Variability of the date of monsoon onset over Kerala (India) of the period 1870–2014 and its relation to sea surface temperature

Monsoon onset over Kerala (India) which occurs every year is a major climatic phenomenon that involves large scale changes in wind, rainfall and sea surface temperature (SST). Over the last 150 years, the date of monsoon onset over Kerala (DMOK) has varied widely, the earliest being 11 May, 1918 and the most delayed being 18 June, 1972. DMOK has a long term (1870–2014) mean of 01 June and standard deviation of 7–8 days. We have studied the inter-annual and decadal time scale variability of DMOK and their relation with SST. We found that SST anomalies of large spatial scale similar to those in El Nino/La Nina are associated with the inter-annual variability in DMOK. Indian Ocean between latitudes \(5^{\circ }\hbox {S}\) and \(20^{\circ }\hbox {N}\) has two episodes of active convection associated with monsoon onset over Kerala (MOK), one around DMOK and the other about six weeks earlier (called pre-monsoon rain peak or bogus monsoon onset) and in between a two week period of suppressed convection occurs over north Indian Ocean. A prominent decadal time scale variability was found in DMOK having large and statistically significant linear correlation with the SST gradient across the equator over Indian and Pacific oceans, the large correlation persisting for several months prior to the MOK. However, no linear trend was seen in DMOK during the long period from 1870 to 2014.     

Changing predictability of Indian monsoon rainfall anomalies?

The predictability of Indian summer monsoon rainfall from pre-season circulation indices is explored from observations during 1939–91. The predictand is the all-India average of June–September precipitation NIR, and the precursors examined are the latitude position of the 500 mb ridge along 75°E in April (L), the pressure tendency April minus January at Darwin (DPT), March-April-May temperature at six stations in west central India (T6), the sea surface temperature (SST) anomaly in the northeastern Arabian Sea in May (ASM), SST anomaly in the Arabian Sea in January (ANJ), northern hemisphere temperature anomaly in January–February (NHT), and Eurasian snow cover in January (SNOW). Monsoon rainfall tends to be enhanced with a more northerly ridge position, small Darwin pressure tendency, warmer pre-season conditions, and reduced winter snow cover. However, relationships have varied considerably over the past half-century, with the strongest associations during 1950–80, and a drastic weakening in the 1980s. Four prediction models were constructed based on stepwise multiple regression, using as predictors combinations of L, DPT, T6, ASM, and NHT, with 1939–68 as “dependent” dataset, or training period, and 1969–91 as “independent” dataset or verification period. For the 1969–80 portion of the verification period calculated and observed NIR values agreed closely, with the models explaining 74–79% of the variance. By contrast, after 1980 predictions deteriorated drastically, with the explained variance for the 1969–89 time span dropping to 25–31%. The monsoon rainfall of 1990 and 1991 turned out to be again highly predictable from models based on stepwise multiple regression and linear discriminant analysis and using as input L + DPT or L + DPT + NHT, and with this encouragement an experimental real-time forecast was issued of the 1992 monsoon rainfall. These results underline the need for investigations into decadal-scale changes in the general circulation setting and raise concern for the continued success of seasonal forecasting.     

Multi-model ensemble schemes for predicting northeast monsoon rainfall over peninsular India

The northeast (NE) monsoon season (October, November and December) is the major period of rainfall activity over south peninsular India. This study is mainly focused on the prediction of northeast monsoon rainfall using lead-1 products (forecasts for the season issued in beginning of September) of seven general circulation models (GCMs). An examination of the performances of these GCMs during hindcast runs (1982–2008) indicates that these models are not able to simulate the observed interannual variability of rainfall. Inaccurate response of the models to sea surface temperatures may be one of the probable reasons for the poor performance of these models to predict seasonal mean rainfall anomalies over the study domain. An attempt has been made to improve the accuracy of predicted rainfall using three different multi-model ensemble (MME) schemes, viz., simple arithmetic mean of models (EM), principal component regression (PCR) and singular value decomposition based multiple linear regressions (SVD). It is found out that among these three schemes, SVD based MME has more skill than other MME schemes as well as member models.     

Surge dynamics of disaster displaced populations in temporary urban shelters: future challenges and management issues

A tropical cyclone (TC) precipitation prediction scheme has been developed based on the physical quantities of the NCEP/NCAR reanalysis data as potential predictors and using fuzzy neural network (FNN) model. TC precipitation samples from 172 tropical cyclones (TCs) affecting Guangxi, China, spanning 1980–2015 are used for model development. The FNN model input is constructed from potential predictors by employing both a stepwise regression method (SRM) and a locally linear embedding (LLE) algorithm. The LLE algorithm is capable of finding meaningful low-dimensional architectures hidden in their nonlinear high-dimensional data space and separating the underlying factors. In this scheme, the newly developed model, which is termed the FNN–LLE model, is used for daily TC precipitation prediction from 20:00 (Beijing Time, or BT) of the previous day to 20:00 BT of the current day at 89 stations covering Guangxi, China. Using identical modeling samples and independent samples, predictions of the FNN–LLE model are compared with the widely used SRM and interpolation method using the fine-mesh data of the European Centre for Medium-Range Weather Forecasts (ECMWF) in terms of the performance of TC rainfall prediction at 89 stations in Guangxi. The root-mean-square error (RMSE), bias, and equitable threat score (ETS) results were employed to assess the predicted outcomes. Results show that the FNN–LLE model is superior to the interpolation method by ECMWF and SRM for TC precipitation prediction with RMSE values of 21.94, 24.07, and 25.22 in FNN–LLE model, interpolation method by ECMWF and SRM, respectively. Moreover, FNN–LLE model having average bias and ETS values close to 1.0 gave better predictions than did the interpolation method by ECMWF and SRM.     

Statistical analyses of Global U-Pb Database 2017

The method of obtaining zircon samples affects estimation of the global U-Pb age distribution.Researchers typically collect zircons via convenience sampling and cluster sampling.When using these techniques,weight adjustments proportional to the areas of the sampled regions improve upon unweighted estimates.Here,grid-area and modern sediment methods are used to weight the samples from a new database of 418,967 U-Pb ages.Preliminary tests involve two age models.Model-1 uses the most precise U-Pb ages as the best ages.Model-2 uses the~(206)Pb/~(238)U age as the best age if it is less than a1000 Ma cutoff,otherwise it uses the~(207)Pb/~(206)Pb age as the best age.A correlation analysis between the Model-1 and Model-2 ages indicates nearly identical distributions for both models.However,after applying acceptance criteria to include only the most precise analyses with minimal discordance,a histogram of the rejected samples shows excessive rejection of the Model-2 analyses around the1000 Ma cutoff point.Because of the excessive rejection rate for Model-2,we select Model-1 as the preferred model.After eliminating all rejected samples,the remaining analyses use only Model-1 ages for five rock-type subsets of the database:igneous,meta-igneous,sedimentary,meta-sedimentary,and modern sediments.Next,time-series plots,cross-correlation analyses,and spectral analyses determine the degree of alignment among the time-series and their periodicity.For all rock types,the U-Pb age distributions are similar for ages older than 500 Ma,but align poorly for ages younger than 500 Ma.The similarities(500 Ma)and differences(500 Ma)highlight how reductionism from a detailed database enhances understanding of time-dependent sequences,such as erosion,detrital transport mechanisms,lithification,and metamorphism.Time-series analyses and spectral analyses of the age distributions predominantly indicate a synchronous period-tripling sequence of~91-Myr,~273-Myr,and~819-Myr among the various rock types.     

Location-specific weather predictions for Sriharikota (13.72°N, 80.22°E) through numerical atmospheric models during satellite launch campaigns

Accurate knowledge of different meteorological parameters over a launch site is very crucial for efficient management of satellite launch operations. Local weather over the Indian satellite launch site located at Sriharikota High Altitude Range (SHAR: 13.72°N, 80.22°E) is very much dependent on the atmospheric circulation prevailing over the Bay of Bengal oceanic region and topography-induced convective activities. With a view to providing severe weather threat prediction in terms of launch commit criteria (LCC), two numerical atmospheric models namely high-resolution regional model (HRM) and advanced regional prediction system (ARPS) are made operational over SHAR in a synoptic and mesoscale domain, respectively. In the present research article, two launch campaigns through Polar Satellite Launch Vehicle (PSLV-C11 and PSLV-C12) when contrasting weather conditions prevailed over the launch site are chosen for demonstration of potential of two models in providing location-specific short-to-medium-range weather predictions meeting the needs of LCC. In the case of PSLV-C11 campaign, when the launch site underwent frequent thundershower-associated rainfall, ARPS model–derived meteorological fields were effectively used in prediction of probability of the wet spells. On the other hand, Bay of Bengal underwent severe cyclonic storm during PSLV-C12 campaign, and its formation was reasonable captured through HRM simulations. It is concluded that a combination of HRM and ARPS provide reliable short-to-medium-range weather prediction over SHAR, which has got profound importance in launch-related activities.     

Multi-Physics ensemble prediction of tropical cyclone movement over Bay of Bengal

Ensemble prediction methodology based on variations in physical process parameterizations in tropical cyclone track prediction has been assessed. Advanced Research Weather Research and Forecasting model with 30-km resolution was used to make 5-day simulation of the movement of Orissa super cyclone (1999), one of the most intense tropical cyclones over the North Indian Ocean. Altogether 36 ensemble members with all possible combinations of three cumulus convection, two planetary boundary layer and six cloud microphysics parameterization schemes were produced. A comparison of individual members indicated that Kain–Fritsch cumulus convection scheme, Mellor–Yamada–Janjic planetary boundary layer scheme and Purdue Lin cloud microphysics scheme showed better performance. The best possible ensemble formulation is identified based on SPREAD and root mean square error (RMSE). While the individual members had track errors ranging from 96–240 km at 24 h to 50–803 km at 120 h, most of the ensemble predictions show significant betterment with mean errors less than 130 km up to 120 h. The convection ensembles had large spread of the cluster, and boundary layer ensembles had significant error disparity, indicating their important roles in the movement of tropical cyclones. Six-member ensemble predictions with cloud microphysics schemes of LIN, WSM5, and WSM6 produce the best predictions with least of RMSE, and large SPREAD indicates the need for inclusion of all possible hydrometeors in the simulation and that six-member ensemble is sufficient to produce the best ensemble prediction of tropical cyclone tracks over Bay of Bengal.     

Seasonal prediction of summer monsoon rainfall over cluster regions of India

Shared nearest neighbour (SNN) cluster algorithm has been applied to seasonal (June–September) rainfall departures over 30 sub-divisions of India to identify the contiguous homogeneous cluster regions over India. Five cluster regions are identified. Rainfall departure series for these cluster regions are prepared by area weighted average rainfall departures over respective sub-divisions in each cluster. The interannual and decadal variability in rainfall departures over five cluster regions is discussed. In order to consider the combined effect of North Atlantic Oscillation (NAO) and Southern Oscillation (SO), an index called effective strength index (ESI) has been defined. It has been observed that the circulation is drastically different in positive and negative phases of ESI-tendency from January to April. Hence, for each phase of ESI-tendency (positive and negative), separate prediction models have been developed for predicting summer monsoon rainfall over identified clusters. The performance of these models have been tested and found to be encouraging.     

Early prediction of onset of south west monsoon from ERS-1 scatterometer winds

Detailed analysis of the surface winds over the Indian Ocean derived from ERS-1 scatterometer data during the years 1993 and 1994 has been used to understand and unambiguously identify the onset phase of south-west monsoon. Five day (pentad) averaged wind vectors for the period April to June during both years have been examined to study the exact reversal of wind direction as well as the increase in wind speed over the Arabian Sea in relation to the onset of monsoon over the Indian west coast (Kerala). The related upper level humidity available from other satellites has also been analysed. The results of our analysis clearly show a consistent dramatic reversal in wind direction over the western Arabian Sea three weeks in advance of the onset of monsoon. The wind speed shows a large increase coinciding with the onset of monsoon. These findings together show the dominant role of sea surface winds in establishing the monsoon circulation. The study confirms that the cross equatorial current phenomenon becomes more important after the onset of monsoon.     

Spring temperature variability in northern Fennoscandia AD 1693–2011

A reconstruction of spring (April–May) temperature for northern Fennoscandia developed from the Tornionjoki (Tornio river) long cryophenological record of ice break‐up dates, back to AD 1693, is presented. The record is strongly climatically sensitive and explains 67% of the variance in the instrumental data over the last 150 years. The record exhibits a stepped decrease in the duration of the river's ice cover by 14 days, equivalent to an increase in April–May mean temperature of approximately 2.5°C over the last three centuries. The relationship between the date of ice break‐up, and accumulated daily mean temperatures (>0°C) is investigated. Uncertainty in the observation of ice break‐up is also considered in addition to the potential of this time series for regional climate model validation. Copyright © 2011 John Wiley & Sons, Ltd.     

Sedimentological evaluation of general circulation model simulations for the “greenhouse” Earth: Cretaceous and Jurassic case studies

Conceptual climate models, based on the workings of the present-day climate system, provided a first-order approach to ancient climate systems. They are potentially very subjective in character. Their main drawback was that they involved the relocation of continents beneath a stable atmospheric circulation modelled upon that of the present. General circulation models (GCMs) use the laws of physics and an understanding of past geography to simulate climatic responses. They are objective in character. However, they require super computers to handle vast numbers of calculations. Nonetheless it is now possible to compare results from different GCMs for a range of times and over a wide range of parameterisations. GCMs are currently producing simulated climate predictions which compare favourably with the distributions of climatically sensitive facies (e.g. coals, evaporites and palaeosols). They have been used effectively in the prediction of oceanic upwelling sites and the distribution of petroleum source-rocks and phosphorites. Parameterisation is the main weakness in GCMs (e.g. sea-surface temperature, orography, cloud behaviour). Sensitivity experiments can be run on GCMs which simulate the effects of Milankovitch forcing and thus provide insights into possible patterns of climate change both globally and locally (i.e. provide predictions that can be evaluated against the rock record). Future use of GCMs could be in the forward modelling of sequence stratigraphic evolution and in the prediction of the diagenetic characteristics of reservoir units in frontier exploration areas. The sedimentary record provides the only way that GCMs may themselves be evaluated and this is important because these same GCMs are being used currently to predict possible changes in future climate.     

Determination and prediction of standardized precipitation index (SPI) using TRMM data in arid ecosystems

Drought over a period threatens the water resources, agriculture, and socioeconomic activities. Therefore, it is crucial for decision makers to have a realistic anticipation of drought events to mitigate its impacts. Hence, this research aims at using the standardized precipitation index (SPI) to predict drought through time series analysis techniques. These adopted techniques are autoregressive integrating moving average (ARIMA) and feed-forward backpropagation neural network (FBNN) with different activation functions (sigmoid, bipolar sigmoid, and hyperbolic tangent). After that, the adequacy of these two techniques in predicting the drought conditions has been examined under arid ecosystems. The monthly precipitation data used in calculating the SPI time series (SPI 3, 6, 12, and 24 timescales) have been obtained from the tropical rainfall measuring mission (TRMM). The prediction of SPI was carried out and compared over six lead times from 1 to 6 using the model performance statistics (coefficient of correlation (R), mean absolute error (MAE), and root mean square error (RMSE)). The overall results prove an excellent performance of both predicting models for anticipating the drought conditions concerning model accuracy measures. Despite this, the FBNN models remain somewhat better than ARIMA models with R?≥?0.7865, MAE?≤?1.0637, and RMSE?≤?1.2466. Additionally, the FBNN based on hyperbolic tangent activation function demonstrated the best similarity between actual and predicted for SPI 24 by 98.44%. Eventually, all the activation function of FBNN models has good results respecting the SPI prediction with a small degree of variation among timescales. Therefore, any of these activation functions can be used equally even if the sigmoid and bipolar sigmoid functions are manifesting less adjusted R² and higher errors (MAE and RMSE). In conclusion, the FBNN can be considered a promising technique for predicting the SPI as a drought monitoring index under arid ecosystems.     

Spatial prediction of landslide susceptibility in western Serbia using hybrid support vector regression (SVR) with GWO,BAT and COA algorithms

In this study, we developed multiple hybrid machine-learning models to address parameter optimization limitations and enhance the spatial prediction of landslide susceptibility models. We created a geographic information system database, and our analysis results were used to prepare a landslide inventory map containing 359 landslide events identified from Google Earth, aerial photographs, and other validated sources. A support vector regression (SVR) machine-learning model was used to divide the landslide inventory into training (70%) and testing (30%) datasets. The landslide susceptibility map was produced using 14 causative factors. We applied the established gray wolf optimization (GWO) algorithm, bat algorithm (BA), and cuckoo optimization algorithm (COA) to fine-tune the parameters of the SVR model to improve its predictive accuracy. The resultant hybrid models, SVR-GWO, SVR-BA, and SVR-COA, were validated in terms of the area under curve (AUC) and root mean square error (RMSE). The AUC values for the SVR-GWO (0.733), SVR-BA (0.724), and SVR-COA (0.738) models indicate their good prediction rates for landslide susceptibility modeling. SVR-COA had the greatest accuracy, with an RMSE of 0.21687, and SVR-BA had the least accuracy, with an RMSE of 0.23046. The three optimized hybrid models outperformed the SVR model (AUC = 0.704, RMSE = 0.26689), confirming the ability of metaheuristic algorithms to improve model performance.     

Method for prediction of landslide movements based on random forests

Prediction of landslide movements with practical application for landslide risk mitigation is a challenge for scientists. This study presents a methodology for prediction of landslide movements using random forests, a machine learning algorithm based on regression trees. The prediction method was established based on a time series consisting of 2 years of data on landslide movement, groundwater level, and precipitation gathered from the Kostanjek landslide monitoring system and nearby meteorological stations in Zagreb (Croatia). Because of complex relations between precipitations and groundwater levels, the process of landslide movement prediction is divided into two separate models: (1) model for prediction of groundwater levels from precipitation data and (2) model for prediction of landslide movements from groundwater level data. In a groundwater level prediction model, 75 parameters were used as predictors, calculated from precipitation and evapotranspiration data. In the landslide movement prediction model, 10 parameters calculated from groundwater level data were used as predictors. Model validation was performed through the prediction of groundwater levels and prediction of landslide movements for the periods from 10 to 90 days. The validation results show the capability of the model to predict the evolution of daily displacements, from predicted variations of groundwater levels, for the period up to 30 days. Practical contributions of the developed method include the possibility of automated predictions, updated and improved on a daily basis, which would be an important source of information for decisions related to crisis management in the case of risky landslide movements.     

Probabilistic multi-model ensemble prediction of Indian summer monsoon rainfall using general circulation models: A non-parametric approach

Probabilistic prediction has the ability to convey the intrinsic uncertainty of forecast that helps the decision makers to manage the climate risk more efficiently than deterministic forecasts. In recent times, probabilistic predictions obtained from the products from General Circulation Models (GCMs) have gained considerable attention. The probabilistic forecast can be generated in parametric (assuming Gaussian distribution) as well as non-parametric (counting method) ways. The present study deals with the non-parametric approach that requires no assumption about the form of the forecast distribution for the prediction of Indian summer monsoon rainfall (ISMR) based on the hindcast run of seven general circulation models from 1982 to 2008. Probabilistic prediction from each of the GCM products has been generated by non-parametric methods for tercile categories (viz. below normal (BN), near-normal (NN), and above normal (AN)) and evaluation of their skill is assessed against observed data. Five different types of PMME schemes have been used for combining probabilities from each GCM to improve the forecast skill as compared to the individual GCMs. These schemes are different in nature of assigning the weights for combining probabilities. After a rigorous analysis through Rank Probability Skill Score (RPSS) and relative operating characteristic (ROC) curve, the superiority of PMME has been established over climatological probability. It is also found that, the performances of PMME1 and PMME3 are better than all the other methods whereas PMME3 has showed more improvement over PMME1.     

Construction of Statistical-Dynamic Prediction Model for the Asian-Australian Summer Monsoon Based on the Predictable Mode Analysis Method and Assessment of Its Predictive Skills

Due to the limitations of model performances, the predictive skills of current climate models for the Asian-Australian summer monsoon precipitation are still poor. The prediction based on the combination of statistical and dynamic approaches is an effective way to improve the predictive skills. We used such method to identify the predictable modes of the Asian-Australian summer monsoon precipitation with clear physical interpretation from the historical observational data. Then we combined the principal components time series of these modes predicted by the coupled models, which is derived from the seasonal prediction experiments in the ENSEMBLES project, and the corresponding spatial patterns derived from the above observational analysis to reconstruct the precipitation field. These formed a statistical-dynamic seasonal prediction model for the Asian-Australian summer monsoon precipitation. We analyzed the predictive skills of the model at 1-, 4-and 7-month leads. The result shows that the forecast skills of the statistical-dynamic prediction model are higher than those of the simple dynamic predictions. In addition, the predictive skills of the Multi-Model Ensemble (MME) mean are superior to those of any individual models. Therefore, it is very necessary to implement multi-model ensemble prediction for the monsoon precipitation.     

Interannual variability of onset of the summer monsoon over India and its prediction

In this article, the interannual variability of certain dynamic and thermodynamic characteristics of various sectors in the Asian summer monsoon domain was examined during the onset phase over the south Indian peninsula (Kerala Coast). Daily average (0000 and 1200 UTC) reanalysis data sets of the National Centre for Environmental Prediction/National Centre for Atmospheric Research (NCEP/NCAR) for the period 1948–1999 were used. Based on 52 years onset date of the Indian summer monsoon, we categorized the pre-onset, onset, and post-onset periods (each an average of 5 days) to investigate the interannual variability of significant budget terms over the Arabian Sea, Bay of Bengal, and the Indian peninsula. A higher difference was noticed in low-level kinetic energy (850 hPa) and the vertically integrated generation of kinetic energy over the Arabian Sea from the pre-onset, onset, and post-onset periods. Also, significant changes were noticed in the net tropospheric moisture and diabatic heating over the Arabian Sea and Indian peninsula from the pre-onset to the post-onset period. It appears that attaining the magnitude of 40 m² s⁻² and then a sharp rise in kinetic energy at 850 hPa is an appropriate time to declare the onset of the summer monsoon over India. In addition to a sufficient level of net tropospheric moisture (40 mm), a minimum strength of low-level flow is needed to trigger convective activity over the Arabian Sea and the Bay of Bengal. An attempt was also made to develop a location-specific prediction of onset dates of the summer monsoon over India based on energetics and basic meteorological parameters using multivariate statistical techniques. The regression technique was developed with the data of May and June for 42 years (1948–1989) and validated with 10 years NCEP reanalysis from 1990 to 1999. It was found that the predicted onset dates from the regression model are fairly in agreement with the observed onset dates obtained from the Indian Meteorology Department.     

青海夏季干旱特征及其预测模型研究

利用1961-2008年青海非干旱区(除柴达木盆地)地面气象观测资料、 74个环流特征量、 海温资料、 北半球500 hPa高度场网格点资料以及500 hPa高度场遥相关, 对夏季干旱的变化趋势和干旱发生的机理进行了研究.结果表明:1961-2008年夏季青海省非干旱区、 东部农业区分别发生干旱15 a、 18 a, 发生干旱的年几率为31.3%、 37.5%; 东部农业区发生干旱的几率较大, 中轻度干旱发生几率大于特大、 重度干旱.夏季典型干旱年500 hPa欧亚中高纬度上空高度距平分布为正距平, 极涡偏弱; 非干旱年蒙古到青藏高原上由负距平控制, 极涡偏强, 偏向东半球, 印缅低压槽十分活跃.当夏季西大西洋型、 上年秋季欧亚纬向环流指数偏弱, 而4月西太平洋型偏强, 8月青藏高原地面加热场强度距平指数偏强, 夏季容易发生干旱; 反之, 当夏季西大西洋型、 上年秋季欧亚纬向环流指数偏强, 而4月西太平洋型偏弱, 8月青藏高原地面加热场强度距平指数偏弱, 则夏季不易发生夏季干旱. 1961-2008年模拟方程的准确率为83.3%, 2009-2010年预测结果与实况接近, 趋势预测准确.     

Prediction on landslide displacement using a new combination model: a case study of Qinglong landslide in China

Prediction on landslide displacement plays an important role in landslide early warning. Many models have been proposed for this purpose. However, the accuracy of the prediction results by these models often varies under different conditions. Rational evaluation and comprehensive consideration of these results still remain a scientific challenge. A new comprehensive combination model is proposed to predict the landslides displacement. The elementary displacement prediction is made by the support vector machine model, the exponential smoothing model, and the gray model (GM)(1,1). The results of the models are comprehensively evaluated by combining the results and introducing the accuracy matrix. The optimal weight in the evaluation work is obtained. A rational prediction result can be attained based on the so-called combination model. The proposed method has been tested by the application of Qinglong landslides in Guizhou Province, China. The comparison between the prediction results and in situ measurement shows that the prediction precision of the proposed model is satisfactory. The root-mean-square error (RMSE) of the combination model can be reduced to 1.4316 (monitoring site JCK2), 1.2623 (monitoring site JCK4), 2.3758 (monitoring site JCK6), 2.2704 (monitoring site JCK8), 1.4247 (monitoring site JCK11), and 0.9449 (monitoring site JCK12), which is much lower than the RMSE of the individual models.

     

A comparative analysis among computational intelligence techniques for dissolved oxygen prediction in Delaware River

Most of the water quality models previously developed and used in dissolved oxygen (DO) prediction are complex. Moreover, reliable data available to develop/calibrate new DO models is scarce. Therefore, there is a need to study and develop models that can handle easily measurable parameters of a particular site, even with short length. In recent decades, computational intelligence techniques, as effective approaches for predicting complicated and significant indicator of the state of aquatic ecosystems such as DO, have created a great change in predictions. In this study, three different AI methods comprising: (1) two types of artificial neural networks (ANN) namely multi linear perceptron (MLP) and radial based function (RBF); (2) an advancement of genetic programming namely linear genetic programming (LGP); and (3) a support vector machine (SVM) technique were used for DO prediction in Delaware River located at Trenton, USA. For evaluating the performance of the proposed models, root mean square error (RMSE), Nash–Sutcliffe efficiency coefficient (NS), mean absolute relative error (MARE) and, correlation coefficient statistics (R) were used to choose the best predictive model. The comparison of estimation accuracies of various intelligence models illustrated that the SVM was able to develop the most accurate model in DO estimation in comparison to other models. Also, it was found that the LGP model performs better than the both ANNs models. For example, the determination coefficient was 0.99 for the best SVM model, while it was 0.96, 0.91 and 0.81 for the best LGP, MLP and RBF models, respectively. In general, the results indicated that an SVM model could be employed satisfactorily in DO estimation.