The primary aim of this work is to study the response of the normalized difference vegetation index (NDVI) of landscapes in the Lower Tigris Basin to current global and regional climate variability presented, respectively, by the global circulation indices and monthly temperatures and precipitation extracted from five observational/reanalysis datasets. The second task is to find the dataset that best reflects the regional vegetation and climate conditions. Comparison of the Köppen-Trewartha bioclimatic landscapes with the positions of botanical districts, land cover types, and streamflow estimates led to the conclusion that only two datasets correctly describe regional climatic zones. Therefore, searching for the NDVI response to regional climate variability requires the use of normalized analogues of temperatures and precipitations, as well as the Spearman Rank Correlation. We found that March/April NDVI, as proxies of the maximum biological productivity of the regional landscapes, are strongly correlated with October- March precipitation derived from three datasets and January-March temperatures derived from one dataset. We discovered the significant impact of autumn-winter El-Niño-Southern Oscillation and winter Indian Oceanic Dipole states on regional weather (e.g. all recent five severe droughts occurred during strong La Niña events). However, the strength of this impact on the vegetation was clearly linked to the zonal landscape type. By selecting pairs of the temperature/precipitation time-series that best correlated with NDVI at a given landscape, we have built a synthetic climate dataset. The landscape approach presented in this work can be used to validate the viability of any dataset when assessing the impacts of climate change and variability on weather-dependent components of the Earth's surface.

No abstract

The overall goal of our work is to find economic and robust supervised machine learning methods which adequate to both individual and collective Student Performance Forecast (SPF). The individual SPF are subject of well-known classification methods but collective SPF is subject of quantification learning algorithms dealing with the novel task to predict the frequency of classes in tested sample e.g. a number of students with unsatisfactory grade. The need for revise of classification methods shows review of 86 SPF in developing countries. The analysis depicts that most of SPF report the high overall accuracy of classifiers based on decision tree J48, Naïve Base NB, Multilayer Perception MLP, k-Nearest Neighbor k-NN, and Support Vector Machine SVM algorithms, but did not take into account the accuracy of the forecast of a minor presented class. So, given the imbalance in the sample, "useful forecast" with the F1 metric above 50% (75%) are given only in ½ (1/5) of cases of forecasts. The pivotal study of the efficacy factors of binary SPF (data type, algorithm, sample balancing, number of classes etc.). Another important finding is that classifiers with the probabilistic Naïve Bayesian kernel, have more stable behavior to classify different EDM datasets, overcoming MLP, J48, SVM and k-NN based classifiers which sometimes achieved good forecast but sometimes failed in prediction. After that, collected all the above experimental finds associated with relationship between algorithm and data information, we construct 3-15 member heterogeneous ensembles contained strong, moderate and weak classifiers for deterministic individual SPF by simple voting and heuristically proposed how individual probabilistic predictions could be generated and how to aggregate them for overall frequency forecasting, i.e. resolve the task of quantification. The proposed methods of ensemble forecasting and ensemble quantification can become the basis for new economic and robust solutions of various real-world problems in the field of machine learning.

This study examines the perspective of artificial neural networks for forecast Normalized Differential Vegetation Index (NDVI) on Diyala River basin and also how information about of bioclimatic landscapes will affect to forecasting performance. To do this, in the first stage of the experiment, a total of 20 perceptrons with different one hidden layer architectures were trained with sitespecific variables (latitude, longitude, minimal, maximal and mean height, landcover type) and seasonal meteorological variables (precipitation sum, and minimal, maximal and average daily temperatures) by error back propagation algorithm on the data of 2000-2010 years and tested on data for 2011-2016 years. It has been shown that the best performance, with determination coefficient R 2 of 0.78, was achieved by perceptron model with 12 hidden neurons the activated by logistic activation function and hyperbolic tangential activation of output value of NDVI. The large spatial heterogeneity of forecasting performance of the best perceptron was detected: in upper part of basin characterized according to Köppen-Trewartha bioclimatic classification, as landscapes of temperate mountain climate and the subtropical climate with dry summers, R 2 was 0.76-0.80, whereas in dry steppe landscapes and semi-desert landscapes of Diyala downstream R 2 was 0.6-0.7. The second stage of experiments with 20 models of perceptrons where the type of landscape was added as input variable or where 150 individual perceptrons were selected for each landscape, have shown that these approaches allows to R 2 increase up to 0.73-0.85. However, the strong contrast between characteristics of individual models complicates their use in the practice and requires to finding of new forecasting approaches.

No abstract