It can be argued that the most stringent test of understanding a system, is to be able to use available observables to forecast a future outcome. It can also be argued that solar flares may not be deterministic, and even if they are, our present understanding is nowhere close to being able to predict the time, location, magnitude and space weather impact of a solar flare with any certainty. Still, solar flare prediction is a needed component of our international space weather infrastructure.

Many research groups around the world are investigating ways to improve flare forecasting methods, especially in light of new observational data available. I present a brief report of the "state of the field", summarizing insights gained from recent workshops (held in 2009 and 2013) aimed at head-to-head comparisons of flare forecasting methods in specific contexts. In summary, today's methods combine sophisticated data analysis with statistical or machine-learning algorithms which generally result in probabilistic forecasts. While the question of whether any of the presently developed methods clearly outperforms the others is still under investigation, the problem of how to answer that question presents interesting challenges for study design, data handling, statistical verification algorithm development (and psychology). I describe the standard skill scores being applied to the careful comparisons and some of the preliminary results, including what appear to be common successes and continued challenges among today's methods.

(Funding for the workshops and subsequent analysis was made possible by NASA/Living with a Star contract NNH09CE72C and NASA/Guest Investigator contract NNH12CG10C.)

Subsurface flows beneath active regions provide information about the processes that occur prior to and during flares. We have shown (Reinard et al. 2010) that subsurface flows exhibit measurable changes prior to flare occurrence. However, this relationship is not yet reliable on a case-by-case basis. We present recent research designed to improve the correlation between subsurface flow properties and solar flaring. We use measurements derived from GONG data using the ring-diagram technique, which provides flow properties between 0-16Mm below the solar surface. Our investigations include the effects of CME-association, hemispheric origin, and nearby active regions.

Several studies show that fractal and multifractal parameters derived from photospheric magnetic field measurements may help assess the eruptive potential of Active Regions (AR). However, the results derived from a recent analysis seem to discard all previously collected evidences, by showing that none of the widely-used indicators analyzed in the study allows to distinguish AR with major flares from flare-quiet ones, though both regions exhibit significant fractality and multifractality. In this framework, we explore whether higher-resolution, higher-cadence, and higher-sensitivity data than those employed in previous analyses might help to clarify the reason of conflicting results presented in the literature. To this aim, we analyzed two datasets of photospheric magnetograms of a flaring and a flare-quiet ARs, NOAA 1158 and NOAA 1117 respectively, which were observed simultaneously with SOHO/MDI and SDO/HMI. The latter observations are characterized by a higher spatial and temporal resolution than the SOHO/MDI data, which are analysed for comparison here and were widely employed in previous studies. Besides, we analyze a larger sample of ARs observed with SDO/HMI and associated with M- and X classes flare events. The results obtained are discussed with respect to those published in the literature.

This study has been carried out in the framework of the EU FP7 project "eHEROES - Environment for Human Exploration and RObotic Experimentation in Space" (grant n. 284461).

The most probable sites of flare onset are the regions of high horizontal magnetic field gradient in solar active regions. Besides the localization of flare producing areas the aim of the present work is to reveal the characteristic temporal variations in these regions prior to flares in order to improve the forecast capability. The method uses sunspot data instead of magnetograms and follows the behaviour of a suitably defined proxy measure representing the horizontal magnetic field gradient. The empirical basis of the analysis is the SDD (SOHO/MDI-Debrecen Data) sunspot catalogue. The following pre-flare signatures seem to be diagnostically useful predicting tools in the variation of the magnetc field gradient: i) steep increase, ii) high maximum, iii) significant fluctuation and iv) a gradual decrease between the maximum and the flare onset. The achived maximum has a linear relationship with the intensity of the flare. After the maximum the gradual decrease of the gradient has a characteristic time of about 10 hours. Close tracking of the variations of the magnetic field gradient allows to esteem the intensity of the flare and the expected time of the flare onset.
The research leading to these results has received funding from the European Community's Seventh Framework Programme (FP7/2012-2015) under grant agreement No. 284461.

Knowledge of the state of the photospheric magnetic field at a single instant in time does not appear sufficient to completely predict solar flares, although it may provide necessary conditions, such as the free magnetic energy needed for a flare to occur. Given the necessary conditions, evolution of the field, possibly by only a small amount, may trigger the onset of a flare. We present the results of a study using time series of photospheric vector field data from the Helioseismic and Magnetic Imager (HMI) on NASA's Solar Dynamics Observatory (SDO) to quantitatively parameterize both the state and evolution of active regions - their complexity, magnetic topology and energy - as related to solar flare events. Statistical tests based on linear and nonparametric Discriminant Analysis were used to compare pre-flare epochs to a control group of flare-quiet active regions. The improvements in forecasts that include evolution of the field, as quantified by standard skill scores, will be presented.

Funding for this work was provided by NASA/Guest Investigator contract NNH12CG10C.

The solar flare mechanism remains incompletely understood, and existing methods of flare prediction are probabilistic, rather than deterministic, relying on imperfect indicators of flare likelihood. The existing methods of prediction are briefly reviewed, and then an approach to the problem, relying only on the past time series of flare occurrence (Wheatland 2004; 2005), is revisited. A specific improvement to the method is presented: an implementation of a new optimal Bayesian change-point algorithm (Scargle et al. 2013) to determine the current rate of occurrence of small flares.

How changes in the three-dimensional magnetic field of solar active region are related to Coronal Mass Ejections (CME) is an important question for contemporary solar physics. Complex active regions are the predominant source of powerful high-speed CMEs, which can result in strong geomagnetic storms. In this paper we present the properties of chromospheric magnetic field of active regions that produced solar flares and CMEs using observations of the Synoptic Optical Long-term Investigations of the Sun (SOLIS) facility operated by the National Solar Observatory. Currently, the SOLIS Vector Spectromagnetograph (VSM) is the only instrument that is capable of obtaining full Stokes profiles in both the photospheric Fe I ë630.2 nm and chromospheric Ca II ë854.2 nm lines on a daily basis. VSM also has the capability of making rapid scans covering an area sufficiently large to contain an active region. We shall present the Stokes profile characteristics of photospheric and chromospheric lines of few CME source regions.

It is widely accepted that magnetic reconnection is the primary mechanism for energy release during solar eruptions. However, the question of how, exactly reconnection works in the solar corona remains very much under discussion. In fact, we can learn much about how reconnection drives flares and CMEs from relatively simple, analytical models. Here we discuss some recent analytical results and the relevance of these results for flare and CME forecasts. In particular we will discuss how the reconnection rate and location, and therefore the rate and locations of energy release, can be influenced by the local conditions and asymmetric magnetic fields in the corona at the time of an eruption. We highlight some lessons that can be gleaned from analytical theory for real-time space weather forecasters and model developers.

Introduction
A possible basis for an early warning system for solar flare activity was observed by happenstance during otherwise routine measurements of Mn-54 decay gammas with an ordinary scintillation detector [Astropart. Phys. 31, 407 (2009)]. Further investigation by our group led to additional evidence pointing to a possible solar influence on terrestrial nuclear decay rates [Astropart. Phys. 32, 42 (2009), Space Sci. Rev. 145, 285 (2009)]. The only known particles emitted by the Sun which could have caused the observed phenomena are neutrinos, particularly in light of the fact that whatever agent is causing the decay rate changes experienced unimpeded passage through the Earth and the Sun. As noted in [Astropart. Phys. 31, 407 (2009)], the observed decrease in the Mn-54 decay rate began to occur up to 36 hours prior to the flares, thus providing a possible predictive method, which could offer stakeholders the ability to protect vulnerable assets and systems.

Background
A significant body of work has been published to date detailing the observations of this putative solar influence beyond the flare-associated decay rate changes. Annual and sub-annual periodicities (~2/yr and 10-12/yr) have also been observed, and these are difficult to attribute to possible known environmental influences. Moreover, the number of observations--by independent groups--of similar periodicities, across a number of different isotopes and diverse detector types [see Table 1 in Appl. Radiat. and Isotop. 74, 50 (2013)], essentially eliminate all known environmental influences as the cause. Although there is no physical mechanism presently to explain these effects, particularly in light of the extremely low cross-section for known neutrino interactions, it is likely that these effects arise from some new phenomenon.

The work done by our group in November and December of 2006, at the end of solar cycle 23, exhibited decay rate changes associated with several flares during the time period. This prompted further investigations of the literature, and analyses of nuclear decay data generously offered by other groups at different institutions world-wide. These analyses led to the identification of the aforementioned periodicities, or non-random behavior, in what should be randomly distributed data [Astropart. Phys. 34, 121 (2010), Solar Phys. 267, 251 (2010), Solar Phys. 272, 1 (2011)].

Method Development
Additional measurements of Mn-54 decays to search for have been underway for the last five years, and recent results are promising. While is it certain that this methodology does not show the same effects for all solar flares, decay rate changes associated with a reasonable percentage of C, M and X class flares can be identified. Analytical methods are being developed that may provide real-time predictive capability for major flares with the ability to attach confidence levels to the flare prediction. Additional detector systems are being installed to increase data flow and offer a better statistical basis for observed flare correlations. As we are nearing the peak of Solar Cycle 24, we have an excellent opportunity for observations and development of the algorithms that will be used.

System Development
Several years of data on Mn-54 decays, taken at 1-hour intervals, are currently available from sites at Purdue University (Indiana, USA) and at the United States Air Force Academy (Colorado, USA). In addition, a third site in Santiago, Chile, has recently been established. Our goal is to establish a world-wide network of similar installations, acquiring data synchronously, and transmitting these data to one or more central servers. Software is currently under development which would allow these data to be combined and analyzed in real-time to search for common features, which would signal an imminent solar storm. Since the statistical power of such an array would increase as N^1/2 for N sites, an order of magnitude improvement could be realized by an appropriate world-wide array of 100 systems.

Future Plans
The fact that changes in decay rates of some radioactive isotopes can precede solar storms by more than a day makes our technology quite promising. Moreover, it seems likely that this technology can be combined with other techniques presently being studied to predict solar flares (See J. Space Weather Space Clim. 3, A17 (2013), and http://www.cost.eu/domains_actions/essem/Actions/ES0803]. Our eventual goal is thus to establish a larger network combining these various technologies. In that network, the obvious scalability of our system utilizing radioactive decay measurements is likely to play a significant role.

Acknowledgements
The authors wish to thank Catherine Ansbro, Eammon Ansbro, Jeff Sinard, John Voeller and Linda Yu for their helpful advice.

Here we propose to review the role played by the primary flux of suprathermal particles, electrons and protons, in the initiation of flares and CMEs. The existence of a critical level of the suprathermals is investigated, giving preliminary estimates for the relationship with the threshold conditions of the wave instabilities responsible for the particle reacceleration at these sites.

Flare-productive active regions are associated with subsurface flows with large values of kinetic helicity density. Kinetic helicity is related to mixing and turbulence of fluids. Reinard et al. 2010 have developed a parameter that captures the variation of kinetic helicity with depth and time, the so-called Normalized Helicity Gradient Variance (NHGV). This parameter increases 2-3 days before a flare occurs and the NHGV values for flaring and non-flaring active regions represent clearly separate populations. We derive subsurface flows from the surface to a depth of 16 Mm using GONG Dopplergrams analyzed with the ring-diagram technique. From the measured velocities, we calculate kinetic helicity density as a function of position on the solar disk. We will then calculate the NHGV parameter exploring different normalization schemes and depth ranges. We will calculate daily NHGV maps of the solar disk for different levels of magnetic activity. We will present the latest results.

We report on the observations of solar type IV bursts and their precursors, namely groups of type III bursts, in the frequency range of 8 - 33 MHz. The observational data obtained with the ground-based radio telescope UTR-2 (Kharkiv, Ukraine) during 2012 campaign completed by spacecrafts data are used. We consider the CMEs (Coronal Mass Ejections), which correspond to type IV radio bursts, and electrons, which generate groups of type III radio bursts observed before type IV bursts, simultaneously originate from the same active regions. The main properties (frequency drift rate, duration, flux) of type III and type IV bursts are analyzed. Several physical characteristics of the CME are estimated.

Almost 70 years after the discovery, the solar coronal heating puzzle is still one of the major challenges in astrophysics. A self-consistent coronal heating model must fulfil a lot of requirements imposed by observational facts. It is commonly accepted that many mechanisms contribute to the temperature of the solar corona.
However, determining which process dominates in what region, has proven to be very difficult. Many heating models have already been proposed for the coronal heating problem, and the two theories that currently stick out are the theory of wave heating and the theory of magnetic reconnection. Although a lot of research is still needed on these two theories, it is very unlikely that they will be able to fully explain the problem since they rely on the continuum or fluid approximation (Magnetohydrodynamics, MHD). Hence, these models cannot really explain coronal heating completely because i) it is clear that the actual heating takes place at length scales much smaller than those on which the (macroscopic) MHD model is justified; and ii) it is obvious that the observed discrepancy between ion and electron temperatures in the corona, as well as iii) the observed large temperature anisotropy in the inner corona and iv) the observed preferential heating of the heavier ions are beyond the (single!) fluid model.
The alternative that is explored here is based on the kinetic theory of drift waves. Assuming drift waves as the cause of the coronal heating implicitly states that the energy source of the heating is located in the corona itself, viz. in the ubiquitous density gradients present there. Though drift waves are studied intensively in the context of nuclear fusion research, not much studies are available on drift waves in solar context. Hence, it is important to look for methods to confirm the presence of drift waves in e.g. the solar corona. Here, the Vlasov theory will be used to study the presence of the lower hybrid drift instability in plasmas with a density gradient perpendicular to the magnetic field. A new approach for this study of Vlasov theory will be used. The main idea consists of expanding the distribution function in a series of Hermite polynomials in velocity space. This approach will facilitate the interpretation of the results obtained, as every expansion term has a specific physical meaning. The study has bot numerical and analytical components in an attempt to quantify the presence of drift waves in the solar corona and their contribution to the heating problem.

We examine the emission of near-relativistic (NR; >50 keV) electrons during solar flares. NR electron events observed in the heliosphere by spacecraft such as ACE, Wind or STEREO provide us with crucial information to unravel the release history of solar energetic particles (SEPs).

We use inversion methods developed within the EU/FP7 SEPServer project to extract, from directional intensities observed near 1 AU, the electron release history for a sample of flare-accelerated events. For each event, we compare the inferred number of released electrons with the intensity of the associated soft X-ray (SXR) flare. We find a close proportionality between the intensity of the SXR flare and the number of released electrons; correlation coefficient 0.949.

This empirical law is combined with an interplanetary transport model database of simulation results, publicily available through the SEPServer website, to predict the expected time of arrival and the in-situ electron event peak intensities for different flare sizes. The tool is used to estimate the maximum amount of NR-electron radiation that solar flares might produce near 1 AU, for an early warning system of flare-related SEP events.

Flare-prediction parameters roughly fall into three distinct classes: proxy parameters, that mostly depend on or quantify the morphology of the source active region, complexity parameters, that explore the fractal/multifractal nature of the region, and fundamental physical parameters, calculated mainly via simplifying assumptions. Recently, it has been suggested that complexity metrics are not expected, and seem to be unable, to distinguish between flaring and non-flaring solar active regions. If this result is further established, flare prediction must rely on proxy quantities and physical parameters, namely the free magnetic energy and, for several models, the magnetic or kinetic helicity of the regions. We examine the pros and cons of each class of predictors, considering ease and speed of the calculations and discussing their predictive ability. Relevant examples are also given. We conclude that the local nature of the flare-triggering process, plus the lack of information to unambiguously calculate fundamental physical quantities makes adequately-defined proxy quantities more efficient for continuous, routine prediction. However, (1) cutting-edge physical understanding is required to produce a suitable proxy, and (2) fundamental physical parameters may enclose a short-term predictive ability for major flares, provided that their timeseries on a candidate flaring region are properly treated for various unrelated effects.

An online solar-flare forecasting utility is currently implemented in the premises of the Academy of Athens. Upon completion, A-EFFort will be used as an ESA federated service in the framework of the Agency's Space Situational Awareness (SSA) Programme. The core of the A-EFFort utility comprises two pattern-recognition and computational-analysis techniques: a proven active-region identification algorithm (ARIA), used to automatically extract solar active regions from full-disk line-of-sight magnetograms acquired by the Helioseismic and Magnetic Imager (HMI) onboard the Solar Dynamics Observatory (SDO), and an automatic calculation of the effective connected magnetic field strength (Beff) for each of the identified active-region patches. From the derived Beff-values and existing statistics we will provide and validate flare probabilities assigned to a given forecast window for each active region separately and the earthward solar hemisphere as a whole. In its nominal operation, A-EFFort will deliver routine 24-hour forecasts for X-, M-, and C-class flares updated every three hours. During periods of exceptional solar activity and enhanced likelihoods of major or great solar flares, automated electronic messages will be sent to a list of subscribed service users. Plans for service updates and maintenance are in place for a number of years after the conclusion of the project's implementation, aiming to establish A-EFFort as a core tool of the ESA/SSA method arsenal.

Two workshops have been held recently, in 2009 and 2013, to begin systematic comparisons of methods for forecasting solar flares. The first, also known as the "All-Clear Forecasting Workshop" was held jointly with NASA/Space Radiation Analysis Group and NOAA/Space Weather Prediction Center, had a focus on predicting "All-Clear periods" from the standpoint of major flares and solar energetic particle events. The second more recent workshop, held at NorthWest Research Associates in Boulder, CO, USA, focused on using and exploiting the recent data from the NASA/Solar Dynamics Observatory (SDO) Helioseismic and Magnetic Imager (HMI), particularly the vector magnetic field time-series now available. For both workshops, many researchers active in flare-forecasting research participated, and diverse methods were represented in terms of both how the methods characterize the Sun, and the statistical approaches used to create a forecast. We describe here the goals of both workshops, the participating methods, and the approaches developed for allowing standardized, testable comparisons between methods.
(Funding for the workshops and the data analysis was provided by NASA/Living with a Star contract NNH09CE72C and NASA/Guest Investigator contract NNH12CG10C.)

We use HMI/SDO line-of-sight magnetograms to analyze the spatial distribution of the magnetic helicity flux in active region (AR) NOAA 11283. Applying an algorithm to identify areas characterized by similar magnetic helicity flux in sign, we determine how fragmented was the helicity flux in the active region. The results clearly show that areas characterized by the lower difference in the number of patches with helicity flux of opposite sign are the most favourable for eruptive events occurrence.
We also show that flares and CMEs occurred in this AR are linked to the interaction between magnetic systems characterized by opposite sign of magnetic helicity flux.

The research leading to these results has received funding from the European Commissions Seventh Framework Programme under the grant agreement no. 284461 (eHEROES project).

Recent years have seen a resurgence in the field of solar flare prediction. Most of these methods aim at evaluating a flare probability in the next 24h based on the current or past status of an active region.
In this project sequences of magnetogram and continuum images are used to distinguish active regions with strong flaring activity.

A homogeneous dataset of magnetogram and continuum images of active regions in their growth phase is produced.
These images are summarized into various scalar predictor variables, which are used as the input for the supervised classification methods.
These methods take into account the time evolution of the active regions through lagged values of the predictors.
The performance of various supervised classification algorithms as well as the predictive power of each predictor variable are assessed.
Special care is taken to handle the imbalance between the number of active regions with and without strong flaring activity.

In our poster we will discuss preliminary results from this project.

Within the previous solar cycle 23, SREM units onboard ESA's INTEGRAL and Rosetta spacecraft detected several tens of Solar Energetic Particle Events (SEPEs) and accurately pinpointed their onset, rise, and decay times. We have undertaken a detailed study to determine the solar sources and the subsequent interplanetary coronal mass ejections (ICMEs) that gave rise to these events, as well as the timing of SEPEs with regard to the onset of possible geomagnetic activity triggered by these ICMEs. We find that virtually all SREM SEPEs can be associated with CME-driven shocks. Moreover, for a number of well-studied INTEGRAL/SREM SEPEs we see an association between the SEPE peak and the shock passage at L1, subject to the heliographic location of the source solar active region. Shortly after the SEPE peak (typically within a few hours), the ICME-driven modulation of the magnetosphere kicks in, often associated with a dip of the Dst index, indicating storm conditions in geospace. In essence we find that SREM SEPEs can be seamlessly fit into a coherent and consistent heliophysical interpretation of solar eruptions all the way from Sun to Earth. Their contribution to space-weather forecasting may be significant and warrants additional investigation.

For more than three years, the radiometer LYRA on the ESA micro-satellite PROBA2 has been observing the Sun in the ultraviolet and gained a considerable data base. On its websites, LYRA presents not only EUV and SXR time series in near real-time, but also information on flare parameters and long-term irradiance and sunspot levels. It will be discussed how these parameters correlate, and whether it is possible to aid space weather forecast with these statistical data, especially for the prediction of expected flare strength.

Several models have been proposed to explain the initiation of CMEs. However, which model better explains the different aspects of the initiation process and the early evolution of the CMEs is a subject of ongoing discussion.

Using the vantage points of the STEREO spacecraft we reconstruct the three-dimensional position of the erupting structure, while the magnetic field evolution is used in order to identify eventual persistent photospheric plasma flows and/or the emergence of new magnetic flux. Finally, potential field extrapolations are used to determine the stability properties of the relevant active region and to disentangle which is the driver of the investigated eruption.

We present the Space Weather and Ultraviolet Solar Variability (SWUSV) Microsatellite Mission proposed to ESA and CNES for a better/earlier prediction of major flares and CMEs and their geoeffectiveness. Mission is an evolution of PROBA-2 and PICARD based on high cadence imaging of the solar disk in Lyman Alpha coupled to H-Alpha from ground network. Objectives, payload and mission, based on a MYRIADE Evolution platform, will be described. Mission is under consideration by CNES in its prospective exercise and could be in service as early as 2018.

Solar energetic particles (SEPs) are an important space weather phenomena. SEPs can damage satellite instrumentation, and they can be hazardous for crews of Low Earth Orbit spacecraft and the International Space Station, especially when engaged in extravehicular activity. They also represent a significant risk to crews of future manned lunar or interplanetary missions. SEPs are one of the most difficult phenomena to predict: their generation and propagation spans very different plasma regimes and large regions of the heliosphere. Successful prediction will ultimately involve many elements: prediction of flares/CMEs prior to eruption, simulating CMEs and associated shocks, and modeling SEP acceleration and propagation. In this paper, we describe an approach to this problem that combines empirical and physics-based models. We discuss how, given a prediction for a flare/CME, CMEs can be modeled in the corona and inner heliosphere. We show initial results for SEP fluxes generated from coupled CME/SEP modeling.

Research supported by NSF and NASA.

We report on the re-implementation of the Wheatland (2004, 2005) solar flare prediction model as an active online space weather service for the daily prediction of GEOS X-ray flares. The model uses a Bayesian approach to predict solar flare activity based on a database of event statistics from GOES measurements spanning over almost 40 years. This approach has the advantage of being objective without the need for users to make ad hoc assumptions. It uses only the past history soft X-ray flare events, and the statistical rules of solar flare occurrence.

We use the model to retrospectively validate M- and X-class event occurrence rates for periods of past solar activity, and also illustrate the results of the online prediction service running daily flare predictions for the present phase of this solar cycle. We also investigate the use of Bayesian analysis to predict occurrence rates for solar energetic particle events, given the similarity of the statistical properties of events for flares and energetic particles.

Wheatland, M.S., A Bayesian approach to solar flare prediction, Astrophys. J., 609, 2004.
Wheatland, M.S., A statistical solar flare forecast method, Space Weather, 3, 2005.

In this study we focus on the analysis of the early phase of flares initiation associated with the CMEs. The aim of this work is to determine kinematical properties and important structural components of these solar eruptions. We analyzed multi-thermal observations of eruptive plasmoids associated with coronal mass ejections and/or flares using the data recorded by the Atmospheric Imaging Assembly (AIA) on board the Solar Dynamics Observatory (SDO). The six coronal EUV channels, high spatial and temporal resolution and large field of view provide us the opportunity for verifying the existing models of solar eruptions. In addition, in order to observe the further propagation of these solar eruptive phenomena in the heliosphere we also use data taken by Sun Earth Connection Coronal and Heliospheric Investigation (SECCHI) on board the Solar TErrestrial RElations Observatory (STEREO).

The aim of this study is to identify the solar origin (flares vs. CMEs) of the solar energetic particles observed during solar cycle 23. For each SEP event, a flare and CME candidate was found by searching into specific temporal and spatial window. A pre-selection of SEPs associated with X and M-class flares was done and thus 180 events in total were analyzed. In order to argue in favor of flare, CME or mixed contribution to the escaping particle fluxes, we utilized all available dynamic radio spectral and single frequency radio (RSTN) records. Our analysis shows that about half of all SEP events are accompanied by high-frequency radio emission signatures that argues in favor of flare acceleration. Less than a third of the events show radio signatures from the high corona only and/or confinement.

It was shown before that long-term fluctuations of geomagnetic field can be considered as prognostic parameter for solar proton flares. This paper is devoted to the comparison of pre-flare behavior of long-term spectral components in the spectrum of the solar ultraviolet radiation and the horizontal component of the geomagnetic field. This study is based on wavelet spectra of solar ultraviolet emission during the period preceding solar proton flares. It is found the congruence in the behavior of spectral components with periods of 30-60 minutes in the ground-based measurements and in UV emission for 3-1 days before the proton flare.