Estimating Soil Temperature Research Articles

Data-driven spatio-temporal estimation of soil moisture and temperature based on Lipschitz interpolation

This study estimates agricultural soil variables using a non-parametric machine learning technique based on Lipschitz interpolation. This method is adapted for the first time to learn spatio-temporal dynamics, accounting for two-dimensional spatial and one temporal coordinate inputs separately. The estimator is validated on real agricultural data, addressing challenges like measurement noise and quantization. The experimental setup, including an edge layer with measurement devices and a cloud layer for data storage and processing, is detailed. Despite its simplicity, the method presents a compelling alternative to Gaussian processes and neural networks.

Prediction of Soil Temperature in Wheat Field Using Machine Learning Models

ABSTRACT Wheat is a very vital cereal crop in the development of a nation’s economy and agricultural sector. A crucial element that profoundly affects the growth and productivity of wheat is the soil temperature. As an insulating layer, straw mulch controls the soil temperature. Field measurement of the soil temperature is time-consuming and costly; therefore, the present study investigates the applicability of random forest (RF), support vector machine (SVM) and generalized regression neural network (GRNN) models for estimating soil temperature in wheat fields covered with straw mulch using meteorological, soil cover and agronomical parameters. A field study was conducted on the wheat cultivar with two soil cover treatments for measuring the soil temperature at three depths (5, 50 and 80 cm). Soil plant analysis development value; soil cover and depth; daily meteorological data such as average air temperature, average relative humidity, sunshine hours and vapor pressure were used as inputs for machine learning models in training and testing. Statistical performance indicators showed applicability of three models with a root mean square error (RMSE) ranging from 0.245–0.595°C, 0.867–0.886°C and 0.557–0.812°C, Nash-Sutcliffe efficiency (NSE) ranging from 0.917–0.987, 0.823–0.836 and 0.804–0.935 and Lin’s concordance correlation coefficient (CCC) ranging from 0.949–0.991, 0.887–0.897 and 0.881–0.957 for the RF, SVM and GRNN, respectively. Results indicated better performance of the RF model for predicting the soil temperature at different depths in the case of both crop and soil cover conditions. Also, the uncertainty analysis indicated the better performance of RF with R-factor and P-factor value of 0.40 and 0.84, respectively.

Temperature and water content estimation in soils of the semi‐arid region of Brazil using finite difference and CFD

AbstractDetermining the temperature and water content of soil, at a given instant or along time, is fundamental to understand several soil‐related phenomena and processes. Evaporation, aeration, chemical‐reaction rates and types, biological processes such as germination and growth of seeds, root development, nutrient and water uptake by roots, and decomposition of organic matter by microbes, are all strongly influenced by soil temperature. On the other hand, infiltration of water through the soil surface allows soil to temporarily store water, making it available for uptake by plants and organisms living in soil. Furthermore, soil water content is closely related to physical and chemical properties of soil, such as oxygen content and demand, which impacts root breathing, microbial activity and soil chemical balance. The accurate evaluation of these two parameters and their interconnection is even relevant in semi‐arid regions, where climate conditions are particularly difficult, such as the north‐eastern zone of Brazil. Thus, the use of computational models and coupled approaches are imperative for rigorous descriptions. This work presents a contribution to estimate soil temperature and water content, by solving the heat transfer equation and the Richards equation, respectively, through finite differences. As input, the model uses the experimental material composition of the soil, the time‐dependent temperature profile at the surface and information about the regional rain regime. Three different numerical approaches were implemented: explicit, simple implicit and the Crank–Nicolson method. The calculations for temperature and water content of the soil obtained with these computational models were compared with the results from Computational Fluid Dynamics (CFD). The relative differences between the numerical methods were less than 0.006% by solving the heat transfer equation and less than 2.75% using the Richards equation. The maximum relative differences within the model, including both a constant and a variable water‐content profile, were 3.28%. The results from the computational model using the CFX tool have maximum relative differences of 0.6%, which contributes to verifying the accuracy of the implemented methods.

Soil temperature detection based on acoustic method and improved Wyllie model

Soil temperature is one of the important environmental factors for the underground parts of plants. It is important to detect soil temperature in agricultural production. Acoustic waves serve as effective carriers of soil information, providing a reliable means to detect soil physical properties. In order to detect the temperature of sandy loam soil based on acoustic technology, this study extends the application of Wyllie model to the temperature measurement of sandy loam soil. The relationship between soil temperature and acoustic velocity is explored. A model of soil moisture content-soil temperature-acoustic velocity (MTV) for temperature measurement in sandy loam soil is proposed by extending the temperature–velocity of sound relationship with the introduction of a variable empirical coefficient related to soil moisture β(θ). In order to determine the parameters of proposed model, a pulsed acoustic velocity detection system was built in this paper. The influence of temperature on the acoustic velocity in sandy loam soil was analyzed through experiments. Based on the experimental results, the key parameters of the MTV model are refined. Finally, a validation experiment was carried out on sandy loam soil. The maximum estimation error of soil temperature based on the MTV model for sandy loam soil temperature estimation is 8.55 %. The results indicate that the MTV model proposed in this study can be used for temperature estimation in sandy loam soil, providing a theoretical basis for the development of acoustic soil physical information detection sensors.

Appraisal of Visible/IR and microwave datasets for land surface fluxes estimation using machine learning techniques

Land surface fluxes such as Soil Moisture (SM) and Soil Temperature (ST) are very important variables for many applications that includes agriculture water management, weather and climate prediction, natural disasters etc. Further, they are important for understanding soil processes, hydrological balances as well as changes in microbial population. Mapping of the soil moisture content at various depth is crucial for the sustenance of water resources and also to understand about the development of crops in forms of quality and yield. With changing environmental conditions, there is a need of approaches for estimating SM and ST in various climatic and geographic situations. Towards this, Earth Observation datasets at higher resolutions from satellites such as Sentinel 1 and 2, could play an important role in the monitoring of SM and ST over the larger areas. For estimation of SM and ST, machine learning approaches could be effective. This research looked into the possibilities of using Earth Observation (EO) data of Sentinel-1 (S1) and Sentinel-2 (S2) simultaneously to estimate SM and ST by using the machine learning methods such as random forest (RF) and Support Vector Machines (SVM). The coefficient of correlation (r), root mean square error (RMSE), and Bias are utilized in model enactment for accuracy and comparative analysis of the models used. The overall analysis indicates that the SVM model (r = 0.85, RMSE = 2.54, Bias = −0.05) is the second most appropriate after the RF model (r = 0.89, RMSE = 2.34, Bias = 0) for estimating land surface fluxes (SM and ST).

Short‐term soil carbon mineralization on golf course sand‐based putting green and its effect on creeping bentgrass nitrogen uptake

AbstractA rapid soil test for accurately estimating nitrogen (N) supply from soil organic matter would help turfgrass managers develop N fertilization programs. In this study, we tested the feasibility of using a quick test, mineralizable carbon (min‐C), to predict turfgrass growth and N uptake from sand‐based putting green soils. A 2‐year field experiment was conducted in Verona, WI, on four sand‐based research golf greens planted with creeping bentgrass. Nitrogen fertilizer was applied at 0, 10, and 20 kg ha−1every 2 weeks as liquid urea during the growing seasons. Soil samples were collected at 0–5 cm and 5–10 cm and tested for min‐C by measuring the flush of CO2following the rewetting dried soil, where the soil was rewetted at 50% water‐filled pore space and incubated at 25°C for 24 h. Min‐C was greater at the shallower soil depth, and unaffected by N rates that were applied. On‐site air temperature, soil moisture content, and estimated soil temperature were negatively related with min‐C, wherer2 = 0.26–0.60 for air temperature,r2 = 0.32–0.71 for soil moisture content, andr2 = 0.16–0.53 for estimated soil temperature, indicating the test is quite sensitive to environmental conditions at the time of sampling. Furthermore, min‐C had weak positive to no relationship with creeping bentgrass growth rate (r2 = 0–0.17) and N uptake (r2 = 0–0.23) on sand‐based putting green, suggesting the test does not hold promise as a method for improving fertilizer decisions to these systems.

Development of a robust daily soil temperature estimation in semi-arid continental climate using meteorological predictors based on computational intelligent paradigms.

Changes in soil temperature (ST) play an important role in the main mechanisms within the soil, including biological and chemical activities. For instance, they affect the microbial community composition, the speed at which soil organic matter breaks down and becomes minerals. Moreover, the growth and physiological activity of plants are directly influenced by the ST. Additionally, ST indirectly affects plant growth by influencing the accessibility of nutrients in the soil. Therefore, designing an efficient tool for ST estimating at different depths is useful for soil studies by considering meteorological parameters as input parameters, maximal air temperature, minimal air temperature, maximal air relative humidity, minimal air relative humidity, precipitation, and wind speed. This investigation employed various statistical metrics to evaluate the efficacy of the implemented models. These metrics encompassed the correlation coefficient (r), root mean square error (RMSE), Nash-Sutcliffe (NS) efficiency, and mean absolute error (MAE). Hence, this study presented several artificial intelligence-based models, MLPANN, SVR, RFR, and GPR for building robust predictive tools for daily scale ST estimation at 05, 10, 20, 30, 50, and 100cm soil depths. The suggested models are evaluated at two meteorological stations (i.e., Sulaimani and Dukan) located in Kurdistan region, Iraq. Based on assessment of outcomes of this study, the suggested models exhibited exceptional predictive capabilities and comparison of the results showed that among the proposed frameworks, GPR yielded the best results for 05, 10, 20, and 100cm soil depths, with RMSE values of 1.814°C, 1.652°C, 1.773°C, and 2.891°C, respectively. Also, for 50cm soil depth, MLPANN performed the best with an RMSE of 2.289°C at Sulaimani station using the RMSE during the validation phase. Furthermore, GPR produced the most superior outcomes for 10cm, 30cm, and 50cm soil depths, with RMSE values of 1.753°C, 2.270°C, and 2.631°C, respectively. In addition, for 05cm soil depth, SVR achieved the highest level of performance with an RMSE of 1.950°C at Dukan station. The results obtained in this research confirmed that the suggested models have the potential to be effectively used as daily predictive tools at different stations and various depths.

Soil temperature estimation at different depths over the central Tibetan Plateau integrating multiple Digital Earth observations and geo-computing

ABSTRACTSoil temperature (ST) plays a critical role in ecosystems. Monitoring high-resolution ST profiles remains challenging due to the inherent heterogeneity of ST in space, time, and depth. To address this challenge, in this study, we integrate remote sensing techniques and deep learning methods to retrieve spatiotemporal continuous ST from 2011 to 2020 at four depths (5, 10, 20, and 40 cm) over the central Tibetan Plateau (TP). Landsat and MODIS observations were fused to obtain land surface properties at different resolutions. The fused variables were integrated with in-situ ST and soil moisture (SM) measurements to estimate spatiotemporally continuous (0.0005°, 0.0025°, and 0.0125°) ST profile through deep learning-based training models. The deep belief network (DBN) was applied to estimate ST: (a) from layer to layer (LW-DBN), and (b) from land surface properties directly (Direct-DBN). The ten-fold cross-validation indicates that both approaches achieve promising results (R2 > 0.836, MAE < 2.152 °C), and the Direct-DBN outperformed the LW-DBN at all spatial scales and depths. ST retrieval at deeper depths and coarser resolutions tend to have better monitoring accuracy. Further analysis indicates an increment of ST at all four depths over the study period, which provides valuable insights into global warming.

Deep learning approach for one-hour ahead forecasting of weather data

ABSTRACT Weather is made up of multiple parameters, including solar radiation (SR), atmospheric pressure (AP), soil temperature (ST), atmospheric temperature (AT), wind speed (WS), relative humidity (RH), and sunshine duration (SD). These factors are also crucial for the renewable energy sector, solar simulation, agriculture, air pollution, water supply and distribution, avalanche warning, forestry, and town and regional planning. A deep learning method based on a neural network with Long Short-Term Memory (LSTM) was employed in this investigation for one-hour-ahead weather data forecasting. The ability of the LSTM model was compared with the Adaptive Neuro-Fuzzy Inference System (ANFIS) with that of the fuzzy c-means (FCM), Autoregressive Integrated Moving Average (ARIMA) model, and the Autoregressive Moving Average (ARMA) model. Mean absolute error (MAE), correlation coefficient (R), root means square error (RMSE), average bias, Nash – Sutcliffe efficiency coefficient (NSE), and mean absolute percentage error (MAPE) were selected as evaluation criteria. Results indicated that the proposed LSTM model presented good enough results compared to other used methods. 7 different types of meteorological data from a total of 4 years (35040 hours) were divided into 25% test data and 75% training data for the models. The best result was obtained for the hourly ST estimation of Adana province using the LSTM method, the MAE, RMSE, R, bias, NSE, and MAPE values were computed as 0.016°C, 0.078°C, 0.9999, −0.00018°C, 0.0805%, and 0.9999, respectively. On the other hand, the worst result was obtained for the hourly SD for Mardin province when ARIMA was used, and the statistical measures were derived as 0.128 hours for MAE, 0.215 hours for RMSE, 0.8851 for R, 0.00091 hours for bias, and 0.7657 for NSE. In this regard, it is demonstrated that the LSTM technique outperformed the other models in terms of all-weather data estimates and delivered highly sensitive outcomes.

A Review of Machine Learning Approaches to Soil Temperature Estimation

Soil temperature is an essential factor for agricultural, meteorological, and hydrological applications. Direct measurement, despite its high accuracy, is impractical on a large spatial scale due to the expensive and time-consuming process. On the other hand, the complex interaction between variables affecting soil temperature, such as topography and soil properties, leads to challenging estimation processes by empirical methods and physical models. Machine learning (ML) approaches gained considerable attention due to their ability to address the limitations of empirical and physical methods. These approaches are capable of estimating the variables of interest using complex nonlinear relationships with no assumptions about data distribution. However, their sensitivity to input data as well as the need for a large amount of training ground truth data limits the application of machine learning approaches. The current paper aimed to provide a review of ML techniques implemented for soil temperature modeling, their challenges, and milestones achieved in this domain.

Investigation of earth air heat exchangers functioning in arid locations using Matlab/Simulink

This study presents a MATLAB/Simulink model to predict the distribution of soil temperatures and design Earth-to-Air Heat Exchangers (EAHEs) efficiently. It compares four EAHE configurations, including single-pipe, multi-pipe, multiple-single pipes (MS-pipes), and twisted-single pipes (TS-pipes) systems. The presented model is validated with reliable results and allows for easy modifications to obtain the final design. Results show that MS-pipe EAHE is the better in terms of pressure losses and good in terms cooling potential. The study proves that the installation of the EAHEs in Kufa, Iraq, is highly efficient, achieving cooling potential of 1626 W in August and 129.8 W in March. Additionally, the study highlights the influence of the geometric configuration of EAHEs on both their flow behavior and thermal efficiency, presenting an equation for comparing pipe lengths at different soil depths. This study fills the knowledge gap in the use of MATLAB simulation for designing EAHEs and estimating soil temperatures distribution and it compares four EAHEs configurations.

Performance of Different Crop Models in Simulating Soil Temperature

Soil temperature is one of the key factors to be considered in precision agriculture to increase crop production. This study is designed to compare the effectiveness of a land surface model (Noah Multiparameterization (Noah-MP)) against a traditional crop model (Environmental Policy Integrated Climate Model (EPIC)) in estimating soil temperature. A sets of soil temperature estimates, including three different EPIC simulations (i.e., using different parameterizations) and a Noah-MP simulations, is compared to ground-based measurements from across the Central Valley in California, USA, during 2000-2019. The main conclusion is that relying only on one set of model estimates may not be optimal. Furthermore, by combining different model simulations, i.e., by taking the mean of two model simulations to reconstruct a new set of soil temperature estimates, it is possible to improve the performance of the single model in terms of different statistical metrics against the reference ground observations. Containing ratio (CR), Euclidean distance (dist), and correlation co-efficient (R) calculated for the reconstructed mean improved by 52%, 58%, and 10%, respectively, compared to both model estimates. Thus, the reconstructed mean estimates are shown to be more capable of capturing soil temperature variations under different soil characteristics and across different geographical conditions when compared to the parent model simulations.

Spatial Interpolation of Soil Temperature and Water Content in the Land-Water Interface Using Artificial Intelligence

The distributed measured data in large regions and remote locations, along with a need to estimate climatic data for point sites where no data have been recorded, has encouraged the implementation of spatial interpolation techniques. Recently, the increasing use of artificial intelligence has become a promising alternative to conventional deterministic algorithms for spatial interpolation. The present study aims to evaluate some machine learning-based algorithms against conventional strategies for interpolating soil temperature data from a region in southeast Canada with an area of 1000 km by 550 km. The radial basis function neural networks (RBFN) and the deep learning approach were used to estimate soil temperature along a railroad after the spline deterministic spatial interpolation method failed to interpolate gridded soil temperature data on the desired locations. The spline method showed weaknesses in interpolating soil temperature data in areas with sudden changes. This limitation did not improve even by increasing the spline nonlinearity. Although both radial basis function neural networks and the deep learning approach had successful performances in interpolating soil temperature data even in sharp transition areas, deep learning outperformed the former method with a normalized RMSE of 9.0% against 16.2% and an R-squared of 89.2% against 53.8%. This finding was confirmed in the same investigation on soil water content.

The Estimation of Monthly Mean Soil Temperature at Different Depths in Sivas Province, Turkey by Artificial Neural Networks

ABSTRACT In this study, soil temperature of Sivas province was estimated by the artificial neural networks (ANNs) method using data obtained from five different meteorological measurement stations situated in provincial borders. Nineteen years of (2000–2018) monthly mean air temperature data obtained from five different soil depths (5, 10, 20, 50 and 100 cm) was used for ANN analysis. Predicted and measured soil temperatures were strongly correlated with determination coefficient (R2) values ranging between 0.9767 and 0.9941. Mean Absolute Error (MAE) ranged from 0.532°C to 1.381°C, while Mean Absolute Percentage Error (MAPE) ranged from 5.692% to 16.263% and Root Mean Squared Error (RMSE) ranged between 0.694°C and 1.666°C. It was found that the predicted values are in good agreement with the measured data. However, there was a tendency to underestimate the soil temperature.

Uzun Kısa Süre Bellek (LSTM) ile Toprak Sıcaklığının Tahmini

Soil temperature not only affects many soil properties, but also has a significant effect on plant development. Knowing and correct estimation of soil temperature is important for both soil management and crop production. The accuracy of temperature forecasts is very important, especially for the countries that stand out with their agriculture-based economies. Therefore, in recent years, different artificial intelligence methods have been used in soil temperature predictions. Deep learning methods lead the way in achieving high prediction accuracy. In this study, a Long Short-Term Memory (LSTM) network, which is a deep learning (DL) sub-architecture, is proposed to create an effective model for soil temperature prediction. The data used in the study are the daily soil temperatures at a depth of 50 cm for the years 2013-2021 of Bingöl province. For the training of the proposed LSTM model, 89% of the data set within the scope of the study was used, and. The remaining 11% was estimated by the model for assessing model success. The RMSE value as a result of the estimation made by the trained LSTM model was obtained as 1,25. The high estimation accuracy of the proposed model showed that this model could be successfully applied in temperature data estimation studies.

Management and environmental factors associated with simulated restoration seeding barriers in sagebrush steppe

Adverse weather conditions, particularly freezing or drought, are often associated with poor seedling establishment following restoration seeding in drylands like the Great Basin sagebrush steppe (U.S.A.). Management decisions such as planting date or seed source could improve restoration outcomes by reducing seedling exposure to weather barriers. We simulated the effects of management and environmental factors on seedling exposure to post‐germination barriers for bottlebrush squirreltail (Elymus elymoides), Sandberg bluegrass (Poa secunda), and bluebunch wheatgrass (Pseudoroegneria spicata). We combined germination timing models with daily soil moisture and temperature estimates to calculate yearly germination favorability and post‐germination freezing and drought barriers for three planting dates (15 October, 15 November, and 15 March) and three seed sources or cultivars per species for 5,000 sites in each of 40 years (water years 1980–2019). We tested the effects of site environmental variables (elevation, mean annual precipitation, heat load, and clay content) and management choices (seed source and planting date) on germination favorability and barrier occurrence (mean) and variability (coefficient of variation). Seedling exposure to barriers was strongly linked to management decisions in addition to site mean precipitation and elevation. Later fall plantings and seed sources with slower germination (lower mean germination favorability) were less likely to encounter freezing and drought barriers. These results suggest that management actions can play a role comparable to site environmental variables in reducing exposure of vulnerable seedlings to adverse weather conditions and subsequent effects on restoration outcomes.

Evaluating the most effective climatic parameters affecting the monthly mean soil temperature estimates using the PLS method

Evaluating the most effective climatic parameters affecting the monthly mean soil temperature estimates using the PLS method

Modelling Soil Temperature by Tree-Based Machine Learning Methods in Different Climatic Regions of China

Accurate estimation of soil temperature (Ts) at a national scale under different climatic conditions is important for soil–plant–atmosphere interactions. This study estimated daily Ts at the 0 cm depth for 689 meteorological stations in seven different climate zones of China for the period 1966–2015 with the M5P model tree (M5P), random forests (RF), and the extreme gradient boosting (XGBoost). The results showed that the XGBoost model (averaged coefficient of determination (R2) = 0.964 and root mean square error (RMSE) = 2.066 °C) overall performed better than the RF (averaged R2 = 0.959 and RMSE = 2.130 °C) and M5P (averaged R2 = 0.954 and RMSE = 2.280 °C) models for estimating Ts with higher computational efficiency. With the combination of mean air temperature (Tmean) and global solar radiation (Rs) as inputs, the estimating accuracy of the models was considerably high (averaged R2 = 0.96–0.97 and RMSE = 1.73–1.99 °C). On the basis of Tmean, adding Rs to the model input had a greater degree of influence on model estimating accuracy than adding other climatic factors to the input. Principal component analysis indicated that soil organic matter, soil water content, Tmean, relative humidity (RH), Rs, and wind speed (U2) are the main factors that cause errors in estimating Ts, and the total error interpretation rate was 97.9%. Overall, XGBoost would be a suitable algorithm for estimating Ts in different climate zones of China, and the combination of Tmean and Rs as model inputs would be more practical than other input combinations.

Modeling Soil Temperature for Different Days Using Novel Quadruplet Loss-Guided LSTM.

Soil temperature (Ts), a key variable in geosciences study, has generated growing interest among researchers. There are many factors affecting the spatiotemporal variation of Ts, which poses immense challenges for the Ts estimation. To enrich processing information on loss function and achieve better performance in estimation, the paper designed a new long short-term memory model using quadruplet loss function as an intelligence tool for data processing (QL-LSTM). The model in this paper combined the traditional squared-error loss function with distance metric learning between the sample features. It can zoom analyze the samples accurately to optimize the estimation accuracy. We applied the meteorological data from Laegern and Fluehli stations at 5, 10, and 15 cm depth on the 1st, 5th, and 15th day separately to verify the performance of the proposed soil temperature estimation model. Meanwhile, this paper inputs the variables into the proposed model including radiation, air temperature, vapor pressure deficit, wind speed, air pressure, and past Ts data. The performance of the model was tested by several error evaluation indices, including root mean square error (RMSE), mean absolute error (MAE), Nash-Sutcliffe model efficiency coefficient (NS), Willmott Index of Agreement (WI), and Legates and McCabe index (LMI). As the test results at different soil depths show, our model generally outperformed the four existing advanced estimation models, namely, backpropagation neural networks, extreme learning machines, support vector regression, and LSTM. Furthermore, as experiments show, the proposed model achieved the best performance at the 15 cm depth of soil on the 1st day at Laegern station, which achieved higher WI (0.998), NS (0.995), and LMI (0.938) values, and got lower RMSE (0.312) and MAE (0.239) values. Consequently, the QL-LSTM model is recommended to estimate daily Ts profiles estimation on the 1st, 5th, and 15th days.

Soil Temperature Prediction via Self-Training: Izmir Case

This paper proposes a new model, called Soil Temperature prediction via Self-Training (STST), which successfully estimates the soil temperature at various soil depths by using machine learning methods. The previous studies on soil temperature prediction only use labeled data which is composed of a variable set X and the corresponding target value Y. Unlike the previous studies, our proposed STST method aims to raise the sample size with unlabeled data when the amount of pre-labeled data is scarce to form a model for prediction. In this study, the hourly soil-related data collected by IoT devices (Arduino Mega, Arduino Shield) and some sensors (DS18B20 soil temperature sensor and soil moisture sensor) and meteorological data collected for nearly nine months were taken into consideration for soil temperature estimation for future samples. According to the experimental results, the proposed STST model accurately predicted the values of soil temperature for test cases at the depths of 10, 20 30, 40, and 50 cm. The data was collected for a single soil type under different environmental conditions so that it contains different air temperature, humidity, dew point, pressure, wind speed, wind direction, and ultraviolet index values. Especially, the XGBoost method combined with self-training (ST-XGBoost) obtained the best results at all soil depths (R2 0.905-0.986, MSE 0.385-2.888, and MAPE 3.109%-8.740%). With this study, by detecting how the soil temperature will change in the future, necessary precautions for plant development can be taken earlier and agricultural returns can be obtained beforehand.