Minimum Surface Air Temperature Research Articles

Moscow Urban Heat Island: Detection, boundaries, and variability

The determination of the anthropogenic impact on the thermal regime of a megalopolis is discussed. The nominal boundaries of the megalopolis thermal effects are determined taking into account orographic and landscape effects and the variability of temperature difference between the center of the city and its suburbs. It is proposed to use the data of stations located within 40-50 km from the urban agglomeration boundaries for calculating the anthropogenic component of the urban heat island. The parameters of the heat island in Moscow are determined from the measurement data on maximum and minimum daily surface air temperature in 2012-2016, and their significant seasonal and intradaily variability is revealed. It is corroborated that anthropogenic causes for the higher air temperature in Moscow as compared with suburbs during the cold season are the release and loss of heat of heating systems; in the absence of snow cover the anthropogenic effect of these sources is enhanced by the thermal radiation of the urban surface.

Changes in climate extremes in Vietnam

This study represents the observed and projected changes in climate extremes in the context of climate change. A linear regression method was used to identify trends and variations of climate extremes in the past, while projected changes in climate extremes were detected based on dynamical downscaling and compared with the baseline period (1986-2005). The results show that daily maximum and minimum surface air temperatures (Txx, Tnn) increased considerably with the highest value of about 0.9oC/decade during the period of 1961-2014. The number of hot days increased across most observed stations, especially in Central Vietnam. Trends of extreme precipitation varied among the regions: decreasing across most stations in the Northwest, Northeast and Red River Delta; and increasing at other stations. In Vietnam, over the period 1959-2015, there were no obvious changes in the frequency of tropical cyclones causing landfall events, including typhoons and tropical depressions. However, the rate of occurrence of the very strongest typhoons are currently on the increase. In recent years, typhoon seasons tend to end later than in the past, and typhoons occur more frequently in Southern regions. Predicted results of climate extremes in the future for the periods 2046-2065 and 2080-2099, compared with the baseline period of 1986-2005 under the medium and high scenario (RCP 4.5 and RCP 8.5) show that temperature extremes will likely increase considerably in the future. The number of hot days will likely increase in both frequency and intensity, while the number of extreme cold days will likely decrease at most stations in the North. A maximum 1-day rainfall will also likely increase across all regions in Vietnam. Results from models show that the number of strong typhoons will likely increase in the mid- and late-century.

Trends in day-to-day variability of surface air temperature in China during 1961–2012

Using daily mean, maximum, and minimum surface air temperature (abbreviated as Tmean, Tmax, and Tmin, respectively) data from CN05.2 and the Met Office Hadley Centre observation data-sets for 1961–...

Projected changes in rainfall and temperature over homogeneous regions of India

The impact of climate change on the characteristics of seasonal maximum and minimum temperature and seasonal summer monsoon rainfall is assessed over five homogeneous regions of India using a high-resolution regional climate model. Providing REgional Climate for Climate Studies (PRECIS) is developed at Hadley Centre for Climate Prediction and Research, UK. The model simulations are carried out over South Asian domain for the continuous period of 1961–2098 at 50-km horizontal resolution. Here, three simulations from a 17-member perturbed physics ensemble (PPE) produced using HadCM3 under the Quantifying Model Uncertainties in Model Predictions (QUMP) project of Hadley Centre, Met. Office, UK, have been used as lateral boundary conditions (LBCs) for the 138-year simulations of the regional climate model under Intergovernmental Panel on Climate Change (IPCC) A1B scenario. The projections indicate the increase in the summer monsoon (June through September) rainfall over all the homogeneous regions (15 to 19%) except peninsular India (around 5%). There may be marginal change in the frequency of medium and heavy rainfall events (>20 mm) towards the end of the present century. The analysis over five homogeneous regions indicates that the mean maximum surface air temperatures for the pre-monsoon season (March–April–May) as well as the mean minimum surface air temperature for winter season (January–February) may be warmer by around 4 °C towards the end of the twenty-first century.

Can Temperature Extremes in East Antarctica be Replicated from ERA Interim Reanalysis?

ABSTRACTBased on daily minimum, maximum, and mean surface air temperatures (T min, T max, T mean) from the European Centre for Medium-Range Weather Forecast’s reanalysis from 1979 onwards (ERA Interim), the accuracy of daily T min and T max reanalysis is assessed against in situ observations from four automatic weather stations (Zhongshan, EAGLE, LGB69, and Dome A) in East Antarctica from 2005 to 2008. ERA Interim generally shows a warm bias for T min and a cool bias for T max, with an underestimation of the diurnal temperature range. The reanalysis explains more than 84% of the daily and annual variance, and the replicating ability decreases gradually from the coast to the interior, with annual root mean square errors of 2.4 °C, 2.6 °C, 3.0 °C, and 4.3 °C for daily T min, and 2.2 °C, 3.1 °C, 3.4 °C, and 4.9 °C for daily T max at Zhongshan, LGB69, Eagle, and Dome A, respectively. ERA Interim shows little seasonal variability, although it performs a little better in the austral spring and worse in winter and autumn at Dome A. An analysis on spatial distribution of temperature and wind field indicates that ERA Interim has successfully replicated the progress of temperature extremes developing, occurring, and disappearing. In addition, weather events extracted from ERA Interim mainly occur on the same day as observations, with high cross-correlation coefficients (R ≧ 0.287, N ≧ 1131, P < 0.001). ERA Interim has, despite its regional limitations and deficiencies, proved to be a powerful tool for weather and climate studies in the Antarctic region.

WATER RESOURCES STATUS AND AVAILABILITY ASSESSMENT IN CURRENT AND FUTURE CLIMATE CHANGE SCENARIOS FOR BEAS RIVER BASIN OF NORTH WESTERN HIMALAYA

Abstract. The water resources status and availability of any river basin is of primary importance for overall and sustainable development of any river basin. This study has been done in Beas river basin which is located in North Western Himalaya for assessing the status of water resources in present and future climate change scenarios. In this study hydrological modelling approach has been used for quantifying the water balance components of Beas river basin upto Pandoh. The variable infiltration capacity (VIC) model has been used in energy balance mode for Beas river basin at 1km grid scale. The VIC model has been run with snow elevation zones files to simulate the snow module of VIC. The model was run with National Centre for Environmental Prediction (NCEP) forcing data (Tmax, Tmin, Rainfall and wind speed at 0.5degree resolution) from 1 Jan. 1999 to 31 Dec 2006 for calibration purpose. The additional component of glacier melt was added into overall river runoff using semi-empirical approach utilizing air temperature and glacier type and extent data. The ground water component is computed from overall recharge of ground water by water balance approach. The overall water balance approach is validated with river discharge data provided by Bhakra Beas Management Board (BBMB) from 1994-2014. VIC routing module was used to assess pixel wise flow availability at daily, monthly and annual time scales. The mean monthly flow at Pandoh during study period varied from 19 - 1581 m3/s from VIC and 50 to 1556 m3/sec from observation data, with minimum water flow occurring in month of January and maximum flow in month of August with annual R2 of 0.68. The future climate change data is taken from CORDEX database. The climate model of NOAA-GFDL-ESM2M for IPCC RCP scenario 4.5 and 8.5 were used for South Asia at 0.44 deg. grid from year 2006 to 2100. The climate forcing data for VIC model was prepared using daily maximum and minimum near surface air temperature, daily precipitation and daily surface wind speed. The GFDL model also gives validation phase scenarios from 2006 to 2015, which are used to test the overall model performance with current data. The current assessment made by hydrological water balance based approach has given reasonable good results in Beas river basin. The main limitation of this study is lack of full representation of glacier melt flow using fully energy balance model. This component will be addressed in coming time and it will be integrated with tradition hydrological and snowmelt runoff models. The other limitation of current study is dependence on NCEP or other reanalysis of climate forcing data for hydrological modelling, this leads to mismatch between actual and simulated water balance components. This problem can be addressed if more ground based and fine resolution grid based hydro meteorological data are used as input forcing data for hydrological modelling.

New gridded daily climatology of Finland: Permutation‐based uncertainty estimates and temporal trends in climate

AbstractLong‐term time series of key climate variables with a relevant spatiotemporal resolution are essential for environmental science. Moreover, such spatially continuous data, based on weather observations, are commonly used in, e.g., downscaling and bias correcting of climate model simulations. Here we conducted a comprehensive spatial interpolation scheme where seven climate variables (daily mean, maximum, and minimum surface air temperatures, daily precipitation sum, relative humidity, sea level air pressure, and snow depth) were interpolated over Finland at the spatial resolution of 10 × 10 km2. More precisely, (1) we produced daily gridded time series (FMI_ClimGrid) of the variables covering the period of 1961–2010, with a special focus on evaluation and permutation‐based uncertainty estimates, and (2) we investigated temporal trends in the climate variables based on the gridded data. National climate station observations were supplemented by records from the surrounding countries, and kriging interpolation was applied to account for topography and water bodies. For daily precipitation sum and snow depth, a two‐stage interpolation with a binary classifier was deployed for an accurate delineation of areas with no precipitation or snow. A robust cross‐validation indicated a good agreement between the observed and interpolated values especially for the temperature variables and air pressure, although the effect of seasons was evident. Permutation‐based analysis suggested increased uncertainty toward northern areas, thus identifying regions with suboptimal station density. Finally, several variables had a statistically significant trend indicating a clear but locally varying signal of climate change during the last five decades.

Development of daily temperature scenarios and their impact on paddy crop evapotranspiration in Kangsabati command area

Production of the principal paddy crop in West Bengal state of India is vulnerable to climate change due to limited water resources and strong dependence on surface irrigation. Therefore, assessment of impact of temperature scenarios on crop evapotranspiration (ETc) is essential for irrigation management in Kangsabati command (West Bengal). In the present study, impact of the projected temperatures on ETc was studied under climate change scenarios. Further, the performance of the bias correction and spatial downscaling (BCSD) technique was compared with the two well-known downscaling techniques, namely, multiple linear regression (MLR) and Kernel regression (KR), for the projections of daily maximum and minimum air temperatures for four stations, namely, Purulia, Bankura, Jhargram, and Kharagpur. In National Centers for Environmental Prediction (NCEP) and General Circulation Model (GCM), 14 predictors were used in MLR and KR techniques, whereas maximum and minimum surface air temperature predictor of CanESM2 GCM was used in BCSD technique. The comparison results indicated that the performance of the BCSD technique was better than the MLR and KR techniques. Therefore, the BCSD technique was used to project the future temperatures of study locations with three Representative Concentration Pathway (RCP) scenarios for the period of 2006–2100. The warming tendencies of maximum and minimum temperatures over the Kangsabati command area were projected as 0.013 and 0.014 °C/year under RCP 2.6, 0.015 and 0.023 °C/year under RCP 4.5, and 0.056 and 0.061 °C/year under RCP 8.5 for 2011–2100 period, respectively. As a result, kharif (monsoon) crop evapotranspiration demand of Kangsabati reservoir command (project area) will increase by approximately 10, 8, and 18 % over historical demand under RCP 2.6, 4.5, and 8.5 scenarios, respectively.

Bayesian geoadditive modelling of climate extremes with nonparametric spatially varying temporal effects

ABSTRACTNon‐stationary modelling of climate extremes has attracted significant attention in recent years. Generalized extreme value (GEV) distribution is the standard approach for modelling block extremes. The non‐stationary form of GEV distribution with location and scale parameters linearly regressing to time has been widely used for single‐site time series of climate extremes. In the present paper, such a model is extended to be geoadditive for regional climate extremes with nonparametric spatially varying temporal effects. The model is implemented through the Bayesian hierarchical approach and is applied to the annual minimum surface air temperature in Mainland China as a demonstration. The five bivariate functions representing the regional regression parameters in the model are approximated by low‐rank thin plate regression splines. All the necessary software tools for producing a thin plate regression spline basis and Markov Chain Monte Carlo sampling can be seamlessly integrated and called through the R software environment. Results show that the model can capture the spatial inhomogeneity and temporal non‐stationarity features of regional climate extremes. Expressions for return levels with different return periods and their trends are also derived accordingly. Trends in different return levels, especially those in the ‘extreme state’ (e.g., 50‐year return level) that cannot be inferred by usual observational trend analysis, can also be derived through samples drawn from posterior predictive distributions. The regional trend in a 50‐year return level is quite different to that in a 2‐year return level. More attention should be paid to trends in the extreme state of climate under climate change.

Role of radiatively forced temperature changes in enhanced semi-arid warming in the cold season over east Asia

Abstract. As climate change has occurred over east Asia since the 1950s, intense interest and debate have arisen concerning the contribution of human activities to the observed warming in past decades. In this study, we investigate regional surface temperature change during the boreal cold season using a recently developed methodology that can successfully identify and separate the dynamically induced temperature (DIT) and radiatively forced temperature (RFT) changes in raw surface air temperature (SAT) data. For regional averages, DIT and RFT contribute 44 and 56 % to the SAT over east Asia, respectively. The DIT changes dominate the SAT decadal variability and are mainly determined by internal climate variability, represented by the North Atlantic Oscillation (NAO), Pacific Decadal Oscillation (PDO), and Atlantic Multi-decadal Oscillation (AMO). Radiatively forced SAT changes have made a major contribution to the global-scale warming trend and the regional-scale enhanced semi-arid warming (ESAW). Such enhanced warming is also found in radiatively forced daily maximum and minimum SAT. The long-term global-mean SAT warming trend is mainly related to radiative forcing produced by global well-mixed greenhouse gases. The regional anthropogenic radiative forcing, however, caused the enhanced warming in the semi-arid region, which may be closely associated with local human activities. Finally, the relationship between the so-called "global warming hiatus" and regional enhanced warming is discussed.

The Effects of Gridding Algorithms on the Statistical Moments and Their Trends of Daily Surface Air Temperature*

Abstract This paper explores the effects from averaging weather station data onto a grid on the first four statistical moments of daily minimum and maximum surface air temperature (SAT) anomalies over the entire globe. The Global Historical Climatology Network–Daily (GHCND) and the Met Office Hadley Centre GHCND (HadGHCND) datasets from 1950 to 2010 are examined. The GHCND station data exhibit large spatial patterns for each moment and statistically significant moment trends from 1950 to 2010, indicating that SAT probability density functions are non-Gaussian and have undergone characteristic changes in shape due to decadal variability and/or climate change. Comparisons with station data show that gridded averages always underestimate observed variability, particularly in the extremes, and have altered moment trends that are in some cases opposite in sign over large geographic areas. A statistical closure approach based on the quasi-normal approximation is taken to explore SAT’s higher-order moments and point correlation structure. This study focuses specifically on relating variability calculated from station data to that from gridded data through the moment equations for weighted sums of random variables. The higher-order and nonlinear spatial correlations up to the fourth order demonstrate that higher-order moments at grid scale can be determined approximately by functions of station pair correlations that tend to follow the usual Kolmogorov scaling relation. These results can aid in the development of constraints to reduce uncertainties in climate models and have implications for studies of atmospheric variability, extremes, and climate change using gridded observations.

Orbital control on late Miocene climate and the North African monsoon: insight from an ensemble of sub-precessional simulations

Abstract. Orbital forcing is a key climate driver over multi-millennial timescales. In particular, monsoon systems are thought to be driven by orbital cyclicity, especially by precession. Here, we analyse the impact of orbital forcing on global climate with a particular focus on the North African monsoon, by carrying out an ensemble of 22 equally spaced (one every 1000 years) atmosphere–ocean–vegetation simulations using the HadCM3L model, covering one full late Miocene precession-driven insolation cycle with varying obliquity (between 6.568 and 6.589 Ma). The simulations only differ in their prescribed orbital parameters, which vary realistically for the selected time period. We have also carried out two modern-orbit control experiments, one with late Miocene and one with present-day palaeogeography, and two additional sensitivity experiments for the orbital extremes with varying CO2 forcing. Our results highlight the high sensitivity of the North African summer monsoon to orbital forcing, with strongly intensified precipitation during the precession minimum, leading to a northward penetration of vegetation up to ~ 21° N. The modelled summer monsoon is also moderately sensitive to palaeogeography changes, but it has a low sensitivity to atmospheric CO2 concentration between 280 and 400 ppm. Our simulations allow us to explore the climatic response to orbital forcing not only for the precession extremes but also on sub-precessional timescales. We demonstrate the importance of including orbital variability in model–data comparison studies, because doing so partially reduces the mismatch between the late Miocene terrestrial proxy record and model results. Failure to include orbital variability could also lead to significant miscorrelations in temperature-based proxy reconstructions for this time period, because of the asynchronicity between maximum (minimum) surface air temperatures and minimum (maximum) precession in several areas around the globe. This is of particular relevance for the North African regions, which have previously been identified as optimal areas to target for late Miocene palaeodata acquisition.

English

The determination of variance of surface air temperature is very essential since it has a direct impact on vegetation, environment and human livelihood. Forecast of surface air temperature is difficult because of the complex physical phenomenon and the random-like behavior of atmospheric system which influences the temperature event on the earth surface. In this study, forecast models based on artificial neural network (ANN) and genetic programming (GP) approaches were proposed to predict lead seven days minimum and maximum surface air temperature using the weather parameters observed at the station Chennai, India. The outcome of this study stated that models formulated using ANN approach are more accurate than genetic programming for all seven days with the highest coefficient of determination (R2), least mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE) on deployment with independent test dataset. ANN models give statistically acceptable mean absolute error of 0.59oC for lead day one in minimum temperature forecast and 0.86oC variance for lead day one in maximum temperature forecast. The study also clarified that the level of accuracy of the proposed prediction models were found to be better for smaller lead days when compared with higher lead days with both approaches.

Changes in the Covariability of Surface Air Temperature and Precipitation over East Asia Associated with Climate Shift in the Late 1970s

AbstractVariations in surface air temperature and precipitation are closely associated because of their thermo-dynamic relations. The climate shift in the late 1970s and associated changes in precipitation over East Asia have been well reported. However, how the covariability of surface air temperature and precipitation responds to the climate shift is not yet well understood. We used the ob-served mean (Tmean), daily maximum (Tmax), and minimum (rmin) surface air temperatures and precipitation during the period of 1953–2000 to explore this issue. Results show that the covariability between Tmeaa and precipitation experienced remarkable changes over certain areas of East Asia after the climate shift with evident seasonal dependencies. In winter, after the climate shift significantly negative correlations occupied more areas over Mongolia and China. By contrast, in summer after the climate shift significantly negative correlations which existed over almost entire East Asia during the pre-shift period were ...

Variability in projected elevation dependent warming in boreal midlatitude winter in CMIP5 climate models and its potential drivers

The future rate of climate change in mountains has many potential human impacts, including those related to water resources, ecosystem services, and recreation. Analysis of the ensemble mean response of CMIP5 global climate models (GCMs) shows amplified warming in high elevation regions during the cold season in boreal midlatitudes. We examine how the twenty-first century elevation-dependent response in the daily minimum surface air temperature [d(ΔTmin)/dz] varies among 27 different GCMs during winter for the RCP 8.5 emissions scenario. The focus is on regions within the northern hemisphere mid-latitude band between 27.5°N and 40°N, which includes both the Rocky Mountains and the Tibetan Plateau/Himalayas. We find significant variability in d(ΔTmin)/dz among the individual models ranging from 0.16 °C/km (10th percentile) to 0.97 °C/km (90th percentile), although nearly all of the GCMs (24 out of 27) show a significant positive value for d(ΔTmin)/dz. To identify some of the important drivers associated with the variability in d(ΔTmin)/dz during winter, we evaluate the co-variance between d(ΔTmin)/dz and the differential response of elevation-based anomalies in different climate variables as well as the GCMs’ spatial resolution, their global climate sensitivity, and their elevation-dependent free air temperature response. We find that d(ΔTmin)/dz has the strongest correlation with elevation-dependent increases in surface water vapor, followed by elevation-dependent decreases in surface albedo, and a weak positive correlation with the GCMs’ free air temperature response.

Daily minimum and maximum surface air temperatures from geostationary satellite data

AbstractObservations of daily minimum and maximum land air temperatures, Tmin and Tmax, have traditionally been obtained through in situ observations at meteorological stations. While the station network is extensive, many land masses are poorly observed. Moreover, observations at stations are “point” observations and may not be representative of air temperatures at neighboring locations. Satellites provide the means to observe surface skin temperatures at spatial scales of tens of meters to kilometers. But although skin and near‐surface air temperatures may be strongly coupled, the two quantities can differ by several degrees over land, where the magnitude of the difference is variable in both space and time. This study describes a method for estimating daily Tmin and Tmax at the pixel scale using geostationary satellite data, providing spatially detailed observations for areas unobserved in situ. A dynamic multiple linear regression model is developed using daily minimum and maximum land surface temperature (LSTmin and LSTmax), fraction of vegetation, distance from coast, latitude, urban fraction, and elevation as predictors. The method is demonstrated over Europe for 2012–2013; evaluation with collocated station observations indicates a mean satellite‐minus‐station bias of 0.0 to 0.5°C with root‐mean‐square difference of 2.3 to 2.7°C. The data derived here are not designed to replace traditional gridded station air temperature data sets, but to augment them. Satellite surface temperature data usually have larger uncertainties than in situ data sets, but they can offer spatial detail and coverage that the latter may not provide.

Investigating diurnal and seasonal climatic response to land use and land cover change over monsoon Asia with the Community Earth System Model

AbstractLand use and land cover change (LULCC) is primarily characterized as forest conversion to cropland for the development of agriculture. Previous climate modeling studies have demonstrated the LULCC impacts on mean climate and its long‐term trends. This study investigates the diurnal and seasonal climatic response to LULCC in monsoon Asia through two numerical experiments with potential and current vegetation cover using the fully coupled Community Earth System Model. Results show that LULCC leads to a reduced diurnal temperature range due to the enhanced (reduced) diurnal cycle of the ground heat flux (sensible heat flux). Daily minimum surface air temperature (Tmin) exhibits a clear seasonality over India as it increases most in the premonsoon season and least during the summer monsoon season. Similarly, a strong anticyclonic anomaly is present at 850 hPa over India in spring and over eastern China in autumn, but weak changes in circulation appear in winter and summer. In addition, the LULCC results in significant changes in the variability of the 2 m air temperature, as characterized by an enhanced variability in India and a reduced variability in northern China to eastern Mongolia in autumn and winter. Possible land‐atmosphere feedback loops involving surface albedo, soil moisture, evapotranspiration, atmospheric circulation, and precipitation are offered as biogeophysical mechanisms that are responsible for the region‐specific LULCC‐induced diurnal and seasonal response.

Climate Change Projections over CORDEX East Asia Domain using Multi-RCMs

In this study, global climate change scenario by HadGEM2-AO is dynamically downscaled using 5 Regional Climate Models (RCMs) for the Coordinated Regional Downscaling Experiment (CORDEX) - East Asia. All RCMs with 50-km horizontal resolution are integrated for continuous 71 years (1979-2049). In the evaluation of historical runs, all models tend to overestimate precipitation over the subtropical western North Pacific. For the surface air temperature, all models have cold biases over the East Asian continent compared to the observation. In the future projections, all models show similar changes in precipitation such as increasing precipitation over the tropical oceans and slightly decreasing precipitation over the East China Sea. In the mid-latitudes, convective precipitation tends to increase due to the enhanced convective instability, while non-convective precipitation tends to decrease slightly due to the changes in subtropical high and monsoon circulation. For surface air temperature, the warming tendency over the entire domain is captured by all models and it is larger in high latitude regions. The warming temperature is related to the increase in longwave radiation, which leads to prominent increase in minimum surface air temperature.

Statistical Variability and Persistence Change in Daily Air Temperature Time Series from High Latitude Arctic Stations

In the last decades, Arctic communities have been reporting that weather conditions are becoming less predictable. Most scientific studies have not been able to consistently confirm such a trend. The question regarding the possible increase in weather variability was addressed here based on daily minimum and maximum surface air temperature time series from 15 high latitude Arctic stations from Canada, Norway, and the Russian Federation. A range of analysis methods were applied, distinguished mainly by the way in which they treat time scale. Statistical L-moments were determined for temporal windows of different lengths. While the picture provided by L-scale and L-kurtosis is not consistent with an increasing variability, L-skewness was found to change towards more positive values, reflecting an enhancement of warm spells. Haar wavelet analysis was applied both to the entire time series and to running windows. Persistence diagrams were generated based on running windows advancing through time and on local slopes of Haar analysis graphs; they offer a more nuanced view on variability by reflecting its change over time on a range of temporal scales. Local increases in variability could be identified in some cases, but no consistent change was detected in any of the stations over the studied temporal scales. The possibility for other intervals of temporal scale (e.g., days, hours, minutes) to potentially reveal a different situation cannot be ruled out. However, in the light of the results presented here, explanations for the discrepancy between variability perception and results of pattern analysis might have to be explored using an integrative approach to weather variables such as air temperature, cloud cover, precipitation, wind, etc.

MARSpline model for lead seven-day maximum and minimum air temperature prediction in Chennai, India

In this study, a Multivariate Adaptive Regression Spline (MARS) based lead seven days minimum and maximum surface air temperature prediction system is modelled for station Chennai, India. To emphasize the effectiveness of the proposed system, comparison is made with the models created using statistical learning technique Support Vector Machine Regression (SVMr). The analysis highlights that prediction accuracy of MARS models for minimum temperature forecast are promising for short term forecast (lead days 1 to 3) with mean absolute error (MAE) less than 1 °C and the prediction efficiency and skill degrades in medium term forecast (lead days 4 to 7) with slightly above 1 °C. The MAE of maximum temperature is little higher than minimum temperature forecast varying from 0.87 °C for day-one to 1.27 °C for lag day-seven with MARS approach. The statistical error analysis emphasizes that MARS models perform well with an average 0.2 °C of reduction in MAE over SVMr models for all ahead seven days and provide significant guidance for the prediction of temperature event. The study also suggests that the correlation between the atmospheric parameters used as predictors and the temperature event decreases as the lag increases with both approaches.