The development of the water usage for planting crops under the smart farm system is based on the measurement of moisture in the soil on actual time using a sensor which is a concept in increasing efficiency of water usage and crop yield in consideration of the frequent drought problem resulting from various conditions occurring in the country. This research was aimed to evaluate a sensor system that can measure soil moisture considering the significantly related factors such as the type of soil and the kind of crop together with analysis of some physical and chemical properties of soil samples from 2 types of planting systems, namely, rice planting and longan cultivation. The resulting soil samples from these 2 planting areas were collected and were used to establish a predicted equation using factors such as area altitude and soil depth in order to create an equation to compare each sample. In this study, there were 16 soil samples from rice planted fields and 10 soil samples taken from longan-planted areas, which were then used to assess and compare the predicted equation on the 3 methods of determining soil moisture, namely: by weight, by volume using SM150 Soil Moisture Kit and by using a sensor that measures actual moisture. Data analysis was conducted by using the Sigma Plot program in order to analyze the 4 designs of equation, namely: Exponential decay (3 parameters); Linear Exponential decay (2 parameters) and Hyperbolic decay (2 parameters) with the use of a sensor that can measure actual soil moisture, by weight, and by volume using SM150 Soil Moisture Kit. The results from the sub-research of the first experiment were later studied in terms of the mathematical equation for the calibration curve and mathematical equation for the prediction to determine soil moisture by considering the correctness and precision through comparison of the resulting soil properties. This was done by applying the resulting mathematical equation with the measured values from the sensor based on statistical techniques for nonparametric regression. The correlation between the level of moisture determined by weight and by volume resulting from the mathematical equation (predicted) with volume of moisture by weight and actual measurement, was done by determining the correlation coefficient (R2) and the probability value (P), root mean square error (RMSE) and ration per standard deviation (RPD), so as to assess the correctness and precision of the predicted equation. The study from laboratory to find out calibration curve and mathematical equation, and assessment of the correctness and precision of the predicted equation.Results of the study showed that soil properties such as texture and amount of organic matter in the soil, were correlated with moisture in the soil and its capacity to store moisture at the field level and the ability to store moisture at the temporary withering point of that type of soil, which may affect the sensor that is used to measure soil moisture as followed. Further results of the study also indicated that different types of soils when applied with similar level of moisture, the amount of moisture measured by the sensor showed different results at the same period of time. In determining the predicted equation, this study was able to establish 128 predicted equations for soil planted to rice and 40 equations for soils planted to longan and also found that the forms of equation were suitable for the prediction of values measured by the sensor and which were changed to become moisture by volume and by weight, including equations of Exponential decay (3 parameters) which are forms of equation that are most suitable in terms of RMSE value. It was also found that the equation in the form of Exponential decay (3 parameters) had the lowest RMSE value in each sample of soil planted to rice and almost each sample of soil planted to longan. This was in confirmation with the RPD value which found that Exponential decay equation (3 parameters) showed the highest values from samples of soils planted to rice and longan. However, Linear type showed also low RMSE values and high RPD also.
As for the assessment of the performance of the sensor at the field level, the study found that moisture measured by the predicted equation of that soil sample altogether considering data of the site planted to crop cultivation, rainfall data, soil types and soil depth, showed agreement with soil moisture measured by the sensor. From this study, it might be said that the use of the sensor to determine soil moisture as actual data reported at that time and resulting soil moisture data directly corresponding to the actual data of the site, has the potential to be used together with watering system of the crop. However, creating the predicted equation should be implemented uniformly together with each type of soil which suggest further study on creating prediction equation of the group type of soil samples that are important to agriculture and which must be created as a ready-to-use tool to make it easy for use in the future.