Search

In recent years, artificial intelligence in geosciences is spreading more and more, thanks to the availability of a large amount of data. In particular, the development of automatic raingauges networks allows to get rainfall data and makes these techniques effective, even if the performance of artificial intelligence models is a consequence of the coherency and quality of the input data. In this work, we intended to provide machine learning models capable of predicting rainfall data starting from the values of the nearest raingauges at one historic time point. Moreover, we investigated the influence of the anomalous input data on the prediction of rainfall data. We pursued these goals by applying machine learning models based on Linear Regression, LSTM and CNN architectures to several raingauges in Tuscany (central Italy). More than 75% of the cases show an R2 higher than 0.65 and a MAE lower than 4 mm. As expected, we emphasized a strong influence of the input data on the prediction capacity of the models. We quantified the model inaccuracy using the Pearson's correlation. Measurement anomalies in time series cause major errors in deep learning models. These anomalous data may be due to several factors such as temporary malfunctions of raingauges or weather conditions. We showed that, in both cases, the data-driven model features could highlight these situations, allowing a better management of the raingauges network and rainfall databases. [ABSTRACT FROM AUTHOR]

In this paper, we describe and analyze two datasets entitled "Homogenised monthly and daily temperature and precipitation time series in China during 1960–2021" and "Homogenised monthly and daily temperature and precipitation time series in Greece during 1960–2010". These datasets provide the homogenised monthly and daily mean (TG), minimum (TN), and maximum (TX) temperature and precipitation (RR) records since 1960 at 366 stations in China and 56 stations in Greece. The datasets are available at the Science Data Bank repository and can be downloaded from https://doi.org/10.57760/sciencedb.01731 and https://doi.org/10.57760/sciencedb.01720. For China, the regional mean annual TG, TX, TN, and RR series during 1960–2021 showed significant warming or increasing trends of 0.27°C (10 yr)−1, 0.22°C (10 yr)−1, 0.35°C (10 yr)−1, and 6.81 mm (10 yr)−1, respectively. Most of the seasonal series revealed trends significant at the 0.05 level, except for the spring, summer, and autumn RR series. For Greece, there were increasing trends of 0.09°C (10 yr)−1, 0.08°C (10 yr)−1, and 0.11°C (10 yr)−1 for the annual TG, TX, and TN series, respectively, while a decreasing trend of −23.35 mm (10 yr)−1 was present for RR. The seasonal trends showed a significant warming rate for summer, but no significant changes were noted for spring (except for TN), autumn, and winter. For RR, only the winter time series displayed a statistically significant and robust trend [−15.82 mm (10 yr)−1]. The final homogenised temperature and precipitation time series for both China and Greece provide a better representation of the large-scale pattern of climate change over the past decades and provide a quality information source for climatological analyses. [ABSTRACT FROM AUTHOR]

In this paper, we study centennial trends of climate and snow cover within the Ossola valley, in the Western Italian Alps. We pursue different tests (Mann Kendall MK, bulk, and sequential/progressive MKprog, Linear Regression, also with change point detection, and moving window average MW) on two datasets, namely (i) dataset1, daily temperature, precipitation, snow depth for 9 stations in the area, during 1930–2018, and (ii) dataset2, snow depth and density, measured twice a month (from February 1st to June 1st) for 47 stations during 2007–2023. We also verify correlation with glacier retreat nearby. In dataset1, we highlight a positive trend for minimum temperature with MK, and Linear Regression. Using MKprog/MW, a negative change of snow cover depth, and duration starting from the late 1980s is found. In dataset2, despite the annual variability in snow cover and 2022–2023 winter drought, we assess the maximum snow water equivalent (SWE) to be delayed with respect to maximum snow depth at high altitude (over a month above 2.700 m a.s.l.), highlighting the effect of settling in decreasing snow depth during spring. We also present a formula linking through Linear Regression the Day of the Year of SWE peak to altitude, relevant to assess the onset of thaw season. Due to the high altitude of the stations, and the paradigmatic nature of the Ossola Valley, hosting Toce River, a main contributor to the Lake Maggiore of Italy, our results are of interest, and can be used as a benchmark for the Italian Alps. [ABSTRACT FROM AUTHOR]

Long memory models can be generalised by the Fractional equal-root Autoregressive Moving Average (FerARMA) process, which displays short memory for a suitable parameter's set. Consequently, the spectrum is bounded, ensuring stationarity also for values of the memory parameter d larger than 0.5. The FerARMA generalization is proposed here to forecast highly persistent time series, as climate records of tree rings and paleo-temperature reconstructions. The main advantage of a bounded spectrum allows for more accurate predictions with respect to standard long memory models, especially if a long prediction horizon is considered. [ABSTRACT FROM AUTHOR]

The joint examination of the climate time series may be efficient methodology for the characterization of extreme weather and climate events. In general, the main difficulties are connected with the different probability distribution of the variables and the handling of the stochastic connection between them. The first problem can be solved by the standardization procedures, i.e., to transform the variables into standard normal ones. For example, there are the Standardized Precipitation Index (SPI) series for the precipitation sums assuming gamma distribution, or the standardization of temperature series assuming normal distribution. In case of more variables, the problem of stochastic connection can be solved on the basis of the vector norm of the transformed variables defined by their covariance matrix. We will present the developed mathematical methodology and some examples for its meteorological applications. [ABSTRACT FROM AUTHOR]

Reliable secular time series of essential climatic variables are a fundamental element for the assessment of vulnerability, impact and adaptation to climate change. Here, we implement a readily portable procedure for building an upgradable long‐term homogeneous climate dataset using monthly and daily observations of temperature and precipitation over a given area of interest, exemplified here with Abruzzo, a region in Central Italy characterized by complex orography. We process the dataset according to a preliminary ranking of stations based on data quantity and quality, and we exploit the Climatol algorithm for inhomogeneity correction. The corrected time series show trends in broad agreement with external databases (CRU, Berkeley Earth, E‐OBS), and highlight the importance of relying on a local network for a better representation of gradients and variability over the territory. We estimate that maximum (TX) and minimum temperature (TN) increased by ~1.6 and ~2.2°C/century, respectively, over the period 1930–2019, while in the recent decades 1980–2019 we found an accelerated trend of ~5.7 and ~3.9°C/century. Precipitation (RR) decreased by ~10%/century in 1930–2019, while it has been increasing at a rate of ~26%/century in 1980–2019. The Köppen–Geiger climate classification is sensitive to the increase of precipitation in the recent decades, which is attributable to decreased summer precipitation overcompensated by more rain in late spring and early autumn. The cold climate types are retreating upwards along the slopes of the mountain ranges. Over the period 1980–2019, extreme values are also displaying significant trends. Every 2 years, there is one less frost day (TN <0°C) and one more summer day (TX >25°C) in the Apennines area, while there is one more tropical night (TN >20°C) in the Adriatic coastal area. Precipitation extremes are increasing, especially along the coast, with rain accumulated in the rainiest days increasing at a rate of 1–2%/year. [ABSTRACT FROM AUTHOR]

The different phases of ENSO (El Niño Southern Oscillation) directly influence the occurrence of natural disasters and global warming. To limit the socio-economic impact, it is essential to develop simple and fast numerical models that can predict these different cycles. Here, we aimed to improve the predictive performance and extracting relevant information from climatic events by applying the signature method to time series models. After transforming the data using this signature method, we performed a comparative analysis of the statistical and machine learning models. In addition, we used PDP (Partial Dependence Plot) and SPRC (Standard Partial Regression Coefficient) to better understand the interactions between different climate indices. Our results showed that the best predictive performance was obtained when we use the signature method with the LSTM (R2 = 0.74) and Lasso (R2 = 0.79) models. Two complementary methods were used to highlight the influence of the following climate indices on ENSO cycle changes: NINO3, NINO3.4 and NPI (North Pacific Index). These methodologies also enabled us to determine the switchover thresholds, and order temporal variations. With this first application of the signature method to this type of time series, we obtained accurate forecasts on a 6-month scale with reduced computation time. This suggests that our methodology can be applied to many other fields of research that use multivariate time-series analyses. • The signature method improves the forecast performance of machine learning. • First application of the signature method to global-scale climate time series. • The signature allows accurate forecasts on a 6-month scale about ENSO cycles. • The LSTM model benefits from signature transformation. • Random forest allows us to better understand climate interactions. [ABSTRACT FROM AUTHOR]

Estimation of Pearson's correlation coefficient between two time series, in the evaluation of the influences of one time-dependent variable on another, is an often used statistical method in climate sciences. Data properties common to climate time series, namely non-normal distributional shape, serial correlation, and small data sizes, call for advanced, robust methods to estimate accurate confidence intervals to support the correlation point estimate. Bootstrap confidence intervals are estimated in the Fortran 90 program PearsonT (Mudelsee, Math Geol 35(6):651-665, ), where the main intention is to obtain accurate confidence intervals for correlation coefficients between two time series by taking the serial dependence of the data-generating process into account. However, Monte Carlo experiments show that the coverage accuracy of the confidence intervals for smaller data sizes can be substantially improved. In the present paper, the existing program is adapted into a new version, called PearsonT3, by calibrating the confidence interval to increase the coverage accuracy. Calibration is a bootstrap resampling technique that performs a second bootstrap loop (it resamples from the bootstrap resamples). It offers, like the non-calibrated bootstrap confidence intervals, robustness against the data distribution. Pairwise moving block bootstrap resampling is used to preserve the serial dependence of both time series. The calibration is applied to standard error-based bootstrap Student's $$t$$ confidence intervals. The performance of the calibrated confidence interval is examined with Monte Carlo simulations and compared with the performance of confidence intervals without calibration. The coverage accuracy is evidently better for the calibrated confidence intervals where the coverage error is acceptably small already (i.e., within a few percentage points) for data sizes as small as 20. [ABSTRACT FROM AUTHOR]