The main objective of this study is to develop a general methodology for predicting soil temperature based on general circulation data. To meet this demand, we used temperature data that can be profitably used to predict soil temperature in a period of 20 years. Accordingly, air temperature data were downscaled to 2016–2025 based on LARS-WG data. The obtained results indicated that the model has precisely predicted minimal and maximal temperatures. According to the results, the best correlation methods are S, cubic, and quadratic. To investigate soil temperature changes, the predicted data were classified and categorized into two separate decades (2016–2025 and 2026–2035). The results showed that air temperature increases to 1 °C and 1.2 °C in the first decade (2016–2025) and the second decade (2026–2035), respectively, but varies in different regions. The predicted air temperature is lower in the eastern part of the region. In the central region, air and soil temperatures are predicted to be greater than that of other regions. It should also be mentioned that a variety of temperature changes are related to the depth of soil.

INTRODUCTION

Increasing global temperatures cause broad, profound changes in the Earth's climate, which results in damage, especially in recent decades. Today, most environmental initiatives have tried to recognize that climate change is an important issue on a global scale. Research performed throughout the world, and in Iran, on soil temperature, is basically limited to regression methods. These methods still represent a meaningful relationship between research data and the acceptable air temperature data. It is also believed that soil temperature varies based on the actual number. Results in the past 20 years (1990–2010) have employed different models based on research data and soil temperature data in analyzing and calculating soil temperature data. According to these results, the prediction of soil temperature is possible. Today, the issue of climate change and its effects on various aspects of human life have become one of the most important environmental, political, economic, and social issues. Changes in climate parameters such as precipitation and temperature are coupled with adverse consequences, including flood, drought, loss of biodiversity, and agricultural production (Karimi et al. 2014; Taxak et al. 2014).

The quality and the quantity of crops depend on many factors such as the soil and its temperature. Soil temperatures significantly affect the budding and growth rates of plants. For example, with the increase in soil temperature, chemical reactions speed up and cause seeds to germinate. Soil temperature plays an important role in the decomposition of soil. It also regulates many processes, including the rate of plant development and growth. Soil temperature also plays an important role for setting the life cycles of small creatures that live in the soil (Fahim Ahmad & Rasul 2008). It also determines the state of the water in the soil (i.e., liquid, gaseous, or frozen state). In cold soil, the rate of decomposition of organic matter is slow because the microorganisms function at a slower rate; as a result, the color of soil is dark. The successful prediction of soil temperature with the help of air temperature minimizes the time, cost, and equipment maintenance necessary for on-site monitoring and helps researchers to use data from other sources as well. Air temperature correlates well with soil temperature because both are determined by the energy balance at ground surface (Buringh 1984; Pritchett & Fisher 1987; Zheng et al. 1993). Climate change as a result of global warming has caused damage in recent decades. Today, most environmental projects attempt to identify climatic variations and their prediction. Thus, the importance of climate change is necessary to make efforts towards the further understanding of how climate change events occur. As predicting future climate is not fully possible due to climate change impacts, there is an alternative solution for specifying different possibilities, climate change scenarios. Currently, the most reliable tool for generating scenarios are general circulation models (GCMs). These models are based on physical laws that are solved by mathematical equations in a 3D grid on the Earth's surface. Some of these models are USCLIMATE Weather Generators (WGEN), GEM, LARS-WG, CLIMGEN, and SDSM. GCMs, powerful hardware, and time-consuming downscaling and statistical models to manufacture the weather have been developed. These models convert general circulation data from large scale to small scale using GCM output models and using specific scenarios of climate data production models (Semenov & Arrow 1997). The advantages of these models are low cost, high speed, and the possibility of use without supercomputers. LARS-WG, AXE-JEN, and WG-PCA are stochastic models that also benefit from the results of GCMs. LARS-WG is one of the most popular models of stochastic weather generator data in the series approach, which is used to produce large amounts of precipitation, solar radiation, and maximum and minimum daily temperatures in a station under baseline and future climate changes. A preliminary version of this model was developed as part of the project that studied the risk assessment of agriculture in Hungary, Budapest, in 1990. The LARS-WG model in the new version used an empirical probability distribution function instead of normal distribution to estimate the temperature. Also, in this version, the range of values was increased close to zero probability and one for better estimation of extreme values. Applying the results of 15 GCMs of the LARS-WG is a possibility that other small-scale models do not have. There are some research studies on Iran and other countries on soil temperature but not on its prediction and possible variation. Semenov (2008), using the LARS-WG model, studied temperature and precipitation at 20 stations in the UK, which were located in different climates. Chen et al. (2013) used the LARS-WG model to predict precipitation and minimum and maximum temperatures in Sudan and South Sudan. Toy et al. (1978) used monthly mean air temperature to predict monthly soil temperature at continental scales. Most studies throughout the world used only one downscaling method, and many researchers applied the change factor (CF) for downscaling climatic variables (Reynard et al. 2001; Diaz Nieto & Wilby 2005; Minville et al. 2008; Tabor & Williams 2010). In other research studies, artificial (stochastic) weather generators, including WGEN, CLIMGEN, CLIGEN, and LARS-WG, were employed for downscaling outputs of GCMs (Babaeian et al. 2004; Dibike & Coulibaly 2005; Abbasi et al. 2011). In this study, the LARS-WG model was used for the purpose of downscaling coarse resolution data of the study area. Two important reasons for using the LARS-WG model include the provision of a means to simulate synthetic weather time-series with certain statistical properties, which are long enough to be used in a risk assessment of hydrological or agricultural applications, and the provision of a means to extend the simulation of weather time-series to unobserved locations. In fact, LARS-WG has been used in various studies, including the assessment of the impacts of climate change (Barrow & Semenov 1995; Dubrovsky 1996; Semenov & Barrow 1997; Weiss et al. 2003; Lawless & Semenov 2005; Khan et al. 2006; Scibek & Allen 2006; Semenov 2007; Semenov & Doblas 2007). The process of generating synthetic weather data can be divided into three distinct steps as follows: (1) model calibration, (2) model validation, and (3) generation of synthetic weather data (Nakicenovic & Swart 2000; Zhang & Garbrecht 2003). Soil surface temperature should correlate well with air temperature due to the energy balance at the ground surface (Zheng et al. 1993). The measuring process of soil temperature is very intricate, expensive, and lengthy. It is too technical to install the thermometer properly (Fahim Ahmad & Rasul 2008). The anticipated outcome is that a strong correlation will be followed with an action to seek an equation by means of forecasting soil temperature from the existing atmospheric temperature through regression analysis. The primary step to validate the correlation is to collect a series of data for two variables (i.e., atmospheric temperature and soil temperature). The coefficient of correlation is determined through Pearson's distribution, whereas atmospheric temperature is considered as an independent variable and soil temperature is a dependent variable. This coefficient is a valid indicator of the correlation between two variables based on statistical data. The coefficients are calculated by distinctly taking soil temperature at several depths as commensurate with air temperature. The procedure is concluded with regression analyses that generate equations through which the dependent variable (soil temperature) is projected corresponding to the independent variable (air temperature). Toy et al. (1978) used monthly mean air temperature to predict monthly soil temperature at continental scales. There is a wide selection of climate models available to provide scenarios of future climate change. All are mathematical models that simulate the function of the global climate system, varying in size (computer space), scope (atmosphere, ocean, sea-ice, and land-surface components), scale (horizontal spacing and grid size), and complexity (parameterization schemes) (Xie & Cheng 2006; Zhao et al. 2006; Muttil & Chau 2007; Wang et al. 2014). Due to the importance and numerous applications of soil temperature prediction in agricultural planning, an effort was made to predict and investigate the soil temperature in the western region of Iran in a 20-year period by employing atmosphere general circulation data and ground observed data of earth surface. We used the model to downscale air temperature data for a time span between 2016 and 2025. As atmospheric general circulation data predict air temperature (Max and Min), to predict soil temperature data, we used a regression method to determine a significant relationship between air temperature data and different depths of soil data by employing ground observed data of 1990–2010.

DATA SETS AND METHODS

Study area

The study area is located between 33–37 northern latitude and 45–50 eastern longitude. The region of study is located in the west of Iran, between Zagros and Alborz mountains, known as the cold region due to its high altitude. Owing to its specific geographical situation, suitable quality of air and water, and suitable soil, the study area has great capabilities of agricultural production and climatic change can impose irreversible damage on the region's economy. In this study, the statistics of the seven different meteorology stations (Hamadan-Airport, Hamadan-Nojeh, Islamabad, Kangavar, Kermanshah, Sanandaj, and Saqez) were used in the years 1990–2010. The stations are listed in Table 1 and shown in Figure 1.
Table 1

Properties of weather stations in studied area

Station Height (m) Longitude Latitude 
Islamabad 1,348 46.28 34.07 
Kermanshah 1,318 15.47 34.35 
Hamedan-Noje 1,679 43.48 15.35 
Hamedan-Airport 1,741 48.32 34.52 
Kangavar 1,468 47.59 34.30 
Sanandaj 1,373 47.00 35.20 
Saqez 1,522 46.16 36.15 
Station Height (m) Longitude Latitude 
Islamabad 1,348 46.28 34.07 
Kermanshah 1,318 15.47 34.35 
Hamedan-Noje 1,679 43.48 15.35 
Hamedan-Airport 1,741 48.32 34.52 
Kangavar 1,468 47.59 34.30 
Sanandaj 1,373 47.00 35.20 
Saqez 1,522 46.16 36.15 
Figure 1

Geographical situation of the studied area.

Figure 1

Geographical situation of the studied area.

Methodology

To predict future changes in soil temperature in the study area, the data stations surveyed (1990–2010) and the data predicted by GCMs over the 20-year period of 2016–2035 were used. In this regard, the following steps were performed.

Validation of the data, the actual air temperature, and soil temperature

To evaluate the accuracy and homogeneity of data before doing any calculations based on actual air temperature and soil temperature (at depths of 5 and 10, 30, 50, and 100 cm) during the years 1990–2010, a homogeneity test was performed.

Prediction air temperature data using LARS-WG model data and ECHAM4 + HOPE-G (ECHO-G) GCM

The decade-scale climate prediction methods are higher due to computational limitations and time, with some problems. In this time scale, there are major limitations to spatial resolution models. Nowadays, the resolution of the GCMs is in the range of several hundred kilometers. There are two approaches, that include using statistical models and statistical downscaling using regional dynamic models. The use of dynamic models of GCMs for downscaling the output of the model is faced with a time limit. In the second method, it could be time statistics for a site-specific statistical downscaling. Among the statistical methods are CLIGEN, GEM, LARS-WG, and SDSM (Babaean et al. 2010). In this study we used the A1 scenario ECHO-G model, previously used at the University of Hamburg, Germany and the South Korea Center for Atmospheric Research, to evaluate regional climate change study using statistical downscaling. The data model ECHO-G GCMs include rainfall, minimum temperature, maximum temperature, and radiation extracted for each network GCM. A particular scenario of the LARS-WG model was developed (Babaean et al. 2010). Precipitation and the occurrence of probability distribution of the quasi-experimental methods and the modeling of radiation Markov chain model based on semi-empirical distribution and temperature using Fourier series were determined (Semenov & Barrow 1997). The LARS model should first determine each station to generate data, including the name, location, and altitude as well as files of daily meteorological data as input to the model (observation period). Then, the LARS model should deploy an alternative analysis of the data from a text file that contains a summary of the statistical properties of the observed average monthly and average quarterly data for the entire period under investigation. The trend in the time-series data model is used to reproduce the observed station data for the same period of time. Finally, using statistical analysis and charting, monthly mean simulated data are compared to the observed data, and the model's ability to simulate meteorological data at these stations is evaluated. After using the capability evaluation model in each station to build the data for the coming period, it is necessary to file the scenario of climate change according to the output of GCMs for the study site as developed and defined for the model (Semenov & Barrow 2002). The ECHO-G model compares the scenario data network in the 2016–2035 period (evaluation period) to the period 1990–2010. The model outputs include only the minimum temperature, maximum temperature, rainfall, and radiation of the weather stations. In this study, the investigated climatic stations were modeled and evaluated. The LARS-WG model for downscaling the statistical model for use as a base case scenario for the period 1980–2010 was prepared. The LARS-WG model was implemented based on these scenarios. The results represent a good agreement between modeled and observed temperatures. Generally, the function of meteorological variable Drmdlsazy LARS is appropriate and it can be used to reconstruct the station data in the last period or extend the length of the data used in future periods. Also, it can be used to evaluate future climate at a local level (Meshkati et al. 2010). The data for the medium-sized minimum and maximum air temperature over the period 2016–2025 based on the study of the stations of this model were extracted.

Investigating different models of regression to study the governing relationships on air and soil temperature

As the soil temperature obeys the air temperature in meteorological stations, the relationships between real data of air temperature and soil temperature were investigated in terms of different models of regression.

If y = α + βx is supposed to be an unknown line equation, the values α and β are obtained by the least-squares method as: 
formula
1
It is proven that the regression line slope is obtained by relationship β = rSy/Sx, where r is the correlation coefficient between x and y. Sx and Sy are the standard deviations (SDs) of x and y, respectively.
When the variation of a parameter (soil temperature) is taken from other factors, ideally we use the determination coefficient instead of the correlation coefficient to assess the effect of the variable from other factors. The value of the determination coefficient is obtained from the following: 
formula
2
where is the predicted value for Y obtained from Equation (1), Y is the real value of y obtained in the problem, and is the mean of the Y value (obtained in the problem).

Other relationships used in this research are summarized in Table 2.

Table 2

Relationships used in this research

Linear  
Quadratic  
Cubic  
Logarithmic  
Exponential  
Inverse  
Power  
Compound  
 
Logistic  
Growth  
Linear  
Quadratic  
Cubic  
Logarithmic  
Exponential  
Inverse  
Power  
Compound  
 
Logistic  
Growth  

Finally, the best method was chosen. To show the method, Sanandaj station was chosen as the representative and the relationships and graphs are shown in terms of regression methods in different depths of soil.

As seen in Tables 38, the maximum determination coefficient (r2) shows that the variable values by the S, cubic, and quadratic methods have similar behavior to the values of Sanandaj soil temperature in different depths.

Table 3

Model description (Sanandaj station)

Model name MOD_35 
Dependent variable 5 cm 
Equation Linear 
Logarithmic 
Inverse 
Quadratic 
Cubic 
Compound 
Power 
Growth 
10 Exponential 
11 Logistic 
Independent variable Mean 
Constant Included 
Variable whose values label observations in plots Unspecified 
Tolerance for entering terms in equations 0.0001 
Model name MOD_35 
Dependent variable 5 cm 
Equation Linear 
Logarithmic 
Inverse 
Quadratic 
Cubic 
Compound 
Power 
Growth 
10 Exponential 
11 Logistic 
Independent variable Mean 
Constant Included 
Variable whose values label observations in plots Unspecified 
Tolerance for entering terms in equations 0.0001 
Table 4

Model summary and parameter estimates in 5 cm soil depth (Sanandaj station)

  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.721 28.374 11 0.0000 6.305 0.868   
Logarithmic 0.721 28.494 11 0.0000 −12.853 11.869   
Inverse 0.720 28.331 11 0.0000 30.021 −161.437   
Quadratic 0.722 12.957 10 0.0020 0.361 1.737 −0.032  
Cubic 0.722 12.960 10 0.0020 2.230 1.315 0.000 −0.001 
Compound 0.733 30.251 11 0.0000 9.459 1.049   
Power 0.735 30.573 11 0.0000 3.297 0.653   
0.735 30.577 11 0.0000 3.551 −8.883   
Growth 0.733 30.251 11 0.0000 2.247 0.048   
Exponential 0.733 30.251 11 0.0000 9.459 0.048   
Logistic 0.733 30.251 11 0.0000 0.106 0.953   
  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.721 28.374 11 0.0000 6.305 0.868   
Logarithmic 0.721 28.494 11 0.0000 −12.853 11.869   
Inverse 0.720 28.331 11 0.0000 30.021 −161.437   
Quadratic 0.722 12.957 10 0.0020 0.361 1.737 −0.032  
Cubic 0.722 12.960 10 0.0020 2.230 1.315 0.000 −0.001 
Compound 0.733 30.251 11 0.0000 9.459 1.049   
Power 0.735 30.573 11 0.0000 3.297 0.653   
0.735 30.577 11 0.0000 3.551 −8.883   
Growth 0.733 30.251 11 0.0000 2.247 0.048   
Exponential 0.733 30.251 11 0.0000 9.459 0.048   
Logistic 0.733 30.251 11 0.0000 0.106 0.953   
Table 5

Model summary and parameter estimates in 10 cm soil depth (Sanandaj station)

  Model summary
 
Parameter estimates
 
Equation R Square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.835 55.831 11 0.000 3.740 0.984   
Logarithmic 0.831 54.190 11 0.000 −19.200 13.914   
Inverse 0.826 52.263 11 0.000 31.553 −196.306   
Quadratic 0.841 26.530 10 0.000 26.506 −2.226 0.113  
Cubic 0.842 26.575 10 0.000 16.189 0.000 −0.047 0.004 
Compound 0.839 57.113 11 0.000 8.065 1.057   
Power 0.835 55.784 11 0.000 2.216 0.783   
0.831 54.115 11 0.000 3.653 −11.058   
Growth 0.839 57.113 11 0.000 2.088 0.055   
Exponential 0.839 57.113 11 0.000 8.065 0.055   
Logistic 0.839 57.113 11 0.000 0.124 0.946   
  Model summary
 
Parameter estimates
 
Equation R Square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.835 55.831 11 0.000 3.740 0.984   
Logarithmic 0.831 54.190 11 0.000 −19.200 13.914   
Inverse 0.826 52.263 11 0.000 31.553 −196.306   
Quadratic 0.841 26.530 10 0.000 26.506 −2.226 0.113  
Cubic 0.842 26.575 10 0.000 16.189 0.000 −0.047 0.004 
Compound 0.839 57.113 11 0.000 8.065 1.057   
Power 0.835 55.784 11 0.000 2.216 0.783   
0.831 54.115 11 0.000 3.653 −11.058   
Growth 0.839 57.113 11 0.000 2.088 0.055   
Exponential 0.839 57.113 11 0.000 8.065 0.055   
Logistic 0.839 57.113 11 0.000 0.124 0.946   
Table 6

Model summary and parameter estimates in 30 cm soil depth (Sanandaj station)

  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.774 37.716 11 0.000 2.798 1.002   
Logarithmic 0.770 36.764 11 0.000 −20.549 14.165   
Inverse 0.764 35.662 11 0.000 31.112 −199.755   
Quadratic 0.782 17.943 10 0.000 30.350 −2.883 0.137  
Cubic 0.782 17.964 10 0.000 16.881 0.000 −069 0.005 
Compound 0.780 38.979 11 0.000 7.407 1.060   
Power 0.776 38.158 11 0.000 1.890 0.828   
0.772 37.162 11 0.000 3.658 −11.690   
Growth 0.780 38.979 11 0.000 2.002 0.059   
Exponential 0.780 38.979 11 0.000 7.407 0.059   
Logistic 0.780 38.979 11 0.000 0.135 0.943   
  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.774 37.716 11 0.000 2.798 1.002   
Logarithmic 0.770 36.764 11 0.000 −20.549 14.165   
Inverse 0.764 35.662 11 0.000 31.112 −199.755   
Quadratic 0.782 17.943 10 0.000 30.350 −2.883 0.137  
Cubic 0.782 17.964 10 0.000 16.881 0.000 −069 0.005 
Compound 0.780 38.979 11 0.000 7.407 1.060   
Power 0.776 38.158 11 0.000 1.890 0.828   
0.772 37.162 11 0.000 3.658 −11.690   
Growth 0.780 38.979 11 0.000 2.002 0.059   
Exponential 0.780 38.979 11 0.000 7.407 0.059   
Logistic 0.780 38.979 11 0.000 0.135 0.943   
Table 7

Model summary and parameter estimates in 50 cm soil depth (Sanandaj station)

  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.776 38.104 11 0.000 3.921 0.964   
Logarithmic 0.766 35.965 11 0.000 −18.416 13.583   
Inverse 0.755 33.846 11 0.000 31.074 −190.841   
Quadratic 0.822 23.034 10 0.000 67.734 −8.033 0.317  
Cubic 0.822 23.123 10 0.000 29.920 0.000 −0.251 0.013 
Compound 0.782 39.391 11 0.000 8.156 1.056   
Power 0.772 37.284 11 0.000 2.322 0.764   
0.762 35.175 11 0.000 3.625 −10.734   
Growth 0.782 39.391 11 0.000 2.099 0.054   
Exponential 0.782 39.391 11 0.000 8.156 0.054   
Logistic 0.782 39.391 11 0.000 0.123 0.947   
  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.776 38.104 11 0.000 3.921 0.964   
Logarithmic 0.766 35.965 11 0.000 −18.416 13.583   
Inverse 0.755 33.846 11 0.000 31.074 −190.841   
Quadratic 0.822 23.034 10 0.000 67.734 −8.033 0.317  
Cubic 0.822 23.123 10 0.000 29.920 0.000 −0.251 0.013 
Compound 0.782 39.391 11 0.000 8.156 1.056   
Power 0.772 37.284 11 0.000 2.322 0.764   
0.762 35.175 11 0.000 3.625 −10.734   
Growth 0.782 39.391 11 0.000 2.099 0.054   
Exponential 0.782 39.391 11 0.000 8.156 0.054   
Logistic 0.782 39.391 11 0.000 0.123 0.947   
Table 8

Model summary and parameter estimates in 100 cm soil depth (Sanandaj station)

  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.805 45.412 11 0.000 3.564 0.990   
Logarithmic 0.798 43.352 11 0.000 −19.440 13.971   
Inverse 0.789 41.191 11 0.000 31.491 −196.683   
Quadratic 0.827 23.878 10 0.000 48.052 −5.283 0.221  
Cubic 0.827 23.900 10 0.000 23.142 0.000 −0.152 0.009 
Compound 0.810 47.002 11 0.000 7.970 1.057   
Power 0.804 45.037 11 0.000 2.175 0.788   
0.796 42.934 11 0.000 3.652 −11.105   
Growth 0.810 47.002 11 0.000 2.076 0.056   
Exponential 0.810 47.002 11 0.000 7.970 0.056   
Logistic 0.810 47.002 11 0.000 0.125 0.946   
  Model summary
 
Parameter estimates
 
Equation R square df1 df2 Sig. Constant b1 b2 b3 
Linear 0.805 45.412 11 0.000 3.564 0.990   
Logarithmic 0.798 43.352 11 0.000 −19.440 13.971   
Inverse 0.789 41.191 11 0.000 31.491 −196.683   
Quadratic 0.827 23.878 10 0.000 48.052 −5.283 0.221  
Cubic 0.827 23.900 10 0.000 23.142 0.000 −0.152 0.009 
Compound 0.810 47.002 11 0.000 7.970 1.057   
Power 0.804 45.037 11 0.000 2.175 0.788   
0.796 42.934 11 0.000 3.652 −11.105   
Growth 0.810 47.002 11 0.000 2.076 0.056   
Exponential 0.810 47.002 11 0.000 7.970 0.056   
Logistic 0.810 47.002 11 0.000 0.125 0.946   

Finally, through the above-mentioned techniques, the most effective technique was identified.

As evident in Tables 4>–8, the highest coefficient of determination appears to indicate that the variables were obtained by methods (cubic, quadratic, and S) that showed more similar behavior to the different depths of soil temperature values in Sanandaj.

Forecasting soil temperature based on the temperature of the soil temperatures

Based on the highest coefficient of determination in correlation (cubic, quadratic, and S) in the actual data (past), soil temperature (at depths of 5 and 10, 30, 50, and 100 cm) based on the output air temperature in the LARS-WG model was predicted. To evaluate changes in soil temperature, forecast data were separated into two decades (2016–2025) and (2026–2035).

Interpolation and data zoning

To identify the soil temperature variations in the area of study, the real and predicted data in two decades were interpolated by applying the Kriging method. This method facilitates the identification of borders with a relatively high accuracy (Figures 24).
Figure 2

Interpolation of the real data by applying the Kriging method (1990–2010).

Figure 2

Interpolation of the real data by applying the Kriging method (1990–2010).

Figure 3

Interpolation of the predicted data in the first decade by applying the Kriging method (2016–2025).

Figure 3

Interpolation of the predicted data in the first decade by applying the Kriging method (2016–2025).

Figure 4

Interpolation of the predicted data in the second decade by applying the Kriging method (2026–2035).

Figure 4

Interpolation of the predicted data in the second decade by applying the Kriging method (2026–2035).

RESULTS AND DISCUSSION

To interpret the results, we present them as follows.

Evaluation of the LARS-WG model

The LARS-WG model evaluation period comparing the data and the data generated by the model was used. 1990–2010 is the period for evaluating the capability model. Data monitoring and parameter modeling for the required statistical characteristics, such as average monthly SD relative errors, mean values, SDs, and correlation relative errors and other tests (Fisher's t-test) to evaluate the model outputs of LARS-WG were prepared.

The results of the modeled values and also those for the stations showed that the minimum and maximum samples are in full compliance with the maximum and minimum temperatures in Sanandaj station to a great extent with the data observed in the same period.

Reviewing and comparing soil temperature changes predicted in the first decade (2016–2025)

The research findings from this study prove that the weather in the next 10 years (2016–2025) will increase the length of this phenomenon by 1° of temperature input data (actual data), as the data for LARS are based on the forecast of Sanandaj station, which shows the lowest rate of increase (Table 9).

Table 9

Difference of first decade of prediction (2016–2025) with real data

Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.6 0.5 0.6 0.8 0.6 0.5 0.7 
Hamedan-Nojeh 0.6 0.7 0.9 1.8 0.8 0.6 0.6 0.6 
Sanandaj 0.9 0.4 0.5 0.4 0.4 0.4 0.4 0.6 
Kermanshah 0.5 0.8 0.4 0.3 0.5 0.5 0.5 0.7 
Saqez 1.8 2.0 2.0 1.7 2.2 1.7 2.0 2.9 
Kangavar 0.8 1.1 0.8 0.9 0.8 0.7 0.8 1.0 
Islamabad 0.6 0.8 0.4 0.3 0.5 0.5 0.5 0.7 
Mean 0.9 0.9 0.8 0.9 0.8 0.7 0.8 1.0 
Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.6 0.5 0.6 0.8 0.6 0.5 0.7 
Hamedan-Nojeh 0.6 0.7 0.9 1.8 0.8 0.6 0.6 0.6 
Sanandaj 0.9 0.4 0.5 0.4 0.4 0.4 0.4 0.6 
Kermanshah 0.5 0.8 0.4 0.3 0.5 0.5 0.5 0.7 
Saqez 1.8 2.0 2.0 1.7 2.2 1.7 2.0 2.9 
Kangavar 0.8 1.1 0.8 0.9 0.8 0.7 0.8 1.0 
Islamabad 0.6 0.8 0.4 0.3 0.5 0.5 0.5 0.7 
Mean 0.9 0.9 0.8 0.9 0.8 0.7 0.8 1.0 

The increasing temperature at different depths of the soil in each station demonstrated similar behaviors to the actual data in the deep soil depths, which have a greater rate of increase than in others. In the actual data, there are clear spatial differences in soil temperature at depths below the surface temperature of the soil in the western region.

Comparative investigation of soil temperature changes anticipated in the second decade (2026–2035)

Saqez station shows that the highest rate of increase arose from the depths of soil at a depth of 1 m or more. The southwest region, such as the east of Hamedan, shows a lower rate of increase in soil temperatures at a depth of more than 1 m (Table 10).

Table 10

Difference of second decade of prediction (2026–2035) with real data

Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.8 0.6 0.7 0.9 0.7 0.6 0.8 
Hamedan-Nojeh 1.0 0.8 1.0 1.7 1.1 0.9 0.8 0.9 
Sanandaj 1.0 0.8 0.7 0.7 0.7 0.7 0.7 0.9 
Kermanshah 0.8 1.1 0.7 0.7 0.7 0.8 0.8 0.9 
Saqez 2.5 2.7 2.4 2.2 2.6 2.6 3.3 2.6 
Kangavar 2.0 1.2 1.3 1.1 1.2 1.1 1.3 1.6 
Islamabad 0.9 1.0 0.7 0.7 0.7 0.8 0.8 0.9 
Mean 1.3 1.2 1.1 1.1 1.1 1.1 1.2 1.2 
Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.8 0.6 0.7 0.9 0.7 0.6 0.8 
Hamedan-Nojeh 1.0 0.8 1.0 1.7 1.1 0.9 0.8 0.9 
Sanandaj 1.0 0.8 0.7 0.7 0.7 0.7 0.7 0.9 
Kermanshah 0.8 1.1 0.7 0.7 0.7 0.8 0.8 0.9 
Saqez 2.5 2.7 2.4 2.2 2.6 2.6 3.3 2.6 
Kangavar 2.0 1.2 1.3 1.1 1.2 1.1 1.3 1.6 
Islamabad 0.9 1.0 0.7 0.7 0.7 0.8 0.8 0.9 
Mean 1.3 1.2 1.1 1.1 1.1 1.1 1.2 1.2 

Comparative investigation of soil temperature in two periods of prediction (2016–2025 and 2026–2035)

Based on the comparison of data from two decades of prediction and their difference in the second period, the increasing trend continues and it seems that warming continues to increase in these regions. At Kangavar station, this increase becomes more remarkable. The temperature increase is likely to display a different behavior in the second decade. The mean difference of air temperature in the two decades is 0.2, and at 50 and 100 cm depth, the increasing trend becomes steeper. Mean temperature increase at Saqez station, and an increase in temperature of 10 cm depth at Hamedan-Noje station, occurs in the lower slope in the first decade. In other regions and at different depths, there will be a significant increasing trend. Table 11 shows the differences of the two periods of prediction.

Table 11

Difference of first and second decade of prediction data

Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.0 0.2 0.1 0.2 0.1 0.1 0.1 0.1 
Hamedan-Nojeh 0.4 0.1 0.2 −0.1 0.4 0.3 0.3 0.3 
Sanandaj 0.1 0.4 0.3 0.3 0.3 0.3 0.3 0.3 
Kermanshah 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 
Saqez 0.7 0.7 0.4 0.5 0.4 0.9 1.3 −0.3 
Kangavar 1.2 0.0 0.5 0.2 0.5 0.4 0.5 0.6 
Islamabad 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 
Mean 0.4 0.3 0.3 0.2 0.3 0.4 0.4 0.2 
Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.0 0.2 0.1 0.2 0.1 0.1 0.1 0.1 
Hamedan-Nojeh 0.4 0.1 0.2 −0.1 0.4 0.3 0.3 0.3 
Sanandaj 0.1 0.4 0.3 0.3 0.3 0.3 0.3 0.3 
Kermanshah 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 
Saqez 0.7 0.7 0.4 0.5 0.4 0.9 1.3 −0.3 
Kangavar 1.2 0.0 0.5 0.2 0.5 0.4 0.5 0.6 
Islamabad 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 
Mean 0.4 0.3 0.3 0.2 0.3 0.4 0.4 0.2 

Comparative investigation of soil temperature changes throughout the forecast period (2016–2035)

The higher latitude, according to the forecast of Hamedan-Noje station, shows the lowest rate of increase in temperature. It also increased more in the deep soil depths than others. Table 12 shows difference of prediction with real data in the entire forecast period.

Table 12

Difference of prediction in the whole period of prediction (2016–2035) with real data

Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.7 0.5 0.7 0.8 0.6 0.6 0.8 
Hamedan-Nojeh 0.8 0.7 1.0 1.8 1.0 0.8 0.7 0.7 
Sanandaj 1.0 0.6 0.6 0.6 0.6 0.5 0.5 0.8 
Kermanshah 0.7 0.9 0.5 0.5 0.6 0.7 0.6 0.8 
Saqez 2.2 2.4 2.2 1.9 2.4 2.1 2.7 2.7 
Kangavar 1.4 1.1 1.0 1.0 1.0 0.9 1.0 1.3 
Islamabad 0.7 0.9 0.5 0.5 0.6 0.7 0.6 0.8 
Mean 1.1 1.0 0.9 1.0 1.0 0.9 1.0 1.1 
Name Max Min 5 cm 10 cm 30 cm 50 cm 100 cm Mean 
Hamadan-Airport 0.9 0.7 0.5 0.7 0.8 0.6 0.6 0.8 
Hamedan-Nojeh 0.8 0.7 1.0 1.8 1.0 0.8 0.7 0.7 
Sanandaj 1.0 0.6 0.6 0.6 0.6 0.5 0.5 0.8 
Kermanshah 0.7 0.9 0.5 0.5 0.6 0.7 0.6 0.8 
Saqez 2.2 2.4 2.2 1.9 2.4 2.1 2.7 2.7 
Kangavar 1.4 1.1 1.0 1.0 1.0 0.9 1.0 1.3 
Islamabad 0.7 0.9 0.5 0.5 0.6 0.7 0.6 0.8 
Mean 1.1 1.0 0.9 1.0 1.0 0.9 1.0 1.1 

CONCLUSION

The abundant production capabilities in the field of agriculture and the considerable potential climate changes could cause irreparable damage to the economy of the region. In this study, based on the actual data of air temperature and its relationship with soil temperature, the soil temperature in the long-term forecast of the region was analyzed at different depths. The data are fed into numerical models to predict the time evolution of weather. Based on the research findings for the next 20 years (2016–2035), an average temperature of 1/1 °C increase was determined. An increase in temperature, especially in winter, reduces the duration of cold and causes a lack of an estimate chilling requirement (cooling degree days) for the plants; their performance is reduced or terminated, and early winter cold causes early flowering, reduces crop yields, and mainly increases the risk of frost. Increased temperature and decreased snow precipitation cause the lack of proper nutrition of underground water in mountainous areas, and consequently, adverse effects on plant growth will emerge. A broader implication of this research study, is applying other models of future climate in the climate and in soil temperatures. The outcomes of this study contribute to the meteorological, ecological, agricultural, and geological divisions within a certain system boundary. The results of this study are important for geological research as well. This is because the soil temperature has a significant role in soil formation as a geological process. This will even facilitate agricultural departments in decision-making to select appropriate sites and which crops to cultivate. Soil temperature determines ecological diversity in soil, thus, it ultimately leads to saving time and money. Air temperature will be required as an input to get the output in the form of soil temperature. The long-term prediction of air temperature and soil temperature in different depths was paid attention to based on the real data of air temperature and its relationship with soil temperature. Air temperature correlates well with soil temperature because both are determined by the energy balance at the ground surface (Buringh 1984; Pritchett & Fisher 1987). Abbasi et al. (2011) reported the assessment of climate change on Zagros area (western Iran) for the time period of 2010–2039 using statistical downscaling of the ECHO-G model over 18 synoptic stations with emission scenario A1. The results showed that mean annual precipitation will be decreased by 2%, while increasing the mean annual temperature by 0.4 °C during the future studied time period. In addition, a study in South Khorasan Province in Iran for the period of 2010–2039 illustrated that climate change is likely to lead to modifications in climate variability and changes in average rainfall (increase of 4%), reducing the number of ice days and increasing the average annual temperature by about 0.3 °C (Abbasi et al. 2011). Yang et al. (2011), by projection of climate change for daily precipitation in a catchment in Taiwan, based on the outputs of GCMs as predictors and using statistical downscaling, revealed that the projected local precipitations under two emission scenarios tend to decrease. Regarding the point that precipitation and temperature data are the most regularly used forcing terms in hydrological models, the choice of the most appropriate downscaling technique is important. A number of researchers have reviewed different downscaling methods (Wilby & Wigley 1997; Semenov et al. 1998; Xu 1999; Qian et al. 2004; Wilby et al. 2004; Fiseha et al. 2012; King et al. 2012; Muluye 2012; Hu et al. 2013). The research of Semenov et al. (1998), in comparison to the downscaling methods, indicated that the LARS-WG generator used more complex distributions for weather variables and tended to match the observed data more closely than WGEN, although there are certain characteristics of the data that neither generator reproduced accurately. These data are entered into the numerical model to predict air temperature change. This process does not give proper responses for the behavior of air over the next few hours, but can be a general prospect of climate in the future, and its results can be used in land survey and regional planning. Based on the findings, there is a 1.1 °C increase in air temperature in 20 years (2016–2035), which reduces the cold period and cooling degree days of plants, leading to a reduction in performance. The early termination of winter cold brings with it early flowering, performance reduction of farming products, and freezing vulnerability. With the increase of soil temperature in the region, especially at higher latitude in the two decades, the agroclimate zoning changes and the crops that have long been planted in specific lands will become unproductive in their permanent habitat. In fact, tropical crops’ growth will be enhanced if the temperature increase is remarkable at 1 m deep, and in Saqez station, air and soil temperature increase occurs in a greater slope than in other stations. Temperature increase leads to the reduction of snowfall, resulting in a lack of feeding underground resources in mountainous regions and dry soil affecting the growth and function of plants. It is suggested that widespread studies are used and other climatic models are applied to predict the climate situation and soil temperature. As a general conclusion, it can be said that the LARS model can be used with high confidence to reproduce daily meteorological data and to study climate change in future periods. The study shows to what extent atmospheric temperature is correlated with soil temperature and, furthermore, the prediction of soil temperature from existing overlying air temperature through a scientific process within a system boundary. The main intention of this study is to substantiate the association between atmospheric temperature and soil temperature in a certain zone. There are many different types of statistical downscaling of climate model software available for climate impact studies. As a next step in the research, these approaches could be applied to the studied region to determine their value in simulating climate values. Many of these approaches are limited in their transferability however, due to their intellectual complexity or their need for expensive software as a platform.

REFERENCES

REFERENCES
Babaeian
I.
Kwon
W. T.
Im
E. S.
2004
Application of weather generator technique for climate change assessment over Korea. Korea Meteorological Research Institute, Climate Research Laboratory, 98 p.
Barrow
E. M.
Semenov
M. A.
1995
Climate change scenarios with high spatial and temporal resolution for agricultural applications
.
Forestry
68
,
349
360
.
Babaean
I.
Najafi Nik
Z.
Zabol Abbasi
F.
Habibi Nokhandan
M.
Adab
H.
Malbosi
Sh.
2010
Hydrophilic assessment of climate change in the period 2039-2010 with the downscaling of general circulation climate models ECHO-G
.
Geography and Development
19
,
135
152
(in Persian).
Buringh
P.
1984
Organic carbon in soils of the world
. In:
The Role of Terrestrial Vegetation in the Global Carbon Cycle: Measurement by Rremote Sensing. SCOPE 23
(
Woodwell
G. M.
, ed.).
John Wiley & Sons
,
Chichester
,
UK
, pp.
91
109
.
Chen
H.
Guo
J.
Zhang
C.
Xu
C.
2013
Prediction of temperature and precipitation in Sudan and South Sudan by using LARS-WG in future
.
Theoretical and Applied Climatology
113
,
363
375
.
Dubrovsky
M.
1996
The stochastic generator of daily weather series for the crop growth model
.
Meteorology
49
,
97
105
.
Fahim Ahmad
M.
Rasul
G.
2008
Prediction of soil temperature by air temperature; a case study for Faisalabad
.
Pakistan Journal of Meteorology
5
(
9
),
1
19
.
Fiseha
B. M.
Melesse
A. M.
Romano
E.
Volpi
E.
Fiori
A.
2012
Statistical downscaling of precipitation and temperature for the Upper Tiber Basin in Central Italy
.
Int. J. Water Sci.
1
(
3
),
1
14
.
Karimi
S. P.
Karimi
S. A.
Yavari
A. R.
Niksokhan
M. H.
2014
Prediction of temperature and precipitation in Damavand Catchment in Iran by using LARS–WG in future
.
Earth Sciences
4
(
3
),
95
100
.
Lawless
C.
Semenov
M. A.
2005
Assessing lead time for predicting wheat growth using a crop simulation model
.
Agricultural and Forest Meteorology
135
,
302
313
.
Meshkati
A. H.
Kord Jazi
M.
Babaean
I.
2010
Evaluation of LARS model to simulate weather data in Golestan province in the period 2007-1993
.
Geographical Sciences Applied Research
19
,
81
96
(in Persian).
Muluye
G.
2012
Comparison of statistical methods for downscaling daily precipitation
.
J. Hydro. Informs.
1006
1023
.
Muttil
N.
Chau
K.
2007
Machine learning paradigms for selecting ecologically significant input variables
.
Engineering Applications of Artificial Intelligence
20
(
6
),
735
744
.
Nakicenovic
N.
Swart
R.
2000
Emissions scenarios. Special Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
.
Pritchett
W. L.
Fisher
R. F.
1987
Properties and Management of Forest Soil
.
John Wiley & Sons
,
New York
,
USA
.
Qian
B.
Gameda
S.
Hayhoe
H.
De Jong
R.
Bootsma
A.
2004
Comparison of LARS-WG and AAFC-WG stochastic weather generators for diverse Canadian climates
.
Climate Research
26
,
175
191
.
Semenov
M. A.
2007
Developing of high-resolution UKCUP02- based climate change scenarios in the UK
.
Journal of Agricultural and Forest Meteorology
144
,
127
138
.
Semenov
M. A.
Barrow
E. M.
2002
Lars a Stochastic Weather Generator for use in Climate Impact Studies
.
User's manual, Version 3.0
, pp.
1
30
.
Semenov
M. A.
Doblas
R. F. J.
2007
Utility of dynamical seasonal forecasts in predicting crop yield
.
Climate Research
34
,
71
81
.
Semenov
M. A.
Brooks
R. J.
Barrow
E. M.
Richardson
C. W.
1998
Comparison of the WGEN and LARS-WG stochastic weather generators for diverse climates
.
Climate Research
10
,
95
107
.
Wang
Y.
Sheng
D.
Wang
D.
Ma
H.
Wu
J.
Xu
F.
2014
Variable fuzzy set theory to assess water quality of the Meiliang Bay in Taihu Lake Basin
.
Water Resources Management
28
,
867
880
.
Wilby
R. L.
Wigley
T. M. L.
1997
Downscaling general circulation model output: a review of methods and limitations
.
Prog. Phys. Geogr.
21
,
530
548
.
Wilby
R. L.
Charles
S. P.
Zorita
E.
Timbal
B.
Whetton
P.
Mearns
L. O.
2004
Guidelines for use of climate scenarios developed from statistical downscaling methods. Supporting material of the Intergovernmental Panel on Climate Change. Available from the DDC of IPCC TGCIA. 27 p
.
Xie
J. X.
Cheng
C. T.
2006
A hybrid adaptive time-delay neural network model for multi-step-ahead prediction of sunspot activity
.
International Journal of Environment and Pollution
28
(
3/4
),
364
381
.
Yang
T. Ch.
Yu
P. Sh.
Wei
Ch. M.
Chen
Sh. T.
2011
Projection of climate change for daily precipitation: a case study in Shih-Men reservoir catchment in Taiwan
.
Hydrological Processes
25
(
8
),
1342
1354
.
Zhang
X. C.
Garbrecht
J. D.
2003
Evaluation of CLIGEN precipitation parameters and their implication on WEPP runoff and erosion prediction
.
Trans. ASAE
46
,
311
320
.
Zhao
C.
Tie
X.
Brasseur
G.
Noone
K. J.
Nakajima
T.
Zhang
Q.
Zhang
R.
Huang
M.
Duan
Y.
Li
G.
Ishizaka
Y.
2006
Aircraft measurements of cloud droplet spectral dispersion and implications for indirect aerosol radiative forcing
.
Geophysical Research Letters
33
,
L16809
.