Comparison of artificial intelligence techniques for prediction of soil temperatures in Turkey

Soil temperature is a meteorological data directly affecting the formation and development of plants of all kinds. Soil temperatures are usually estimated with various models including the artificial neural networks (ANNs), adaptive neuro-fuzzy inference system (ANFIS), and multiple linear regression (MLR) models. Soil temperatures along with other climate data are recorded by the Turkish State Meteorological Service (MGM) at specific locations all over Turkey. Soil temperatures are commonly measured at 5-, 10-, 20-, 50-, and 100-cm depths below the soil surface. In this study, the soil temperature data in monthly units measured at 261 stations in Turkey having records of at least 20 years were used to develop relevant models. Different input combinations were tested in the ANN and ANFIS models to estimate soil temperatures, and the best combination of significant explanatory variables turns out to be monthly minimum and maximum air temperatures, calendar month number, depth of soil, and monthly precipitation. Next, three standard error terms (mean absolute error (MAE, °C), root mean squared error (RMSE, °C), and determination coefficient (R ²)) were employed to check the reliability of the test data results obtained through the ANN, ANFIS, and MLR models. ANFIS (RMSE 1.99; MAE 1.09; R ² 0.98) is found to outperform both ANN and MLR (RMSE 5.80, 8.89; MAE 1.89, 2.36; R ² 0.93, 0.91) in estimating soil temperature in Turkey.

     

Evaluation and estimation of daily global solar radiation from the estimated direct and diffuse solar radiation

Active layer plays a key role in regulating the dynamics of hydrothermal processes and ecosystems that are sensitive to the changing climate in permafrost regions. However, little is known about the hydrothermal dynamics during freeze-thaw processes in permafrost regions with different vegetation types on the Qinghai-Tibetan Plateau (QTP). In the present study, the freezing and thawing processes at four sites (QT01, 03, 04, and 05) with different vegetation types on the QTP was analyzed. The results indicated that the impact on the soil water and heat during the summer thawing process was markedly greater than that during the autumn freezing process. Furthermore, the thermal-orbit regression slopes for all sites exhibited a homologous variation as the depth increased, with the slowest attenuation for the meadow sites (QT01 and QT03) and a slightly faster attenuation for the desert steppe site (QT05). The air and ground surface temperatures were similar in winter, but the ground surface temperature was significantly higher than the air temperature in summer in the radiation-rich environment at all sites on the QTP. The results also indicated that the n-factors were between 0.36 and 0.55 during the thawing season, and the annual mean temperature near the permafrost table was between − 1.26 and − 1.84 °C. In the alpine desert steppe region, the thermal conditions exhibited to show a warming trend, with a current permafrost table temperature of − 0.22 °C. The annual changing amplitude of the ground temperature at the permafrost table was different for different vegetation types.     

Temperature and thermal growing season variations along elevational gradients on a sub-alpine,temperate China

Fully and accurately studying temperature variations in montane areas are conducive to a better understanding of climate modeling and climate-growth relationships on regional scales. To explore the spatio-temporal changes in air and soil temperatures and their relationship in montane areas, on-site monitoring over 2 years (2015 and 2016) was conducted at nine different elevations from 2040 to 2740 m a.s.l. on Luya Mountain in the semiarid region of China. The results showed that the annual mean of air temperature lapse rate (ATLR) was 0.67 °C/100 m. ATLR varied obviously in different months within a range of 0.57~0.79 °C/100 m. The annual mean of the soil temperature lapse rate (STLR) was 0.48 °C/100 m. Seasonally, monthly mean soil temperature did not show a consistent pattern with regard to elevation. The relationships between air and soil temperatures showed piecewise changes. Soil was decoupled from the air temperature in cold winter and early spring. The parameters of the growing season based on the two temperature types had no corresponding relations, and seasonal mean of soil temperature showed the smallest value at mid-elevation rather than in the treeline ecotone. Based on these changes, our results emphasized that altitudinal and seasonal variability caused by local factors (such as snow cover and soil moisture) should be taken into full consideration in microclimate extrapolation and treeline prediction in montane areas, especially in relation to soil temperature.     

Hourly air temperature driven using multi-layer perceptron and radial basis function networks in arid and semi-arid regions

This study employed two artificial neural network (ANN) models, including multi-layer perceptron (MLP) and radial basis function (RBF), as data-driven methods of hourly air temperature at three meteorological stations in Fars province, Iran. MLP was optimized using the Levenberg–Marquardt (MLP_LM) training algorithm with a tangent sigmoid transfer function. Both time series (TS) and randomized (RZ) data were used for training and testing of ANNs. Daily maximum and minimum air temperatures (MM) and antecedent daily maximum and minimum air temperatures (AMM) constituted the input for ANNs. The ANN models were evaluated using the root mean square error (RMSE), the coefficient of determination (R ²) and the mean absolute error. The use of AMM led to a more accurate estimation of hourly temperature compared with the use of MM. The MLP-ANN seemed to have a higher estimation efficiency than the RBF ANN. Furthermore, the ANN testing using randomized data showed more accurate estimation. The RMSE values for MLP with RZ data using daily maximum and minimum air temperatures for testing phase were equal to 1.2°C, 1.8°C, and 1.7°C, respectively, at Arsanjan, Bajgah, and Kooshkak stations. The results of this study showed that hourly air temperature driven using ANNs (proposed models) had less error than the empirical equation.     

Spatial and temporal variations in soil temperatures over the Qinghai–Tibet Plateau from 1980 to 2017 based on reanalysis products

A change in soil temperature (ST) is a significant indicator of climate change, so understanding the variations in ST is required for studying the changes of the Qinghai–Tibet Plateau (QTP) permafrost. We investigated the performance of three reanalysis ST products at three soil depths (0–10 cm, 10–40 cm, and 40–100 cm) on the permafrost regions of the QTP: the European Centre for Medium-Range Weather Forecasts interim reanalysis (ERA-Interim), the second version of the National Centers for Environmental Prediction Climate Forecast System (CFSv2), and the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). Our results indicate that all three reanalysis ST products underestimate observations with negative mean bias error values at all three soil layers. The MERRA-2 product performed best in the first and second soil layers, and the ERA-Interim product performed best in the third soil layer. The spatiotemporal changes of annual and seasonal STs on the QTP from 1980 to 2017 were investigated using Sen’s slope estimator and the Mann–Kendall test. There was an increasing trend of ST in the deeper soil layer, which was less than that of the shallow soil layers in the spring and summer as well as annually. In contrast, the first-layer ST warming rate was significantly lower than that of the deeper soil layers in the autumn and winter. The significantly (P < 0.01) increasing trend of the annual ST indicates that the QTP has experienced climate warming during the past 38 years, which is one of the factors promoting permafrost degradation of the QTP.     

A permafrost warming in a cooling Antarctica?

The magnitude and even direction of recent Antarctic climate change is still debated because the paucity of long and complete instrumental data records. While along Antarctic Peninsula a strong warming coupled with large retreat of glaciers occurred, in continental Antarctica a cooling was recently detected. Here, the first existing permafrost data set longer than 10 years recorded in continental Antarctica is presented. Since 1997 summer ground surface temperature showed a strong warming trend (0.31°C per year) although the air temperature was almost stable. The summer ground surface temperature increase seemed to be influenced mainly by the increase of the total summer radiation as confirmed also by the increase of the summer thawing degree days. In the same period the active layer exhibited a thickening trend (1 cm per year) comparable with the thickening rates observed in several Arctic locations where air warming occurred. At all the investigated depths permafrost exhibited an increase of mean annual temperature of approximately 0.1°C per year. The dichotomy between active layer thickness and air temperature trends can produce large unexepected and unmodelled impacts on ecosystems and CO₂ balance.     

Climate changes and its impact on tundra ecosystem in Qinghai-Tibet Plateau,China

Alpine ecosystems in permafrost region are extremely sensitive to climate change. The headwater regions of Yangtze River and Yellow River of the Qinghai-Tibet plateau permafrost area were selected. Spatial-temporal shifts in the extent and distribution of tundra ecosystems were investigated for the period 1967–2000 by landscape ecological method and aerial photographs for 1967, and satellite remote sensing data (the Landsat’s TM) for 1986 and 2000. The relationships were analyzed between climate change and the distribution area variation of tundra ecosystems and between the permafrost change and tundra ecosystems. The responding model of tundra ecosystem to the combined effects of climate and permafrost changes was established by using statistic regression method, and the contribution of climate changes and permafrost variation to the degradation of tundra ecosystems was estimated. The regional climate exhibited a tendency towards significant warming and desiccation with the air temperature increased by 0.4–0.67°C/10a and relative stable precipitation over the last 45 years. Owing to the climate continuous warming, the intensity of surface heat source (HI) increased at the average of 0.45 W/m² per year, the difference of surface soil temperature and air temperature (DT) increased at the range of 4.1°C–4.5°C, and the 20-cm depth soil temperature within the active layer increased at the range of 1.1°C–1.4°C. The alpine meadow and alpine swamp meadow were more sensitive to permafrost changes than alpine steppe. The area of alpine swamp meadow decreased by 13.6–28.9%, while the alpine meadow area decreased by 13.5–21.3% from 1967 to 2000. The contributions of climate change to the degradation of the alpine meadow and alpine swamp was 58–68% and 59–65% between 1967 and 2000. The synergic effects of climate change and permafrost variation were the major drivers for the observed degradation in tundra ecosystems of the Qinghai-Tibet plateau.     

Soil temperature modeling at different depths using neuro-fuzzy,neural network,and genetic programming techniques

The applicability of artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS), and genetic programming (GP) techniques in estimating soil temperatures (ST) at different depths is investigated in this study. Weather data from two stations, Mersin and Adana, Turkey, were used as inputs to the applied models in order to model monthly STs. The first part of the study focused on comparison of ANN, ANFIS, and GP models in modeling ST of two stations at the depths of 10, 50, and 100 cm. GP was found to perform better than the ANN and ANFIS-SC in estimating monthly ST. The effect of periodicity (month of the year) on models’ accuracy was also investigated. Including periodicity component in models’ inputs considerably increased their accuracies. The root mean square error (RMSE) of ANN models was respectively decreased by 34 and 27 % for the depths of 10 and 100 cm adding the periodicity input. In the second part of the study, the accuracies of the ANN, ANFIS, and GP models were compared in estimating ST of Mersin Station using the climatic data of Adana Station. The ANN models generally performed better than the ANFIS-SC and GP in modeling ST of Mersin Station without local climatic inputs.     

An Amplified Signal of Climatic Change in Soil Temperatures during the Last Century at Irkutsk, Russia

Climatic changes at the Earth's surface propagate slowly downward into theground and modify the ambient ground thermal regime. However, causes of soiltemperature changes in the upper few meters are not well documented. One majorobstacle to understanding the linkage between the soil thermal regime andclimatic change is the lack of long-term observations of soil temperatures andrelated climatic variables. Such measurements were made throughout the formerSoviet Union with some records beginning at the end of the 19th century. Inthis paper, we use records from Irkutsk, Russia, to demonstrate how the soiltemperature responded to climatic changes over the last century. Both airtemperature and precipitation at Irkutsk increased from the late 1890s to the1990s. Changes in air temperature mainly occurred in winter, while changes inprecipitation happened mainly during summer. There was an anti-correlationbetween mean annual air temperature and annual total precipitation, i.e., more(less) precipitation during cold (warm) years. There were no significanttrends of changes in the first day of snow on the ground in autumn, but snowsteadily disappeared earlier in spring, resulting in a reduction of the snowcover duration. A grass-covered soil experiences seasonal freezing for morethan nine months each year and the long-term average maximum depth ofseasonally frozen soils was about 177 cm with a range from 91 cm to 260 cm.The relatively lower soil temperature at shallow depths appears to representthe so-called `thermal offset' in seasonally frozen soils. Changes in meanannual air temperature and soil temperature at 40 cm depth were about the samemagnitude (2.0 °C to 2.5 °C) over the common period of record, but thepatterns of change were substantially different. Mean annual air temperatureincreased slightly until the 1960s, while mean annual soil temperatureincreased steadily throughout the entire period. This leads to the conclusionthat changes in air temperature alone cannot explain the changes in soiltemperatures at this station. Soil temperature actually decreased duringsummer months by up to 4 °C, while air temperature increased slightly.This cooling in the soil may be explained by changes in rainfall and hencesoil moisture during summer due to the effect of a soil moisture feedbackmechanism. While air temperature increased about 4 °C to 6 °C duringwinter, soil temperature increased by up to 9 °C. An increase in snowfallduring early winter (October and November) and early snowmelt in spring mayplay a major role in the increase of soil temperatures through the effects ofinsulation and albedo changes. Due to its relatively higher thermalconductivity compared to unfrozen soils, seasonally frozen ground may enhancethe soil cooling, especially in autumn and winter when thermal gradient isnegative.     

Comparison of artificial neural network and multivariate linear regression methods for estimation of daily soil temperature in an arid region

Soil temperature (T _S) strongly influences a wide range of biotic and abiotic processes. As an alternative to direct measurement, indirect determination of T _S from meteorological parameters has been the focus of attention of environmental researchers. The main purpose of this study was to estimate daily T _S at six depths (5, 10, 20, 30, 50 and 100?cm) by using a multilayer perceptron (MLP) artificial neural network (ANN) model and a multivariate linear regression (MLR) method in an arid region of Iran. Mean daily meteorological parameters including air temperature (T _a), solar radiation (R _S), relative humidity (RH) and precipitation (P) were used as input data to the ANN and MLR models. The model results of the MLR model were compared to those of ANN. The accuracy of the predictions was evaluated by the correlation coefficient (r), the root mean-square error (RMSE) and the mean absolute error (MAE) between the measured and predicted T _S values. The results showed that the ANN method forecasts were superior to the corresponding values obtained by the MLR model. The regression analysis indicated that T _a, RH, R _S and P were reasonably correlated with T _S at various depths, but the most effective parameters influencing T _S at different depths were T _a and RH.     

Comprehensive modeling of monthly mean soil temperature using multivariate adaptive regression splines and support vector machine

Soil temperature (T _s) and its thermal regime are the most important factors in plant growth, biological activities, and water movement in soil. Due to scarcity of the T _s data, estimation of soil temperature is an important issue in different fields of sciences. The main objective of the present study is to investigate the accuracy of multivariate adaptive regression splines (MARS) and support vector machine (SVM) methods for estimating the T _s. For this aim, the monthly mean data of the T _s (at depths of 5, 10, 50, and 100 cm) and meteorological parameters of 30 synoptic stations in Iran were utilized. To develop the MARS and SVM models, various combinations of minimum, maximum, and mean air temperatures (T _min, T _max, T); actual and maximum possible sunshine duration; sunshine duration ratio (n, N, n/N); actual, net, and extraterrestrial solar radiation data (R _s, R _n, R _a); precipitation (P); relative humidity (RH); wind speed at 2 m height (u ₂); and water vapor pressure (V_p) were used as input variables. Three error statistics including root-mean-square-error (RMSE), mean absolute error (MAE), and determination coefficient (R ²) were used to check the performance of MARS and SVM models. The results indicated that the MARS was superior to the SVM at different depths. In the test and validation phases, the most accurate estimations for the MARS were obtained at the depth of 10 cm for T _max, T _min, T inputs (RMSE = 0.71 °C, MAE = 0.54 °C, and R ² = 0.995) and for RH, V _p, P, and u ₂ inputs (RMSE = 0.80 °C, MAE = 0.61 °C, and R ² = 0.996), respectively.     

Permafrost Thaw Accelerates in Boreal Peatlands During Late-20th Century Climate Warming

Permafrost covers 25% of the land surface in the northern hemisphere, where mean annual ground temperature is less than 0°C. A 1.4–5.8 °C warming by 2100 will likely change the sign of mean annual air and ground temperatures over much of the zones of sporadic and discontinuous permafrost in the northern hemisphere, causing widespread permafrost thaw. In this study, I examined rates of discontinuous permafrost thaw in the boreal peatlands of northern Manitoba, Canada, using a combination of tree-ring analyses to document thaw rates from 1941–1991 and direct measurements of permanent benchmarks established in 1995 and resurveyed in 2002. I used instrumented records of mean annual and seasonal air temperatures, mean winter snow depth, and duration of continuous snow pack from climate stations across northern Manitoba to analyze temporal and spatial trends in these variables and their potential impacts on thaw. Permafrost thaw in central Canadian peatlands has accelerated significantly since 1950, concurrent with a significant, late-20th-century average climate warming of +1.32 °C in this region. There were strong seasonal differences in warming in northern Manitoba, with highest rates of warming during winter (+1.39 °C to +1.66 °C) and spring (+0.56 °C to +0.78 °C) at southern climate stations where permafrost thaw was most rapid. Projecting current warming trends to year 2100, I show that trends for north-central Canada are in good agreement with general circulation models, which suggest a 4–8 °C warming at high latitudes. This magnitude of warming will begin to eliminate most of the present range of sporadic and discontinuous permafrost in central Canada by 2100.     

Prediction of soil temperature using regression and artificial neural network models

In this study, monthly soil temperature was modeled by linear regression (LR), nonlinear regression (NLR) and artificial neural network (ANN) methods. The soil temperature and other meteorological parameters, which have been taken from Adana meteorological station, were observed between the years of 2000 and 2007 by the Turkish State Meteorological Service (TSMS). The soil temperatures were measured at depths of 5, 10, 20, 50 and 100 cm below the ground level. A three-layer feed-forward ANN structure was constructed and a back-propagation algorithm was used for the training of ANNs. In order to get a successful simulation, the correlation coefficients between all of the meteorological variables (soil temperature, atmospheric temperature, atmospheric pressure, relative humidity, wind speed, rainfall, global solar radiation and sunshine duration) were calculated taking them two by two. First, all independent variables were split into two time periods such as cold and warm seasons. They were added to the enter regression model. Then, the method of stepwise multiple regression was applied for the selection of the “best” regression equation (model). Thus, the best independent variables were selected for the LR and NLR models and they were also used in the input layer of the ANN method. Results of these methods were compared to each other. Finally, the ANN method was found to provide better performance than the LR and NLR methods.     

EnKF中误差协方差优化方法及在资料同化中应用

集合卡尔曼滤波 (the Ensemble Kalman Filter,简称EnKF) 中将预报集合的统计协方差作为预报误差协方差,但该估计可能严重偏离真实的预报误差协方差,影响同化精度。基于极大似然估计理论,发展了一种优化预报误差协方差矩阵的实时膨胀方法,即MLE (the Maximum Likelihood Estimation) 方法。利用蒙古国基准站Delgertsgot (简称DGS站) 观测资料,基于EnKF方法和MLE方法,在通用陆面模式 (the Common Land Model,简称CoLM) 中同化了地表温度和10 cm土壤温度观测资料,建立了土壤温度同化系统。结果表明:MLE方法对地表温度和各层土壤温度 (尤其深层土壤温度) 的估计比EnKF方法准确。考虑到浅层和深层土壤温度的差别,在实施MLE方法时对浅层和深层土壤温度采用了不同的膨胀因子。对比膨胀因子为单一标量时的结果,多因子膨胀能缓解深层土壤温度的不合理膨胀,改善同化效果。     

Local indications of climate changes in Turkey: Bursa as a case example

This study aims to put out on what ratio Bursa province, one of the important heavy industry regions of Turkey, has been affected climatic process called “Global Warming” or “Climate Change”. For this intend climatic measurement results from Bursa center, top of Uludağ Mount, Yenişehir and Keles meteorological stations were used. These measurements were taken as minimum temperature at night-time, maximum temperature at day-time, and mean temperature, mean pressure, insolation intensity, insolation duration, mean wind speed, minimum temperature above soil, soil temperatures at depths of 5, 10, and 20 cm rainfall. Overall, our statistical results showed that there was a considerable warming at statistically 1% and 5% levels in summer months, particularly in July Almost all performed measurements confirm this result. According to climatic data for thirty years (1975–2005), in the last twelve years contrary to previous 18 years, mean temperature values were higher than long-term mean value nine times (years) repetitively. Temperatures did not deviated higher than 0.5°C in six of these. At the temperatures below mean, The maximum deviation was −0.4°C.     

Climate change impact on snow and soil temperature in boreal Scots pine stands

Scenarios indicate that the air temperature will increase in high latitude regions in coming decades, causing the snow covered period to shorten, the growing season to lengthen and soil temperatures to change during the winter, spring and early summer. To evaluate how a warmer climate is likely to alter the snow cover and soil temperature in Scots pine stands of varying ages in northern Sweden, climate scenarios from the Swedish regional climate modelling programme SWECLIM were used to drive a Soil-Vegetation-Atmosphere Transfer (SVAT)-model (COUP). Using the two CO₂ emission scenarios A and B in the Hadley centres global climate model, HadleyA and HadleyB, SWECLIM predicts that the annual mean air temperature and precipitation will increase at most 4.8°C and 315 mm, respectively, within a century in the study region. The results of this analysis indicate that a warmer climate will shorten the period of persistent snow pack by 73–93 days, increase the average soil temperature by 0.9–1.5°C at 10 cm depth, advance soil warming by 15–19 days in spring and cause more soil freeze–thaw cycles by 31–38%. The results also predict that the large current variations in snow cover due to variations in tree interception and topography will be enhanced in the coming century, resulting in increased spatial variability in soil temperatures.     

Incorporating organic soil into a global climate model

Organic matter significantly alters a soil’s thermal and hydraulic properties but is not typically included in land-surface schemes used in global climate models. This omission has consequences for ground thermal and moisture regimes, particularly in the high-latitudes where soil carbon content is generally high. Global soil carbon data is used to build a geographically distributed, profiled soil carbon density dataset for the Community Land Model (CLM). CLM parameterizations for soil thermal and hydraulic properties are modified to accommodate both mineral and organic soil matter. Offline simulations including organic soil are characterized by cooler annual mean soil temperatures (up to ∼2.5°C cooler for regions of high soil carbon content). Cooling is strong in summer due to modulation of early and mid-summer soil heat flux. Winter temperatures are slightly warmer as organic soils do not cool as efficiently during fall and winter. High porosity and hydraulic conductivity of organic soil leads to a wetter soil column but with comparatively low surface layer saturation levels and correspondingly low soil evaporation. When CLM is coupled to the Community Atmosphere Model, the reduced latent heat flux drives deeper boundary layers, associated reductions in low cloud fraction, and warmer summer air temperatures in the Arctic. Lastly, the insulative properties of organic soil reduce interannual soil temperature variability, but only marginally. This result suggests that, although the mean soil temperature cooling will delay the simulated date at which frozen soil begins to thaw, organic matter may provide only limited insulation from surface warming.     

Bulgarian mountains air temperatures and precipitation—statistical downscaling of global climate models and some projections

Measured air temperature and precipitation data from three high mountainous Bulgarian stations were used along with data from 18 global climate models (GCMs). Air temperature and precipitation outputs of preindustrial control experiment were compared with actually observed values. GCM with the best overall performance is BCCR BCM 2.0 for air temperatures (period 1941?C2009) and CGCM 3.1/T47 for precipitation (period 1947?C2009). Statistical methods were used in this research??nonparametric Spearman correlation, Mann?CWhitney test, multiple linear regression, etc. Projections were made for the following future decades: 2015?C2024, 2045?C2054 and 2075?C2084. The best months, described by multiple linear regression (MLR) model of air temperatures, are November, January, March, and May. The worst described are summer months. There is not any pattern in the relationship between constructed MLR models and measured precipitation. Models that perform the best in different months at the three investigated stations are MIUB ECHO-G, GISS AOM, CGCM 3.1/T63, and CNRM CM3 for air temperatures and GFDL CM 2.1, GISS AOM, and MIUB ECHO-G for precipitation. The fit between statistical models' outputs and values observed at stations is different, better in cold part of the year. There will be mixed future changes of air temperatures at all the three high mountainous stations. An increase of temperatures is expected in April, November, and December. A decrease will happen in February, July, and October. Mean annual temperatures are expected to rise by 0.1?°C (Botev) to 0.2?°C (Musala and Cherni vrah) in the decade 2075?C2084, but mean annual temperatures at the end of the period with measurements (2009) has already exceeded by far projected values. Trends in precipitation are mixed both in spatial and in temporal directions. Observed decrease of precipitation, especially in the warm half of the year, is not described well in MLR models. The same is valid for annual amounts, which are projected to be higher than those measured in the end of instrumental period (2009). This is opposite to observed trends in recent decades, especially at stations Cherni vrah and Botev, where a significant decrease of precipitation amounts has happened.     

Soil thermal dynamics of terrestrial ecosystems of the conterminous United States from 1948 to 2008: an analysis with a process-based soil physical model and AmeriFlux data

The spatiotemporal distribution characteristics of soil temperature are a significant, but seldom described signal of climate warming. This study examines the spatiotemporal trends in soil temperature at depths of 10, 20, and 50 cm in the conterminous US during 1948–2008. We find a warming trend of between 0.2 and 0.4 °C at all depths from 1948 to 2008. The lowest soil temperatures are in Colorado and the area where Wyoming, Idaho, and Montana meet. The coastal areas, such as Texas, Florida, and California, experienced the highest soil temperature. In addition, areas that experienced weak cooling in summer soil temperature include Texas, Oklahoma, and Arkansas. Warming was recorded in Arizona, Nevada, and Oregon. In winter, Mississippi, Alabama, and Georgia show a cooling trend, and Montana, North Dakota, and South Dakota have been warming over the 61-year period. Additionally, mix-forest areas experience slightly cooler soil temperature in comparison with air temperature. Shrubland areas experience slightly warmer soil temperature in comparison with air temperature. This study is among the first to analyze the spatiotemporal distribution characteristics of soil temperature in the conterminous US by using multiple site observational data. Improved understanding of the spatially complex responses of soil temperature shall have significant implications for future studies in climate change over the region.     

Spatiotemporal modeling of monthly soil temperature using artificial neural networks

Soil temperature data are critical for understanding land–atmosphere interactions. However, in many cases, they are limited at both spatial and temporal scales. In the current study, an attempt was made to predict monthly mean soil temperature at a depth of 10 cm using artificial neural networks (ANNs) over a large region with complex terrain. Gridded independent variables, including latitude, longitude, elevation, topographic wetness index, and normalized difference vegetation index, were derived from a digital elevation model and remote sensing images with a resolution of 1 km. The good performance and robustness of the proposed ANNs were demonstrated by comparisons with multiple linear regressions. On average, the developed ANNs presented a relative improvement of about 44 % in root mean square error, 70 % in mean absolute percentage error, and 18 % in coefficient of determination over classical linear models. The proposed ANN models were then applied to predict soil temperatures at unsampled locations across the study area. Spatiotemporal variability of soil temperature was investigated based on the obtained database. Future work will be needed to test the applicability of ANNs for estimating soil temperature at finer scales.