IoT-Enabled Instrumentation for Real-Time In-Situ Soil Electrical Conductivity Monitoring for Precision Agriculture

Precision agriculture aims to manage a farm by utilizing advanced technology to monitor and assess land and crops and adjust the treatment according to the land and crop conditions. Precision agriculture uses spatial and temporal information to characterize variability within agricultural fields and to support management decisions based on observed field conditions. Technologies such as sensing, remote monitoring, geospatial analysis, and variable-rate application can provide information needed to better understand and manage this variability. Moreover, the projected increase in global food demand associated with population growth toward 9–10 billion by 2050 has encouraged the adoption of technologies such as the Internet of Things (IoT), remote sensing, and variable-rate application in agriculture [1, 2]. These technologies are increasingly used to improve the management of agricultural resources while accounting for variability within cultivated areas. Spatial variability is an important consideration in precision agriculture because differences in soil and crop conditions can be used to identify areas requiring different management practices [3-5].

Corn (Zea mays L.) is not only consumed widely by humans and animals but also exploited in various industrial fields. Therefore, there is a high demand for its yields and productivity [6, 7]. The more yields and productivity are achieved, the more nutrients are needed in the soil to outgrow the plants. As shown in several studies, the use of NPK fertilizer has successfully increased the number of the yields [8-10]. Soil electrical conductivity (EC) can be used as an indirect proxy for the concentration of dissolved ions in the soil, which is influenced by multiple factors such as moisture, temperature, and salinity. Soil EC is a useful proxy for characterizing soil physicochemical conditions and can provide information related to salinity, soil moisture, and dissolved-ion dynamics [11, 12]. However, its interpretation as an indicator of nutrient availability or soil fertility requires consideration of other soil properties and site-specific conditions.

Many farmers merely use fertilizer without adjusting it to the soil conditions in some areas. Occasionally, they only observe the plant condition manually in determining to use fertilizer. This indeed impacts the harvest [13]. The assessment of the nutrient content of soil can only be done scientifically in a laboratory by analyzing soil samples. Undoubtedly, it requires funds and allocated time. Furthermore, the samples do not immediately cover a larger area; as a result, it is nearly impossible to obtain high spatial accuracy data [14-16]. In other words, laboratory analysis cannot be the sole indicator for determining the type and amount of fertilizer for the soil [17, 18].

The real-time monitoring of soil EC serves as a vital approach for understanding the dynamic changes in the soil environment while supporting more sustainable agricultural management. Soil EC is influenced by dissolved salt content, moisture levels, soil texture, and other physicochemical characteristics; consequently, it acts as an indicator for identifying spatial and temporal variations in soil conditions and potential salinity risks [19, 20]. Continuous EC monitoring enables the rapid detection of soil condition changes related to nutrient movement, salt accumulation, and irrigation processes, allowing water and fertilizer application to be adjusted to actual field conditions. Real-time monitoring of soil EC provides information on spatial and temporal changes in soil conditions. Soil EC is influenced by dissolved ions, soil water content, temperature, texture, and other physicochemical properties [19, 20]. Consequently, repeated EC measurements can be used to characterize changes in field conditions over time and to identify spatial patterns that may be relevant to subsequent soil and crop assessment. When integrated with georeferenced measurements, EC monitoring can provide a spatial basis for evaluating field variability and supporting further soil management analysis. Thus, integrating real-time EC monitoring with precision agriculture technologies can provide additional spatial and temporal information for characterizing field variability and supporting subsequent soil-management assessment [19, 20]. This study additionally supports SDG 2 (Zero Hunger) through the provision of field-scale insights that can refine soil and fertilizer management, and SDG 13 (Climate Action) by allowing for ongoing georeferenced assessments of soil EC, which can assist in recognizing variations in soil conditions related to fertilizer application and rainfall variability.

This has prompted the adoption of sensing technologies to obtain timely information on soil conditions and nutrient-related indicators. The technology facilitates the development of an instrumentation system for measuring soil EC in situ and obtaining spatially referenced EC data under field conditions. The system is also able to manage the spatial variability by integrating geographic coordinates from a Global Positioning System (GPS) module. The resulting georeferenced measurements can subsequently be processed in a Geographic Information System (GIS) to visualize the spatial distribution of measured soil EC. The integration of soil sensing, GPS-based georeferencing, and IoT data acquisition enables EC measurements to be associated with specific sampling locations and displayed as spatial information. Such measurements can be used to characterize field variability and temporal changes in soil EC during crop cultivation. Based on this approach, the present study aimed to develop an IoT-enabled instrumentation system for in-situ soil EC measurement and to evaluate its performance for monitoring the spatial and temporal variability of EC in a maize field. Therefore, this study aimed to develop an IoT-driven instrumentation system and then do an assessment to its use by testing the soil EC in a real-time setting.

3.1 Sensor calibration

The calibration analysis was performed using 10 paired EC measurements obtained from the developed sensor and the YY-1080 reference instrument. The measurements covered the observed EC range of the calibration dataset. The results showed a strong linear relationship between the reference EC measurements and the readings obtained from the developed sensor, with a regression equation of y = 0.5626x and R² of 0.9966 (Figure 5). In this relationship, x represents the reference EC value, and y represents the raw EC reading obtained from the developed sensor. Because the regression equation expresses the sensor reading as a function of the reference measurement, the equation was rearranged to obtain the reference-equivalent EC value from the raw sensor reading. Accordingly, the corrected EC value was calculated by dividing the raw sensor reading by 0.5626. The corrected EC values were then used in the subsequent spatial and statistical analyses. The RMSE and MAE obtained from the calibration dataset were 2.102 mS/cm and 1.963 mS/cm, respectively. These metrics were calculated using the same 10 paired measurements used to establish the linear regression equation and therefore represent the calibration-fit errors rather than independent validation errors. The results indicate the magnitude of the differences between the sensor-derived EC values and the reference measurements within the calibration dataset. However, because an independent dataset was not used for validation, these results do not establish the accuracy or generalizability of the developed sensor under other soil and environmental conditions. Further validation using independent measurements and laboratory-based EC analysis is required.

The calibration assessment was based on 10 paired field measurements and comparison with a reference field instrument. Therefore, the results should be interpreted within the EC conditions represented by the calibration dataset. The calibration procedure did not constitute independent laboratory validation of the sensor. Further validation using laboratory-based EC measurements and a broader range of soil and environmental conditions would be required to establish the general measurement accuracy of the sensor. There was a strong correlation between EC and capacitance-based sensors and the concentrations of dissolved nutrient ions, with R² values between 0.97 and 0.99 [22]. The Modbus RS485 sensor demonstrated stable performance throughout the study, requiring less frequent recalibration despite potential electrical noise and interference. These findings are consistent with those reported in a previous study [23], which also demonstrated the effectiveness of the sensor for EC measurement. Therefore, it was confirmed that the EC sensor was reliable to be used as an integrative parameter in the real-time monitoring of soil EC.

005.png

Figure 5. Calibration relationship between the developed sensor and the reference electrical conductivity (EC) instrument

3.2 Real-time monitoring of electrical conductivity levels

A real-time monitoring of soil EC was able to retrieve thorough spatial data of the soil. The georeferenced EC measurements provided spatial data for characterizing field variability. The data were processed using Kriging interpolation to visualize the distribution of measured EC across the study plots and cultivation phases. These spatial patterns were used to describe changes in EC before and after fertilization and to examine the variability observed among sampling locations. The resulting maps provide spatial information that can support subsequent assessment of soil conditions and field management (Figure 6). The data collection became the basis for the spatial analysis of the EC distribution. The Kriging interpolation technique helped to generate the map estimation of the overall EC value throughout the study sites. The measured EC value was visually indicated by the color gradient representing low (light yellow), medium (orange), and high (dark brown) EC classes. This observation covered four research plots (plot 1, plot 2, plot 3, and plot 4) throughout 3 phases of corn planting: prior to planting, after the first fertilization, and after the second fertilization (Figures 7–9).

006.png

Figure 6. Field installation of the Internet of Things (IoT)-based electrical conductivity (EC) measurement system

Spatial interpolation of soil EC was performed using ordinary Kriging in ArcGIS ArcMap 10.8. The EC measurements were georeferenced using the WGS 1984 UTM Zone 47S coordinate reference system. For each plot and observation phase, 20 georeferenced sampling points were used to generate the corresponding interpolated EC map. The same sampling locations were revisited during each observation phase. The experimental variogram was automatically fitted using a spherical model. The final nugget, sill, and range parameters were obtained from the fitted variogram model rather than specified manually. A variable search radius was used with a maximum of 12 neighboring points and one search sector. The search neighborhood was circular. The selected variogram model and its fitted parameters are reported in Table 3. The performance of the Kriging interpolation was assessed using cross-validation. Prediction performance was evaluated using mean error (ME), RMSE, mean standardized error (MSE), and root mean square standardized error (RMSSE).

Cross-validation was performed using 80 observations. The ME was 0.00037 mS/cm, while the RMSE was 0.0296 mS/cm. The MSE was 0.0136, and RMSSE was 0.978. The RMSSE value close to 1 indicates that the estimated prediction uncertainty was broadly consistent with the observed prediction errors. The regression slope of 0.1003 indicates a smoothing effect; therefore, the interpolated EC maps should be interpreted as representations of spatial patterns rather than exact point values. The interpolated EC maps should be interpreted as representations of spatial patterns rather than exact point values.

71.png

72.png

81.png

82.png

91.png

92.png

Table 3. Variogram parameters and cross-validation results of ordinary Kriging for soil electrical conductivity (EC) interpolation

In the pre-fertilization phase, the EC distribution was heterogeneous. Unlike plot 2, which was dominated by low EC levels, plots 1, 3, and 4 showed spatial variability across low, medium, or high EC levels. The observed pattern indicates that the measured EC distribution was heterogeneous across the plots. However, after the first treatment in the fertilizing process, the number of EC values rose significantly across all plots. Notable shifts were found mostly in plots 1, 2, and 4, where the conductivity areas were in more medium (orange) and high (dark brown) levels, particularly in plot 2, which was at a low value once and then promptly shifted to the medium EC level. The measured EC values were higher in several plots after the first fertilization phase than in the pre-fertilization maps.

A further change in measured EC was observed following the second fertilization treatment, with higher EC values recorded across the study plots. All four plots displayed a dominant dark brown coverage, which indicated that they were in the highest range of EC level. This coincided with broader areas of higher measured EC across the study-site maps. Because each fertilizer dosage was represented by a single plot and other soil and environmental factors were not independently controlled, the observed spatial changes cannot be attributed solely to fertilizer application. The soil EC patterns across the four plots showed noticeable temporal and spatial variation across the fertilization phases and were also influenced by intrinsic spatial heterogeneity. The pre-fertilization maps showed predominantly low measured EC values in several areas, represented by the light-yellow gradient. These patterns indicate spatial variation in EC rather than directly indicating low nutrient concentrations. This phenomenon is in accordance with what the literature suggests: soil texture, moisture, and residual nutrient salts are the main factors influencing EC values [24]. Due to differences in soil chemical composition or moisture retention, plot 1 and plot 3, before fertilization, exhibited a moderately higher level of EC. This confirmed that the site was not uniformly homogeneous, aligning with typical spatial heterogeneity found in farming soil.

Following the first fertilization, a broader distribution of moderate EC values was observed in the Kriging maps. This change may reflect changes in dissolved-ion conditions associated with fertilization, although the specific chemical changes were not directly measured in this study. The observed spatial pattern after fertilization was consistent with the formation of spatially distinct EC zones, with the observed distribution potentially reflecting differences in soil moisture, ionic composition, texture, and field management [25]. Unlike plots 1 to 3, plot 4 was cultivated traditionally, which displayed a more variable level of EC compared to the other plots, which may reflect inconsistencies in traditional farmer fertilization methods [26].

The IoT-based EC sensor captured changes in the spatial distribution of measured EC across the study sites, with greater variation observed after the fertilization phases. The most substantial increase in all plots was shown after the second fertilization phase, as the gradient of darker brown can be seen across the map. This observation was consistent with the temporal change in measured EC following repeated fertilization [27]. The EC levels were more consistent and moderate in plots 1 and 2, indicating that fertilization was performed regularly and uniformly. Meanwhile, EC levels in plot 4 were much higher, as it was expected, since the spatial pattern of EC in plot 4 was more heterogeneous, which may reflect differences in field management and underlying soil conditions [28]. High EC values may reflect differences in soil properties such as moisture conditions, texture, organic matter, and microtopography; however, these factors were not independently evaluated in the present study [29].

The 3-phase observation revealed changes in the spatial distribution of measured EC from lower to higher EC classes across the observation periods. These changes may reflect temporal variation in dissolved-ion conditions, soil moisture, and other site factors; however, the specific chemical changes were not directly measured in this study. These changes were effectively captured and shown by the Kriging interpolation method, which is consistent with former studies that have reported the suitability of Kriging for characterizing the spatial variability of soil-related properties [30]. The three-phase observations also showed that the spatial distribution of measured EC was not uniform across the study plots, reflecting spatial heterogeneity in the observed field conditions. These findings support the use of precision agriculture technologies for obtaining spatial information that may assist further field and soil-management assessment [31]. Overall, integrating in-situ EC sensing with geostatistical analysis provides a practical approach for characterizing the spatial variability of measured soil EC. The resulting information can be used as supporting data for further evaluation of field conditions within a precision agriculture framework.

3.3 Changes in measured soil electrical conductivity across fertilization phases

Corn growth was classified by identifying the parameters after data collection. Each planting phase had a single classification system, which was performed at 30 and 45 DAP (Tables 4 and 5). Subsequently, the data from each phase were presented in the form of a bar graph, illustrating the crop growth during the observation. EC reflects the capacity of the soil environment to conduct electrical current and is affected by dissolved ions, soil moisture, texture, temperature, and other physicochemical properties. Therefore, changes in EC were analyzed as changes in measured soil conditions rather than as direct measurements of nutrient availability. The analysis was conducted to examine changes in measured EC across the different fertilization phases (Figures 7–9). Table 6 presents the ANOVA results for measured soil EC under variations in crop age, fertilizer dosage, and sampling point (Figure 10). The table also shows that plant age is the most influential factor, as it has the highest F-value (156.219; p < 0.001). The interaction of plant age and fertilizing treatment comes second (F = 5.144; p = 0.002), followed by the fertilizer treatment itself (F = 4.087; p = 0.008). On the other hand, there is no significant difference in the EC value for the sampling point factor (p = 0.682). In two growth periods, data were collected: 30 DAP and 45 DAP (Table 7). The variables of corn growth were categorized into 2 indicators: plant growth at 30 DAP (PG30) and plant growth at 45 DAP (PG45) (Figure 11).

Table 4. Classification of plant growth parameters at 30 days after planting (DAP)

In the exploratory ANOVA, there was a statistically significant crop-age term (F = 156.219, p < 0.001), fertilizer-dosage term (F = 4.087, p = 0.008), and an interaction term of crop age and fertilizer dosage (F = 5.144, p = 0.002), while the sampling-point term was not statistically significant (F = 0.819, p = 0.682). The statistically significant fertilizer-dosage variable captures differences in recorded soil EC for the observed fertilizer-dosage plots in the exploratory model. However, this finding cannot be taken as evidence for a replicated effect of the fertilizer treatment, as there was only one plot per fertilizer application. The interaction of crop age and fertilizer dosage should also be considered a pattern associated with exploratory plots rather than a direct causal effect of fertilizer.

Table 5. Classification of plant growth parameters at 45 days after planting (DAP)

Table 6. Summary of exploratory analysis of variance (ANOVA) for measured soil electrical conductivity (EC) by fertilizer dosage, plant age, and sampling point

010.png

Figure 10. Corn plant measurement for classification

Table 7. Correlation coefficient for soil electrical conductivity (EC) and corn plant growth

011.png

Figure 11. Average plant classification at 30 days after planting (DAP) per plot of land

These results suggest that soil EC varied with plant age and fertilizer treatment, while the relationship between fertilizer treatment and EC depended on plant age. However, the results should be interpreted cautiously because each fertilizer treatment was applied to only one plot. Therefore, treatment effects cannot be separated from plot-specific characteristics, and the sampling points within each plot should not be considered independent experimental replicates. The analysis is consequently interpreted as exploratory evidence of differences among the monitored plots rather than as definitive evidence of causal fertilizer effects.

The correlation analysis revealed a strong positive relationship between soil EC and maize plant growth at both observation stages. At 30 DAP, the correlation coefficient was r = 0.79 (p = 0.001, n = 80), while at 45 DAP, the correlation coefficient was r = 0.72 (p = 0.004, n = 80). Both correlations were statistically significant. However, correlation does not imply causation. Other environmental and agronomic factors, including soil moisture, temperature, sunlight exposure, soil texture, and pH, may also influence plant growth and were not independently evaluated in this study. EC should be considered as one of several soil-condition indicators associated with maize growth rather than as a direct determinant of plant growth. Further research is recommended, including the study of additional environmental factors and direct analysis of soil chemistry, to determine the causal mechanisms behind the observed correlation between EC and growth.

The spatial EC mapping, plant-growth classification, exploratory ANOVA, and correlation analysis collectively characterize the spatial and temporal variation in measured soil EC and its association with maize growth. The Kriging-based interpolation showed changes in measured EC patterns across the observation phases, while the exploratory ANOVA identified crop age and fertilizer dosage as statistically significant terms within the model. However, because each fertilizer dosage was represented by a single plot, the fertilizer-related statistical results are interpreted as exploratory differences among the monitored plots rather than as replicated treatment effects. The correlation analysis further showed a positive association between measured EC and maize growth at 30 and 45 DAP.

It should be noted that soil EC is not a direct measure of nutrient availability but rather an indirect proxy influenced by multiple factors, including soil moisture content, temperature, salinity, soil texture, and the concentration of dissolved ions [24]. Corwin and Lesch [19] emphasized that apparent soil EC should be interpreted through its correlation with specific soil properties at a given site, since the same EC value can result from different combinations of salinity, moisture, and texture. Consequently, a high EC value does not necessarily indicate higher nutrient availability or improved soil fertility; it may instead reflect elevated salinity or moisture conditions unrelated to nutrient content. In this study, EC measurements and the available NPK sensor readings were considered as complementary soil measurements, while plant growth observations were used as biological response indicators. However, these sensor-derived measurements were not independently validated against laboratory soil chemical analyses. In addition, since no direct soil chemical analysis (e.g., laboratory-based nutrient testing) was conducted to confirm actual nutrient concentrations, the relationship between EC, nutrient availability, and crop response reported here should be interpreted with caution. Future studies incorporating laboratory soil chemical analysis are recommended to validate and strengthen the reliability of EC-based nutrient status estimation.

The authors gratefully acknowledge the financial support provided by the Directorate of Research, Technology, and Community Service (DRTPM), Directorate General of Higher Education, Research, and Technology, Ministry of Education, Culture, Research, and Technology of the Republic of Indonesia, through the Fundamental-Regular Research Scheme, under Research Contract No. 87/UN16.19/PT.01.03/PL/2025, Fiscal Year 2025. The authors also gratefully acknowledge the financial support for the presentation of this research at the international conference provided by the Indonesian Endowment Fund for Education (LPDP), on behalf of the Ministry of Higher Education, Science and Technology of the Republic of Indonesia, and managed under the EQUITY Program, Contract No. 4304/B3/DT.03.08/2025.