Edward Island and New Brunswick, contribute 24.5% and 13.6%, respectively, in the total potato production of Canada [2]. However, the yield per unit area of potatoes in these provinces is less than that of most American States, emphasizing the need for better management of potato cultivation in Canada.
For precise management of crop production within-field, it is very important to know about the spatial variability of crop and soil properties [3]. Several  A grid sampling approach was established by determining the coordinates of grids with Real-Time Kinematic Global Position System (RTK-GPS) (Benchmark, Calgary, AB, Canada). The horizontal coplanar geometry (HCP) of the field was measured using ground conductivity surveys with DUALEM-2 sensor (DUALEM Inc., Milton, ON, Canada) in both fields. The grid pattern for sampling was one-third, or half of the range of variability [32], based on which a grid size of 25 m × 25 m was fixed for Field 1 (n = 40) and Field 2 (n = 39); where n is the number of grids nodes).
Soil samples (about 500 g each) were collected from each grid from a 15 cm depth with a soil auger prior to sowing. The soils of PEI potato fields are sandy loams (Orthic Humo-Ferric Podzol) with large gravels and stones contents, and with more than 2% of SOM. The sandy soils have good drainage, aeration, and thus can result in drought stress in absence of supplemental irrigation. However, sandy loam soils with adequate presence of SOM can retain θ and do not drain as rapidly. Potatoes prefer a well-drained, medium textured and loose loam, fertile soil that is high in organic matter [33].
Post-sowing data collection was done twice during the growing season. For Field 1, the first and second samplings were on 8 June 2017 and 13 July 2017, respectively. For Field 2, the first and second samplings were on 11 June 2018 and 10 July 2018, respectively. At each sampling time, HCP was measured using DUALEM-2 sensor. A FieldScout TDR-300 probe (Spectrum Technologies, Inc., Aurora, CO, USA) was used to measure θ at each soil sampling point by inserting TDR rods to 15 cm depths. The data for normalized difference vegetative index (NDVI) and yield were collected at the end of growing seasons. GreenSeeker™ handheld sensor was used to measure NDVI. To collect the tuber yield samples a 1 m × 1 m plot was dug manually in each grid. A digital weighing balance was used to weigh those samples. After yield measurements, tubers were buried back in soil for farmers to collect their farm yield. Harvesting was performed on 2 October 2017 and 4 October 2018 in Fields 1 and 2, respectively.
The soil samples were analyzed for SOM, soil pH, θ, P, K, Fe, CEC, and LI. Phosphorus was measured using Olsen method [34], K using Flame photometric method [35], Fe with colorimetric method [36], CEC with BaCl2compulsive exchange procedure [37], and pH was determined using a pH multimeter. The LI was determined by mixing the SMP buffer solution with soil and followed by measuring its pH [38]. Soil organic matter was determined using standard methods [39,40]. Five A grid sampling approach was established by determining the coordinates of grids with Real-Time Kinematic Global Position System (RTK-GPS) (Benchmark, Calgary, AB, Canada). The horizontal coplanar geometry (HCP) of the field was measured using ground conductivity surveys with DUALEM-2 sensor (DUALEM Inc., Milton, ON, Canada) in both fields. The grid pattern for sampling was one-third, or half of the range of variability [32], based on which a grid size of 25 m × 25 m was fixed for Field 1 (n = 40) and Field 2 (n = 39); where n is the number of grids nodes).
Soil samples (about 500 g each) were collected from each grid from a 15 cm depth with a soil auger prior to sowing. The soils of PEI potato fields are sandy loams (Orthic Humo-Ferric Podzol) with large gravels and stones contents, and with more than 2% of SOM. The sandy soils have good drainage, aeration, and thus can result in drought stress in absence of supplemental irrigation. However, sandy loam soils with adequate presence of SOM can retain θ and do not drain as rapidly. Potatoes prefer a well-drained, medium textured and loose loam, fertile soil that is high in organic matter [33].
Post-sowing data collection was done twice during the growing season. For Field 1, the first and second samplings were on 8 June 2017 and 13 July 2017, respectively. For Field 2, the first and second samplings were on 11 June 2018 and 10 July 2018, respectively. At each sampling time, HCP was measured using DUALEM-2 sensor. A FieldScout TDR-300 probe (Spectrum Technologies, Inc., Aurora, CO, USA) was used to measure θ at each soil sampling point by inserting TDR rods to 15 cm depths. The data for normalized difference vegetative index (NDVI) and yield were collected at the end of growing seasons. GreenSeeker™ handheld sensor was used to measure NDVI. To collect the tuber yield samples a 1 m × 1 m plot was dug manually in each grid. A digital weighing balance was used to weigh those samples. After yield measurements, tubers were buried back in soil for farmers to collect their farm yield. Harvesting was performed on 2 October 2017 and 4 October 2018 in Fields 1 and 2, respectively.
The soil samples were analyzed for SOM, soil pH, θ, P, K, Fe, CEC, and LI. Phosphorus was measured using Olsen method [34], K using Flame photometric method [35], Fe with colorimetric method [36], CEC with BaCl 2 -compulsive exchange procedure [37], and pH was determined using a pH multimeter. The LI was determined by mixing the SMP buffer solution with soil and followed by measuring its pH [38]. Soil organic matter was determined using standard methods [39,40]. Five grams of ground air-dried soil <2 mm diameter were placed in 15 mL ceramic cups and oven-dried at 105 • C for 24 h and cooled in a desiccator before weighing for its mass (M 105 ) prior to combusting it for 5 h at 550 • C in a muffle furnace. The soil was then cooled down to 105 • C before putting it in a desiccator for further cooling down. The cooled samples were weighed again for their mass (M 550 ) to determine SOM, using a formula introduced by [41].

Statistical Analysis
Values of the mean, maximum, minimum, coefficient of variance (CV), and variance were calculated for all the crop and soil properties using Minitab 18 (Minitab, INC., State College, PA, USA). Minitab provides numerous statistical analysis tools helpful for conducting regression, analysis of variance (ANOVA), and other statistical measures [42]. Stepwise regression, which is a method of fitting regression models while making choice of predictive variables, was conducted in Minitab. In each step, a variable (from the list of variables given in Tables 1 and 2) were considered for addition to or subtraction from the set of explanatory variables mentioned in Equations (1)-(4), based on values of coefficient of determination (R 2 ) of the regression models. Stepwise regression helped in assessing the overall contribution of all soil properties toward soil variability in a Fields 1 and 2. Stepwise regression was performed, which reduced the number of parameters to have ones with more effect on the variability. Coefficient of variation (CV) is the first indicator of variability. If the CV < 15%, it means the parameter is the least variable; if CV = 15-35%, it means the parameter is moderately variable. If the CV > 35%, it means the parameter is highly variable [43]. Geostatistical analysis was performed using GeoStatistics Software (GS+) to describe the spatial dependencies of soil and crop parameters. GS+ is a geostatistical analysis program that allows readily measuring and illustrating spatial relationships in geo-referenced data. GS+ analyzes spatial data for autocorrelation and uses the generated information to make optimal (as well as statistically precise) maps of the area of interest [44].
ArcMap 10.7 was used to analyze the spatial variability of soil properties and crop yield. All of the parameters were interpolated using the Inverse Distance Weighted (IDW) interpolation technique. The IDW was preferred over other interpolation methods including kriging based on its simple approach [45] and several error statistics used to assess the performance of interpolation methods to represent the analysis of spatially continuous phenomena [46,47]. Since, lower values of the error statistics indicate higher accuracy of spatial interpolation, the lower error statistics of IDW were considered as basis for evaluating prediction accuracy of the interpolated data by this method [48].
All of the maps were produced on the same scale and the equal number of classes to allow easier comparison. Cluster analysis was performed on parameters reduced by stepwise regression analysis to observe the spatial patterns of natural productivity groups, and cluster dendrograms were developed with the help of which the data was classified in the form of groups. A dendrogram represents different clusters and the distinctness of the cluster from its closest neighbor. The multiple means comparison was performed using Minitab 18 software and the means of yield were compared to check the low yielding and high yielding areas of the fields. The clustered data based on the class membership was imported into ArcMap to develop the MZs.

Variability in Crop and Soil Parameters
The soil and crop properties were least to moderately variable, except HCP and K that were highly variable (CV > 35%) during the second sampling of Field 1 (Table 1). Whereas, LI was least variable during first and second samplings in Field 1 (CV = 1.06 and 1.68%) and Field 2 (CV = 1.95 and 1.55%). The SOM showed the least variability and θ was moderately variable in both fields during both samplings ( Table 1). The SOM reported in Table 1 does not reflect that the soils of the experimental fields were ideal for potato production. Soils with higher levels of organic matter (over 3.5%) have an ability to mineralize nitrogen for use by a growing potato crop [49]. Loss of SOM leads to poor water infiltration through soil and slow drainage, increase in saturated soil conditions, and electrical conductivity and resultantly soil's ability to buffer electrical conductivity [50]. Farooque et al. [31] evaluated spatial variability of HCP and θ measured in two potato fields of New Brunswick and two in PEI, different from the fields used in this study, and reported average values of HCP and θ in the range of this study.
Potassium concentration in soil has been negatively related to potato tuber yield or count by Redulla et al. [51] who reported that K in soils of their experimental fields might have been present in excess, since when there was a significant correlation between tuber yield or tuber count and K, it was negative. Their four potato experimental fields had K ranging from 168 to 233 mg kg −1 and in the range of K concentration of four dates of measurements in the two studied fields (i.e., 173 to 222 mg kg −1 in Field 1, 169 to 201 mg kg −1 in Field 2). Tuber yield in this study was also negatively related with K concentration; e.g., Field 1 with comparatively higher concentration of K than Field 2 had lower tuber yield than Field 2 ( Table 1).
Soil pH and CEC showed least variability (CV < 15%) in both fields during both samplings and it had almost the same values as discussed reported by Khan et al. [28]; their values were non-significantly variable across the potato fields of New Brunswick and PEI. The pH ranged from 4.50 to 6.20 in accordance with the findings of [36], who reported soil pH of potato fields of PEI in this range. These pH ranges are lower than the optimal values of pH for potato production, i.e., 7 to 9 [52]. Soil liming is one of the ways to bring alkaline soil to the optimal level of soil pH for potato production. Statistics Canada reports state that 1.1% of cropland across Canada were treated with lime during 2010, whereas PEI's 20.2% croplands went under these treatments [53]. The uniform application of lime is reflected by the least variable readings of LI; i.e., CV = 1.06% during first sampling, CV = 1.68% during second sampling in Field 1, and CV = 1.95% during first sampling and CV = 1.55% during second sampling in Field 2. The P showed the least to moderate variability during both years.

Spatial Dependencies of Soil Parameters
Geostatistical analysis was performed in which spatial dependencies of soil parameters were evaluated to check the spatial variability within fields ( Table 2). For Field 1 data, a Gaussian was the best-fitted model for all of the parameters, except CEC, which had the best fit with a linear model. All of the parameters had strong spatial dependencies except CEC. The range of all the parameters varied from 27 to 34 m except for CEC, which was 66 m. All of the parameters had strong R 2 ranging from 0.85 to 0.98, except CEC, which had the weakest R 2 ; i.e., 0.28. For Field 2 data, linear model had the best fit to pH, P, K, lime index, CEC, and yield variables. Moisture content had the best fit with the exponential model. The spherical model was fitted for NDVI and Gaussian showed the best fit to the data of SOM and Fe. The range and, thus, spatial variability for HCP data, was least; i.e., 19 m. Yield, pH, P, K, LI, and CEC had strong spatial dependencies, having a range of 65.8 m.

Regression Analysis
It was suggested by Tang et al. [30] that HCP has a positive correlation with potato tuber yield, which showed that HCP played a very important role in predicting potato yield. Regression analysis at α = 0.05 resulted in the parameters describing 90% of the variability in the Field 1 (Equation (1)) and 89% variability in Field 2 (Equation (2)).
It is difficult for growers to manage too many parameters as mentioned in Equations (1) and (2) to increase crop yield. The large list of variables can be shortened without compromising quality of prediction through stepwise regression to eliminate the least important parameters, while keeping the R 2 value close to high. Managing a lesser number of variables becomes easier for growers. Equations (3) and (4) present the most important soil and crop properties including HCP, θ, and NDVI that explain 86 and 71% of the within-field variability for Fields 1 and 2, respectively.
For farmers' ease, the component of HCP may be removed from Equations (3) and (4) to get simplified forms of the equations as Yield = −7.99 + 0.26 × θ + 22 × NDVI (R 2 = 83%) and Yield = 2.36 + 0.57 × θ + 9.05 × NDVI (R 2 = 52%) that use the easily measured variables θ and NDVI. Removing θ and NDVI from these equations and using HCP may only reduce strength (R 2 ) of these equations. Farooque et al. [29] reported strong relationships of ground electrical conductivity (ECa)-HCP is a component of ECa-with tuber yield with R 2 ranging from 55 to 62% using linear cubic models, and from 57 to 0.66% using cubic models for potato fields of Atlantic provinces in Canada (PEI and New Brunswick).

Interpolation for Within-Field Variability
The GIS interpolated maps of soil and crop properties, produced in ArcGIS using the IDW showed significant spatial variability for the selected properties across Fields 1 and 2 ( Figure 2). A higher value of HCP was observed in the southeast part of Field 1, the lower value was observed in the northwest part of this field (Figure 2). A higher value of yield was observed in the south and west parts of the field than those observed in the north and east parts of the field. It was because of high SOM in the southeastern and western parts of the field. Soil organic matter that affects crop productivity has been related with spatial yield variations for effective nutrient management [54,55].
The GIS interpolated maps of soil and crop properties, produced in ArcGIS using the IDW showed significant spatial variability for the selected properties across Fields 1 and 2 ( Figure 2). A higher value of HCP was observed in the southeast part of Field 1, the lower value was observed in the northwest part of this field (Figure 2). A higher value of yield was observed in the south and west parts of the field than those observed in the north and east parts of the field. It was because of high SOM in the southeastern and western parts of the field. Soil organic matter that affects crop productivity has been related with spatial yield variations for effective nutrient management [54,55].
High values of θ were observed in the southwest of the field; whereas, low value was observed in the northeast part. Southwestern parts of the field had the higher NDVI than the northern parts. Yield maps showed substantial spatial variation across the field. High yielding areas were in the southwest of the field, and low yielding areas were in the north of the field in most of the parameters for Field 1. Maps presented in Figure 2 also showed that the tuber yield was higher in the areas with more HCP, θ, and NDVI. The overall results of descriptive statistics and the interpolated maps suggested moderate variability in soil properties and tuber yield within-field, except for HCP, θ, and NDVI, which was highly variable for Field 1. High values of θ were observed in the southwest of the field; whereas, low value was observed in the northeast part. Southwestern parts of the field had the higher NDVI than the northern parts. Yield maps showed substantial spatial variation across the field. High yielding areas were in the southwest of the field, and low yielding areas were in the north of the field in most of the parameters for Field 1. Maps presented in Figure 2 also showed that the tuber yield was higher in the areas with more Agronomy 2020, 10, 1854 9 of 17 HCP, θ, and NDVI. The overall results of descriptive statistics and the interpolated maps suggested moderate variability in soil properties and tuber yield within-field, except for HCP, θ, and NDVI, which was highly variable for Field 1.
The maps of Field 2 have been shown in Figure 3. Higher values of HCP, θ, and yield were observed in western parts of the field and lower value of HCP and θ was observed in the east of the field, whereas the lower value of yield was observed in the southeast part of the field. The presence of soil moisture content enhances the crop growth by improving nutrient use efficiency of soils/crops [54,55]. The maps of Field 2 have been shown in Figure 3. Higher values of HCP, θ, and yield were observed in western parts of the field and lower value of HCP and θ was observed in the east of the field, whereas the lower value of yield was observed in the southeast part of the field. The presence of soil moisture content enhances the crop growth by improving nutrient use efficiency of soils/crops [54,55].
Higher value of NDVI was observed in the northwest parts of the field where lower value of NDVI was observed in the southern part of the field. Maps of Field 2 also showed that tuber yield was higher in the areas with high θ, NDVI, and HCP values. Higher value of NDVI was observed in the northwest parts of the field where lower value of NDVI was observed in the southern part of the field. Maps of Field 2 also showed that tuber yield was higher in the areas with high θ, NDVI, and HCP values.
The MZs maps or the relationships of yield developed with θ, NDVI, and HCP can help growers to manage their practices according to variations in field HCP, θ, and NDVI. The HCP maps of a field may be obtained either commercially or upon request from governments' agricultural services related ministries. For example, the department of Agriculture and Agri-Food Canada (https: //www.agr.gc.ca/eng/?id=1343066456961) provides interactive agriculture-related maps, geospatial data, and tools to help you make better decisions for environmentally responsible yet competitive agriculture. There exist simple methods to measure θ and NDVI using a variety of sensors available in the market. Gozdowski et al. [21] also showed significant variations in NDVI within the MZs. This vegetation index reflects actual crop status, which relates to the soil properties that are important to satisfy crop requirements. Consequently, the NDVI may be a useful tool for MZ delineation, especially because their measurements are simple to make. The moisture content of soil can also be determined by oven dry method. Moreover, the simplest method to determine θ in the soil is "the feel and appearance method", which is one of several irrigation scheduling methods used worldwide. It is a way of monitoring soil moisture to determine when to irrigate and how much water to apply.

Cluster Analysis and Management Zones
The results of interpolated maps confirmed that there exists spatial variations in the fields, which require the development of MZs for site-specific nutrient management. Cluster analysis was performed and soil and crop properties were clustered in the form of groups that differentiate the high yielding and low yielding areas of the fields, by separating the high yielding sampling points and low yielding sampling points. The dendrograms of the Fields 1 and 2 are shown in Figures 4 and 5, representing clustered soil and tuber yield data into three groups based on their similarity level. The productivity levels to develop MZs were decided based on tuber yield data, i.e., excellent (yield > 65,000 kg ha −1 ), medium (yield 55,000-65,000 kg ha −1 ), poor (yield < 55,000 kg ha −1 ) for both fields. The grouping of soil and crop properties in dendrogram showed that most of the sampling points fall in medium productivity area (Figures 4 and 5) and a of couple data points fell in excellent productivity area and a very few points lied in the poor productivity area of the field, which indicated that both were high yielding fields. Moreover, the cluster analysis differentiated the high yielding and low yielding areas of the fields.     After cluster analysis, the comparison of mean tuber yield and soil properties for MZs in Field 1 are shown in Table 3. In this, field HCP and yield were significantly different in all zones having the highest values in excellent zones. Highest values were observed for θ and SOM in the medium zone. Lime index and pH had maximum values in the poor zone. Phosphate, K, Fe, and NDVI were not significantly variable in all the zones.
Similarly, in Field 2 (Table 4), HCP, θ, pH, K, and yield were significantly different in all of the zones having the highest values in the excellent zones. Phosphate, Fe, and CEC had higher values in excellent zones; however, their values differ moderately across the field. Soil organic matter and NDVI were the least significant in the field. Since higher values of HCP, θ, pH, K, P, Fe, and CEC After cluster analysis, the comparison of mean tuber yield and soil properties for MZs in Field 1 are shown in Table 3. In this, field HCP and yield were significantly different in all zones having the highest values in excellent zones. Highest values were observed for θ and SOM in the medium zone. Lime index and pH had maximum values in the poor zone. Phosphate, K, Fe, and NDVI were not significantly variable in all the zones. Similarly, in Field 2 (Table 4), HCP, θ, pH, K, and yield were significantly different in all of the zones having the highest values in the excellent zones. Phosphate, Fe, and CEC had higher values in excellent zones; however, their values differ moderately across the field. Soil organic matter and NDVI were the least significant in the field. Since higher values of HCP, θ, pH, K, P, Fe, and CEC determine good soil health for potato production [30,56] their presence in the excellent zone ensures higher tuber yield in excellent zones as compared to the poor and medium zones [29,31]. The tuber yield obtained from Fields 1 and 2 was higher than the country average vales of tuber yield of nations from Europe and America that adopt the best nutrient management practices, as tabulated by Koch et al. [57]. The values of SOM and pH reported in Tables 1, 3 and 4 of this study are in the neighborhood of the values reported by Farooque et al. [31], measured in two potato fields of New Brunswick and two in PEI. Moreover, the values of P and K reported by Zare et al. [56] that resulted in enhanced tuber yield in potato fields of PEI and New Brunswick are in the range of the values of P and K of excellent and medium zones, or high yield producing portions of Fields 1 and 2 of the current study. Stark et al. [58] commended that potatoes have high nutrient demands, as they require approximately 247 kg of N, 33.5 kg of P, and 335 kg of K per hectare to produce approximately 56,491 kg/ha yield of Russet Burbank in southern Idaho, USA. The clustered data were imported into ArcGIS where MZs were developed indicating the different levels of productivity that required the variable application of nutrients across the fields. The comparison of HCP and yield mapped in MZs for the Fields 1 and 2 are shown in Figures 6 and 7, respectively. The information regarding tuber yield related with MZs is important as a part of a successful agronomic strategy to adequately manage nutrients for potato crop, which is essential throughout the whole growth period [57]. Development of MZs with the help of soil characteristics, especially from the proximal sensing information of soil's HCP can help farmers manage crop inputs other than nutrients during the whole growth period. The following information generated from MZs can help producers reduce the gap between the actual yield and the yield potential [59]. The results confirm the findings of [26,27] who delineated MZs using EM38 measured soil electrical conductivity, and related it with soil properties used to related MZs and tuber yield. The approach is recommended to have a control on managing within-field variability of soil and crop characteristics for enhancing potato productivity and profitability, while reducing the environmental impacts of agricultural practices.

Conclusions
This research revealed that there exists spatial variability within potato fields that require precise agricultural inputs. Geostatistical analysis showed that most of the parameters had strong spatial dependencies, and stepwise regression explained 89 to 90% of within-field variability, while considering all of the studied variables (i.e., HCP, θ, SOM, pH, P, K, Fe, LI, P/Al, CEC, and NDVI), 71 to 86% while considering the most significant variables (i.e., HCP, θ, and NDVI), and 52 to 83% while considering the easily measureable variables (θ and NDVI). The results of cluster analysis and comparison of means of tuber yield with other soil variables suggested that ground conductivity; i.e., HCP, θ, NDVI, and tuber yield data can be utilized to delineate MZs for site-specific nutrient management in potato fields. Field implications of this research include the fact that MZs can provide a better way to manage the spatial variability of soil properties and tuber yield within fields. The use of MZs would decrease the cost of production while increasing farm profitability. However, the farmers may face challenges in determining the recommended soil properties to construct MZ maps. There may exist commercial services for assisting farmers in obtaining maps of variable soil properties of their fields, or such maps may be provided by government departments responsible for guiding farmers to make informed decisions about agricultural management practices. Among various examples related to other countries for this purpose is the example of the department of Agriculture and Agri-Food Canada (https://www.agr.gc.ca/eng/?id=1343066456961) that provides interactive agriculture-related maps, geospatial data, and tools to help you make better decisions for environmentally responsible yet competitive agriculture. For properties, such as θ and NDVI, there exist simple methods and tools of measurements, such as commercially available sensors. The moisture content of soil can also be determined by the oven dry method, or, simply, by the feel and appearance method.