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.     

Data selection using support vector regression

Geophysical data sets are growing at an ever-increasing rate, requiring computationally efficient data selection(thinning)methods to preserve essential information. Satellites, such as Wind Sat, provide large data sets for assessing the accuracy and computational efficiency of data selection techniques. A new data thinning technique, based on support vector regression(SVR), is developed and tested. To manage large on-line satellite data streams, observations from Wind Sat are formed into subsets by Voronoi tessellation and then each is thinned by SVR(TSVR). Three experiments are performed. The first confirms the viability of TSVR for a relatively small sample, comparing it to several commonly used data thinning methods(random selection, averaging and Barnes filtering), producing a 10% thinning rate(90% data reduction), low mean absolute errors(MAE) and large correlations with the original data. A second experiment, using a larger dataset, shows TSVR retrievals with MAE < 1 m s-1and correlations 0.98. TSVR was an order of magnitude faster than the commonly used thinning methods. A third experiment applies a two-stage pipeline to TSVR, to accommodate online data. The pipeline subsets reconstruct the wind field with the same accuracy as the second experiment, is an order of magnitude faster than the nonpipeline TSVR. Therefore, pipeline TSVR is two orders of magnitude faster than commonly used thinning methods that ingest the entire data set. This study demonstrates that TSVR pipeline thinning is an accurate and computationally efficient alternative to commonly used data selection techniques.     

Application of random forest time series,support vector regression and multivariate adaptive regression splines models in prediction of snowfall (a case study of Alvand in the middle Zagros,Iran)

We have conducted a case study to investigate the performance of support vector machine, multivariate adaptive regression splines, and random forest time series methods in snowfall modeling. These models were applied to a data set of monthly snowfall collected during six cold months at Hamadan Airport sample station located in the Zagros Mountain Range in Iran. We considered monthly data of snowfall from 1981 to 2008 during the period from October/November to April/May as the training set and the data from 2009 to 2015 as the testing set. The root mean square errors (RMSE), mean absolute errors (MAE), determination coefficient (R ²), coefficient of efficiency (E%), and intra-class correlation coefficient (ICC) statistics were used as evaluation criteria. Our results indicated that the random forest time series model outperformed the support vector machine and multivariate adaptive regression splines models in predicting monthly snowfall in terms of several criteria. The RMSE, MAE, R ², E, and ICC for the testing set were 7.84, 5.52, 0.92, 0.89, and 0.93, respectively. The overall results indicated that the random forest time series model could be successfully used to estimate monthly snowfall values. Moreover, the support vector machine model showed substantial performance as well, suggesting it may also be applied to forecast snowfall in this area.     

Daily soil temperature modeling using neuro-fuzzy approach

Soil temperature is an important meteorological parameter which influences a number of processes in agriculture, hydrology, and environment. However, soil temperature records are not routinely available from meteorological stations. This work aimed to estimate daily soil temperature using the coactive neuro-fuzzy inference system (CANFIS) in arid and semiarid regions. For this purpose, daily soil temperatures were recorded at six depths of 5, 10, 20, 30, 50, and 100 cm below the surface at two synoptic stations in Iran. According to correlation analysis, mean, maximum, and minimum air temperatures, relative humidity, sunshine hours, and solar radiation were selected as the inputs of the CANFIS models. It was concluded that, in most cases, the best soil temperature estimates with a CANFIS model can be provided with the Takagi–Sugeno–Kang (TSK) fuzzy model and the Gaussian membership function. Comparison of the models’ performances at arid and semiarid locations showed that the CANFIS models’ performances in arid site were slightly better than those in semiarid site. Overall, the obtained results indicated the capabilities of the CANFIS model in estimating soil temperature in arid and semiarid regions.     

Climatic noise and potential predictability of monthly mean temperature over China

Summary In this paper, based on the data at 162 stations selected over China from 1960 to 1991 the climatic noise and potential predictability of monthly mean temperature have been studied. The method of estimating climatic noise is based on the idea of Yamamoto et al. (1985) and the potential predictability is expressed by the ratio of the estimated inter-annual variation to the estimated natural variation (or climatic noise). Generally the climatic noise of monthly mean temperature increases with latitude and altitude and varies with season. The continental air from Siberia and Mongolia plays a significant role and the ocean acts as an adjustor and a reductor in the climatic noise except for the tropical Pacific ocean in transitional season. The potential predictability is diversified from month to month and one station to another, but generally the monthly mean temperature over China is potentially predictable at statistical significance level 0.10. The results suggest that we could not ask a climate model to predict the climate with satisfactory results worldwide in all seasons and that the regional model could be a hopeful way to predict the climate.With 3 Figures     

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.     

Estimation of the monthly and annual mean maximum and mean minimum air temperature values in Greece

In this study, regression equations to estimate the monthly and annual values of the mean maximum and mean minimum air temperatures in Greece are derived. For this purpose, data from 87 meteorological stations distributed all over Greece are used. Geographical parameters, i.e., altitude, latitude, longitude, minimum distance from the sea and an index of terrain morphology, are used as independent variables. These equations explain 79?C97% of the variance of the temperature values and have standard error of estimate between 0.59 and 1.20°C. Data from 37 other meteorological stations are used to validate the accuracy of the equations. Topographic or climatic factors, which could not be introduced into the equations, are responsible for most temperature residuals >0.5°C or <?0.5°C. Moreover, some particular emphasis has been given to the values of the regression coefficient for the altitude, since it is the estimator for the mean lapse rate of air temperature.     

Retrieving soil water contents from soil temperature measurements by using linear regression

A simple linear regression method is developed to retrieve daily averaged soil water content from diurnal variations of soil temperature measured at three or more depths. The method is applied to Oklahoma Mesonet soil temperature data collected at the depths of 5, 10, and 30 cm during 11-20 June 1995. The retrieved bulk soil water contents are compared with direct measurements for one pair of nearly collocated Mesonet and ARM stations and also compared with the retrievals of a previous method at 14 enhanced Oklahoma Mesonet stations. The results show that the current method gives more persistent retrievals than the previous method. The method is also applied to Oklahoma Mesonet soil temperature data collected at the depths of 5, 25, 60, and 75 cm from the Norman site during 20-30 July 1998 and 1-31 July 2000. The retrieved soil water contents are verified by collocated soil water content measurements with rms differences smaller than the soil water observation error (0.05m~3 m~(-3)). The retrievals are     

Intercomparison of prediction skills of ensemble methods using monthly mean temperature simulated by CMIP5 models

This study focuses on an objective comparison of eight ensemble methods using the same data, training period, training method, and validation period. The eight ensemble methods are: BMA (Bayesian Model Averaging), HMR (Homogeneous Multiple Regression), EMOS (Ensemble Model Output Statistics), HMR+ with positive coefficients, EMOS+ with positive coefficients, PEA_ROC (Performance-based Ensemble Averaging using ROot mean square error and temporal Correlation coefficient), WEA_Tay (Weighted Ensemble Averaging based on Taylor’s skill score), and MME (Multi-Model Ensemble). Forty-five years (1961-2005) of data from 14 CMIP5 models and APHRODITE (Asian Precipitation- Highly-Resolved Observational Data Integration Towards Evaluation of Water Resources) data were used to compare the performance of the eight ensemble methods. Although some models underestimated the variability of monthly mean temperature (MMT), most of the models effectively simulated the spatial distribution of MMT. Regardless of training periods and the number of ensemble members, the prediction skills of BMA and the four multiple linear regressions (MLR) were superior to the other ensemble methods (PEA_ROC, WEA_Tay, MME) in terms of deterministic prediction. In terms of probabilistic prediction, the four MLRs showed better prediction skills than BMA. However, the differences among the four MLRs and BMA were not significant. This resulted from the similarity of BMA weights and regression coefficients. Furthermore, prediction skills of the four MLRs were very similar. Overall, the four MLRs showed the best prediction skills among the eight ensemble methods. However, more comprehensive work is needed to select the best ensemble method among the numerous ensemble methods.     

Prediction of extreme rainfall event using weather pattern recognition and support vector machine classifier

A major component of flood alert broadcasting is the short-term prediction of extreme rainfall events, which remains a challenging task, even with the improvements of numerical weather prediction models. Such prediction is a high priority research challenge, specifically in highly urbanized areas like Mumbai, India, which is extremely prone to urban flooding. Here, we attempt to develop an algorithm based on a machine learning technique, support vector machine (SVM), to predict extreme rainfall with a lead time of 6–48 h in Mumbai, using mesoscale (20–200 km) and synoptic scale (200–2,000 km) weather patterns. The underlying hypothesis behind this algorithm is that the weather patterns before (6–48 h) extreme events are significantly different from those of normal weather days. The present algorithm attempts to identify those specific patterns for extreme events and applies SVM-based classifiers for extreme rainfall classification and prediction. Here, we develop the anomaly frequency method (AFM), where the predictors (and their patterns) for SVM are identified with the frequency of high anomaly values of weather variables at different pressure levels, which are present before extreme events, but absent for non-extreme conditions. We observe that weather patterns before the extreme rainfall events during nighttime (1800 to 0600Z) is different from those during daytime (0600 to 1800Z) and, accordingly, we develop a two-phase support vector classifier for extreme prediction. Though there are false alarms associated with this prediction method, the model predicts all the extreme events well in advance. The performance is compared with the state-of-the-art statistical technique fingerprinting approach and is observed to be better in terms of false alarm and prediction.     

On the mean monthly equivalent potential temperature and rainfall in West Africa

Summary Rainfall in West Africa is examined in relation to monthly mean equivalent potential temperature ( _e )at the earth's surface. The study revealed that monthly mean equivalent potential temperature ( _e ) and monthly rainfall (R) generally decreased northwards from the equator.A good relationship existed betweenR and _e in the northern zone of West Africa (i.e., north of 7.5° N). No definite relationship existed in the southern zone. In the northern zone, the departure of _e from its annual mean ( ) first became positive about a month before the onset of the rains. Positive departures from ) generally resulted in more than normal (or average) rainfall in this zone. In general, little or no rainfall occurred in West Africa whenever _e was less than 320 K.

---

Zusammenfassung Der Niederschlag (MonatssummeR) in Westafrika wird in Zusammenhang mit der mittleren monatlichen Äquivalent-temperatur ( _e ) an der Erdoberfläche untersucht. Es zeigte sich, daß die Monatswerte beider Elemente im allgemeinen vom Äquator nach Norden abnehmen.ZwischenR und _e ergab sich für das nördliche Westafrika (nördlich von 7.5° N) eine gute, für die südliche Zone jedoch keine beweisbare Übereinstimmung. In der nördlichen Zone übertraf _e das Jahresmittel erstmals etwa einen Monat vor Beginn der Regenzeit. Positive Abweichungen vom mittleren _e hatten immer übernormalen Niederschlag in dieser Zone zur Folge. Dagegen gab es wenig oder keinen Niederschlag in Westafrika, wenn _e unter 320 K lag.

---

---

With 7 Figures     

Specification of monthly mean anomalies of surface air temperature in Canada and Alaska

Abstract

Anomalies of monthly mean surface temperature observed at 55 stations in Canada and 13 in Alaska from 1951 through 1980 are related to concurrent anomalies of monthly mean 700‐mb height at a network of 107 grid points in North America and the surrounding oceans. The data are screened by a stepwise forward selection procedure to yield multiple regression equations for specifying the monthly mean temperature anomaly at each city and for each month from the field of simultaneous 700‐mb heights plus the previous month's local temperature anomaly. On the average, the specification equations explain 70% of the temperature variance and select as predictors approxiamtely 2 heights to the west of the reference station, 1.5 heights in the vicinity, 1 height to the east, and 0.5 previous temperatures.

Most of this paper describes various properties of the specification equations and related atmospheric characteristics on a regional, seasonal and month‐to‐month basis. Five statistical features are mapped for the months of January, April, July and October, and marked regional differences are noted. The above features are then averaged for the entire region and graphed month by month; the annual cycle of other properties is also described. Systematic spatial and temporal variations in the characteristics of temperature variability, persistence, correlation with height, and specification equations are illustrated.     

Prediction of Monthly Mean Surface Air Temperature in a Region of China

In conventional time series analysis, a process is often modeled as three additive components： linear trend, seasonal effect, and random noise. In this paper, we perform an analysis of surface air temperature in a region of China using a decomposition method in time series analysis. Applications to the National Centers for Environmental Prediction/the National Center for Atmospheric Research （NCEP/NCAR） Collaborative Reanalysis data in this region of China are discussed. The main finding was that the surface air temperature trend estimated for January 1948 to February 2006 was not statistically significant at 0.5904℃ （100 yr）^-1. Forecasting aspects are also considered.     

南方涛动与我国大尺度季、月气温的关系

南方涛动与我国月平均气温的相关,在当年2月、4月、9月以正相关为主,9月以后出现持续的负相关。南方涛动与我国季平均气温相关最显著的季节是当年秋季,其次是次年春季。南方涛动与次年长江下游、广东、福建、山东的年平均气温有良好的负相关。这些关系均可在预报中利用。此外,还指出,在我国4、5月,9、10月的大范围气温记录中存在早期识别厄尔尼诺的信号。     

河北省月平均气温分布式模型应用研究

利用1971-2000年河北省及周边126个气象台站的常规观测资料,应用月平均气温分布式模型,实现了起伏地形下河北省月平均气温的分布式模拟,制作出100 m×100 m分辨率的气温空间制图。结果表明：坝上高原和河北省平原地区地势平坦,气温分布比较均匀;燕山山脉和太行山山脉地形复杂,气温受局地地形影响显著。在角度相同的坡地上,偏南坡与偏北坡的气温差异1月>10月>4月>7月。在同一时段,偏南坡与偏北坡的气温差异随坡度的增加而增加;张家口地区多盆地河谷,气温分布均匀且较周围地区高;月平均气温分布式模型在河北省具有良好的模拟精度、时间维和空间维模拟性能及山地扩展性能。     

SVM和Elman神经网络在乌鲁木齐污染物预报中的应用研究

选用2012年11月1日-2013年1月31日的逐6 h的空气污染物（SO2、NO2、PM10）和实况气象要素（温度、湿度、能见度、风速和气压）资料,利用支持向量机和Elman神经网络方法建立空气污染物预报模型。结果表明,支持向量机和Elman神经网络方法都可以得到较为理想的预测结果,支持向量机在泛化能力方面具有显著优势,预测结果更加准确。