Marco Verdecchia

A Neural Network Approach to Real-Time Rainfall Estimation for Africa Using Satellite Data

Abstract Operational, real-time rainfall estimation on a daily timescale is potentially of great benefit for hydrological forecasting in African river basins. Sparseness of ground-based observations often means that only methodologies based predominantly on satellite data are feasible. An approach is presented here in which Cold Cloud Duration (CCD) imagery derived from Meteosat thermal infrared imagery is used in conjunction with numerical weather model analysis data as the input to an artificial neural network. Novel features of this approach are the use of principal component analysis to reduce the data requirements for the weather model analyses and the use of a pruning technique to identify redundant input data. The methodology has been tested using 4 yr of daily rain gauge data from Zambia in central Africa. Calibration and validation were carried out using pixel area rainfall estimates derived from daily rain gauge data. When compared with a standard CCD approach using the same dataset, the neural ...

Changing hydrological conditions in the Po basin under global warming.

The Po River is a crucial resource for the Italian economy, since 40% of the gross domestic product comes from this area. It is thus crucial to quantify the impact of climate change on this water resource in order to plan for future water use. In this paper a mini ensemble of 8 hydrological simulations is completed from 1960 to 2050 under the A1B emission scenario, by using the output of two regional climate models as input (REMO and RegCM) at two different resolutions (25 km-10 km and 25 km-3 km). The river discharge at the outlet point of the basin shows a change in the spring peak of the annual cycle, with a one month shift from May to April. This shift is entirely due to the change in snowmelt timing which drives most of the discharge during this period. Two other important changes are an increase of discharge in the wintertime and a decrease in the fall from September to November. The uncertainty associated with the winter change is larger compared to that in the fall. The spring shift and the fall decrease of discharge imply an extension of the hydrological dry season and thus an increase in water stress over the basin. The spatial distributions of the discharge changes are in agreement with what is observed at the outlet point and the uncertainty associated with these changes is proportional to the amplitude of the signal. The analysis of the changes in the anomaly distribution of discharge shows that both the increases and decreases in seasonal discharge are tied to the changes in the tails of the distribution, i.e. to the increase or decrease of extreme events.

Analysis of surface ozone using a recurrent neural network

Hourly concentrations of ozone (O₃) and nitrogen dioxide (NO₂) have been measured for 16 years, from 1998 to 2013, in a seaside town in central Italy. The seasonal trends of O₃ and NO₂ recorded in this period have been studied. Furthermore, we used the data collected during one year (2005), to define the characteristics of a multiple linear regression model and a neural network model. Both models are used to model the hourly O₃ concentration, using, two scenarios: 1) in the first as inputs, only meteorological parameters and 2) in the second adding photochemical parameters at those of the first scenario. In order to evaluate the performance of the model four statistical criteria are used: correlation coefficient, fractional bias, normalized mean squared error and a factor of two. All the criteria show that the neural network gives better results, compared to the regression model, in all the model scenarios. Predictions of O₃ have been carried out by many authors using a feed forward neural architecture. In this paper we show that a recurrent architecture significantly improves the performances of neural predictors. Using only the meteorological parameters as input, the recurrent architecture shows performance better than the multiple linear regression model that uses meteorological and photochemical data as input, making the neural network model with recurrent architecture a more useful tool in areas where only weather measurements are available. Finally, we used the neural network model to forecast the O₃ hourly concentrations 1, 3, 6, 12, 24 and 48 h ahead. The performances of the model in predicting O₃ levels are discussed. Emphasis is given to the possibility of using the neural network model in operational ways in areas where only meteorological data are available, in order to predict O₃ also in sites where it has not been measured yet.

Diurnal temperature range for a doubled carbon dioxide concentration experiment: Analysis of possible physical mechanisms

An analysis of the results of a climate simulation for a doubling of atmospheric carbon dioxide concentration over the European region is reported. Physical mechanisms are sought which could explain possible changes in the diurnal temperature range (DTR) under conditions of increased atmospheric greenhouse gas content. We show that an important contribution to changes in DTR is given by soil moisture. In areas where soil moisture increases due to an increase in precipitation there is a positive change in latent heat flux and a decrease in sensible heat flux. As a result, in areas with increasing soil moisture, the increase in maximum daytime temperature will be smaller than that in minimum temperature, thereby causing a decrease in the DTR. The opposite occurs for areas which undergo soil drying. This process amplifies the effect of cloud changes on surface solar and infrared radiation and dominates the direct effect of downward infrared radiation associated with increasing greenhouse gas concentration. Because the soil water content is largely controlled by precipitation, our results are consistent with early observational findings of negative correlation between changes in precipitation and in diurnal temperature range.

Continuous lidar measurements of stratospheric aerosols and ozone after the Pinatubo eruption Part II: Time evolution of ozone profiles and of aerosol properties

Two lidar systems, an aerosol lidar and an O3 Differential Absorption Lidar (DIAL), have been routinely operated at the same site (L‧Aquila, Italy; 42°N, 13°E) since August 1991. The multiwavelength analysis of the lidar signals allows to retrieve parameters related to equivalent aerosol size distributions and their optical properties. These are needed to correct the ozone DIAL profiles from the disturbance introduced by the stratospheric volcanic aerosols. The method and the confidence of the retrieved ozone profiles are discussed in a companion paper. Here we present the whole measurement series of ozone and backscattering ratio profiles during the period from August 1991 to December 1992. In addition, for some observations, the mode radius and the dispersion of the representative aerosol size distribution are reported. The time evolutions of aerosol surface area density and mass mixing ratio are also discussed within the uncertainties of the retrieval algorithm.

A Neural Network Approach for blocking recognition

We propose to use an Artificial Neural Network (ANN) for meteorological blocking recognition. The network output is presented as a number which ranges between 0 (absence of blocking) and 1 (blocked situation). This output is then compared with the step function obtained with a blocking index used in meteorological analysis and in the recognition of synoptic maps. We show that the ANN can pick events which are disregarded by the TM index and that ANN performances are equivalent and in some cases better than those indicated by an analytically computed blocking index.

Mapping of Alternaria and Pleospora concentrations in Central Italy using meteorological forecast and neural network estimator

Airborne particles (pollens and fungal spores) are recognized as important causes of allergies and many other pathologies whose main symptoms are usually associated with respiratory problems. In addition, these particles seem to be responsible for clinical symptoms of oculorhinitis and bronchial asthma. Many authors showed how pollen and spore concentrations are critically linked to meteorological conditions, while other studies investigated the possibility to estimate these concentrations through meteorological parameters. So, many different approaches have been proposed, and one of the most sophisticated is based on the use of a complex artificial neural network architecture. Once the neural device is calibrated using simultaneous time series of observed meteorological parameters and airborne biological particles, it is straightforward to use the Neural Network to predict spore concentrations using operational Limited Area Meteorological Model. In a previous work, it has been shown that the MM5 meteorological model developed by National Center for Atmospheric Research and Pennsylvania State University can be coupled with the above-cited neural predictor to provide a good prediction of Alternaria and Pleospora spore in the location of L’Aquila (Central Italy). Following the same approach, this work aims to provide the mapping of spore concentration over a wide area covered by high-resolution meteorological prediction in Central Italy. The complex patterns of fungal spore concentrations in selected areas will be described, and the high temporal variability of such fields will be discussed as well. The possibility to infer useful information from the predicted pattern of spore concentrations is discussed, as an example it appears that for people suffering from allergy to fungal spores is more comfortable to spend summertime close to the east coast of Italian Peninsula respect to the west coast. A further step of this work may easily lead to an operational use of the model for supporting the clinical management of allergies and for establishing a preventive strategy in agriculture to avoid unsafe and useless pollution of atmosphere, crops and fields.

Assessment of multiple daily precipitation statistics in ERA-Interim driven Med-CORDEX and EURO-CORDEX experiments against high resolution observations

We assess the statistics of different daily precipitation indices in ensembles of Med-CORDEX and EURO-CORDEX experiments at high resolution (grid spacing of ~0.11°, or RCM11) and medium resolution (grid spacing of ~0.44°, or RCM44) with regional climate models (RCMs) driven by the ERA-Interim reanalysis of observations for the period 1989–2008. The assessment is carried out by comparison with a set of high resolution observation datasets for nine European subregions. The statistics analyzed include quantitative metrics for mean precipitation, daily precipitation probability density functions (PDFs), daily precipitation intensity, frequency, 95th percentile and 95th percentile of dry spell length. We assess an ensemble including all Med-CORDEX and EURO-CORDEX models together and others including the Med-CORDEX and EURO-CORDEX separately. For the All Models ensembles, the RCM11 one shows a remarkable performance in reproducing the spatial patterns and seasonal cycle of mean precipitation over all regions, with a consistent and marked improvement compared to the RCM44 ensemble and the ERA-Interim reanalysis. A good consistency with observations by the RCM11 ensemble (and a substantial improvement compared to RCM44 and ERA-Interim) is found also for the daily precipitation PDFs, mean intensity and, to a lesser extent, the 95th percentile. A general improvement by the RCM11 models is also found when the data are upscaled and intercompared at the 0.44° and 1.5° resolutions. For some regions the RCM11 ensemble overestimates the occurrence of very high intensity events while for one region the models underestimate the occurrence of the most intense extremes. The RCM11 ensemble still shows a general tendency to underestimate the dry day frequency and 95th percentile of dry spell length over wetter regions, with only a marginal improvement compared to the lower resolution models. This indicates that the problem of the excessive production of low precipitation events found in many climate models persists also at relatively high resolutions, at least in wet climate regimes. Concerning the Med-CORDEX and EURO-CORDEX ensembles we find that their performance is of similar quality over the Mediterranean regions analyzed. Finally, we stress the need of consistent and quality checked fine scale observation datasets for the assessment of RCMs run at increasingly high horizontal resolutions.

Global Ozone Depletion and the Antarctic Ozone Hole

The secular trend of the Antarctic ozone hole has been studied with a two-dimensional model which can simulate formation of polar stratospheric clouds and includes heterogeneous chemical reactions. Results from the numerical simulation have been validated by comparison with available experimental data. Trends up to the year 2010 using standard (i.e., homogeneous) and heterogeneous chemistry have been compared and show that global ozone depletion reached 5–6% in the last 30 years and will average 8% for the next 20 years. Subtracting a 2% loss due to standard chemistry in the presence of trace gas increase in the last 30 years, we find a 3–4% global ozone loss due to heterogeneous chemistry. The depletion is evident even outside the southern hemisphere spring season and at mid-latitudes, pointing to an increase in global ozone sink.

An estimate of the Antarctic ozone modulation by the QBO

The possible effects of the QBO on the ozone distribution have been studied including in a 2D model a parameterization of Kelvin and Rossby-gravity wave forcing in the lower equatorial stratosphere. A chemical code complete with heterogeneous reactions allows a simulation of the ozone depletion due to the increase of stratospheric chlorine. With this model, the authors study the possible modulation of the secular trend in the Antarctic ozone hole by the QBO. When heterogeneous chemistry is not included, the model shows a polar ozone oscillation ({plus minus}6 Dobson Units) comparable to that deduced from early measurements (1970-1975). When heterogeneous reactions are taken into account, the model predicts a larger ozone oscillation in the Southern Hemisphere ({plus minus}12 Dobson Units) comparable to that obtained from recent observations. This behavior seems to point out a QBO induced temperature effect and its feedback on PSC with activation of heterogeneous chemistry.