@article{RinaldiFormisanoKappeletal.2019, author = {Rinaldi, G. and Formisano, M. and Kappel, David and Capaccioni, F. and Bockelee-Morvan, D. and Cheng, Y-C and Vincent, J-B and Deshapriya, P. and Arnold, G. and Capria, M. T. and Ciarniello, M. and De Sanctis, M. C. and Doose, L. and Erard, S. and Federico, C. and Filacchione, G. and Fink, U. and Leyrat, C. and Longobardo, A. and Magni, G. and Mighorini, A. and Mottola, S. and Naletto, G. and Raponi, A. and Taylor, F. and Tosi, F. and Tozzi, G. P. and Salatti, M.}, title = {Analysis of night-side dust activity on comet 67P observed by VIRTIS-M}, series = {Astronomy and astrophysics : an international weekly journal}, volume = {630}, journal = {Astronomy and astrophysics : an international weekly journal}, publisher = {EDP Sciences}, address = {Les Ulis}, issn = {1432-0746}, doi = {10.1051/0004-6361/201834907}, pages = {16}, year = {2019}, abstract = {On 2015 July 18, near perihelion at a heliocentric distance of 1.28 au, the Visible InfraRed Thermal Imaging Spectrometer (VIRTIS-M) on board the Rosetta spacecraft had the opportunity of observing dust activity in the inner coma with a view of the night side (shadowed side) of comet 67P/Churyumov-Gerasimenko. At the time of the measurements we present here, we observe a dust plume that originates on the far side of the nucleus. We are able to identify the approximate location of its source at the boundary between the Hapi and Anuket regions, and we find that it has been in darkness for some hours before the observation. Assuming that this time span is equal to the conductive time scale, we obtain a thermal inertia in the range 25-36 W K-1 m(-2) s(-1/2). These thermal inertia values can be used to verify with a 3D finite-element method (REM) numerical code whether the surface and subsurface temperatures agree with the values found in the literature. We explored three different configurations: (1) a layer of water ice mixed with dust beneath a dust mantle of 5 mm with thermal inertia of 36 J m(-2) K-1 S-0.5 ; (2) the same structure, but with thermal inertia of 100 J m(-2) K-1 S-0.5; (3) an ice-dust mixture that is directly exposed. Of these three configurations, the first seems to be the most reasonable, both for the low thermal inertia and for the agreement with the surface and subsurface temperatures that have been found for the comet 67P/Churyumov-Gerasimenko. The spectral properties of the plume show that the visible dust color ranged from 16 +/- 4.8\%/100 nm to 13 +/- 2.6\%/100 nm, indicating that this plume has no detectable color gradient. The morphology of the plume can be classified as a narrow jet that has an estimated total ejected mass of between 6 and 19 tons when we assume size distribution indices between -2.5 and -3.}, language = {en} } @phdthesis{Zoeller1999, author = {Z{\"o}ller, Gert}, title = {Analyse raumzeitlicher Muster in Erdbebendaten}, url = {http://nbn-resolving.de/urn:nbn:de:kobv:517-0000122}, school = {Universit{\"a}t Potsdam}, year = {1999}, abstract = {Die vorliegende Arbeit besch{\"a}ftigt sich mit der Charakterisierung von Seismizit{\"a}t anhand von Erdbebenkatalogen. Es werden neue Verfahren der Datenanalyse entwickelt, die Aufschluss dar{\"u}ber geben sollen, ob der seismischen Dynamik ein stochastischer oder ein deterministischer Prozess zugrunde liegt und was daraus f{\"u}r die Vorhersagbarkeit starker Erdbeben folgt. Es wird gezeigt, dass seismisch aktive Regionen h{\"a}ufig durch nichtlinearen Determinismus gekennzeichent sind. Dies schließt zumindest die M{\"o}glichkeit einer Kurzzeitvorhersage ein. Das Auftreten seismischer Ruhe wird h{\"a}ufig als Vorl{\"a}uferphaenomen f{\"u}r starke Erdbeben gedeutet. Es wird eine neue Methode pr{\"a}sentiert, die eine systematische raumzeitliche Kartierung seismischer Ruhephasen erm{\"o}glicht. Die statistische Signifikanz wird mit Hilfe des Konzeptes der Ersatzdaten bestimmt. Als Resultat erh{\"a}lt man deutliche Korrelationen zwischen seismischen Ruheperioden und starken Erdbeben. Gleichwohl ist die Signifikanz daf{\"u}r nicht hoch genug, um eine Vorhersage im Sinne einer Aussage {\"u}ber den Ort, die Zeit und die St{\"a}rke eines zu erwartenden Hauptbebens zu erm{\"o}glichen.}, language = {en} } @article{HeWangHeetal.2022, author = {He, Yushuang and Wang, Feipeng and He, Li and Wang, Qiang and Li, Jian and Qian, Yihua and Gerhard, Reimund and Plath, Ronald}, title = {An insight Into the role of Nano-Alumina on DC Flashover-Resistance and surface charge variation of Epoxy Nanocomposites}, series = {IEEE transactions on dielectrics and electrical insulation}, volume = {29}, journal = {IEEE transactions on dielectrics and electrical insulation}, number = {3}, publisher = {Inst. of Electr. and Electronics Engineers}, address = {Piscataway}, issn = {1070-9878}, doi = {10.1109/TDEI.2022.3173510}, pages = {1022 -- 1029}, year = {2022}, abstract = {The addition of nano-Al2O3 has been shown to enhance the breakdown voltage of epoxy resin, but its flashover results appeared with disputation. This work concentrates on the surface charge variation and dc flashover performance of epoxy resin with nano-Al2O3 doping. The dispersion of nano-Al2O3 in epoxy is characterized by scanning electron microscopy (SEM) and atomic force microscopy (AFM). The dc flashover voltages of samples under either positive or negative polarity are measured with a finger-electrode system, and the surface charge variations before and after flashovers were identified from the surface potential mapping. The results evidence that nano-Al2O3 would lead to a 16.9\% voltage drop for the negative flashovers and a 6.8\% drop for positive cases. It is found that one-time flashover clears most of the accumulated surface charges, regardless of positive or negative. As a result, the ground electrode is neighbored by an equipotential zone enclosed with low-density heterocharges. The equipotential zone tends to be broadened after 20 flashovers. The nano-Al2O3 is noticed as beneficial to downsize the equipotential zone due to its capability on charge migration, which is reasonable to maintain flashover voltage at a high level after multiple flashovers. Hence, nano-Al2O3 plays a significant role in improving epoxy with high resistance to multiple flashovers.}, language = {en} } @article{AbdallaAdamAharonianetal.2020, author = {Abdalla, H. and Adam, R. and Aharonian, Felix A. and Benkhali, F. Ait and Ang{\"u}ner, Ekrem Oǧuzhan and Arcaro, C. and Armand, C. and Armstrong, T. and Ashkar, H. and Backes, M. and Baghmanyan, V. and Martins, V. Barbosa and Barnacka, A. and Barnard, M. and Becherini, Y. and Berge, D. and Bernlohr, K. and Bi, B. and Bottcher, M. and Boisson, C. and Bolmont, J. and de Lavergne, M. de Bony and Bordas, Pol and Breuhaus, M. and Brun, F. and Brun, P. and Bryan, M. and Buchele, M. and Bulik, T. and Bylund, T. and Caroff, S. and Carosi, A. and Casanova, Sabrina and Chand, T. and Chandra, S. and Chen, A. and Cotter, G. and Curylo, M. and Mbarubucyeye, J. Damascene and Davids, I. D. and Davies, J. and Deil, C. and Devin, J. and deWilt, P. and Dirson, L. and Djannati-Atai, A. and Dmytriiev, A. and Donath, A. and Doroshenko, V. and Duffy, C. and Dyks, J. and Egberts, Kathrin and Eichhorn, F. and Einecke, S. and Emery, G. and Ernenwein, J. -P. and Feijen, K. and Fegan, S. and Fiasson, A. and de Clairfontaine, G. Fichet and Fontaine, G. and Funk, S. and Fussling, Matthias and Gabici, S. and Gallant, Y. A. and Giavitto, G. and Giunti, L. and Glawion, D. and Glicenstein, J. F. and Gottschall, D. and Grondin, M. -H. and Hahn, J. and Haupt, M. and Hermann, G. and Hinton, J. A. and Hofmann, W. and Hoischen, Clemens and Holch, T. L. and Holler, M. and Horbe, M. and Horns, D. and Huber, D. and Jamrozy, M. and Jankowsky, D. and Jankowsky, F. and Jardin-Blicq, A. and Joshi, V. and Jung-Richardt, I. and Kasai, E. and Kastendieck, M. A. and Katarzynski, K. and Katz, U. and Khangulyan, D. and Khelifi, B. and Klepser, S. and Kluzniak, W. and Komin, Nu. and Konno, R. and Kosack, K. and Kostunin, D. and Kreter, M. and Lamanna, G. and Lemiere, A. and Lemoine-Goumard, M. and Lenain, J. -P. and Levy, C. and Lohse, T. and Lypova, I. and Mackey, J. and Majumdar, J. and Malyshev, D. and Malyshev, D. and Marandon, V. and Marchegiani, P. and Marcowith, Alexandre and Mares, A. and Marti-Devesa, G. and Marx, R. and Maurin, G. and Meintjes, P. J. and Meyer, M. and Mitchell, A. and Moderski, R. and Mohamed, M. and Mohrmann, L. and Montanari, A. and Moore, C. and Morris, P. and Moulin, Emmanuel and Muller, J. and Murach, T. and Nakashima, K. and Nayerhoda, A. and de Naurois, M. and Ndiyavala, H. and Niederwanger, F. and Niemiec, J. and Oakes, L. and O'Brien, Patrick and Odaka, H. and Ohm, S. and Olivera-Nieto, L. and Wilhelmi, E. de Ona and Ostrowski, M. and Oya, I. and Panter, M. and Panny, S. and Parsons, R. D. and Peron, G. and Peyaud, B. and Piel, Q. and Pita, S. and Poireau, V. and Noel, A. Priyana and Prokhorov, D. A. and Prokoph, H. and Puhlhofer, G. and Punch, M. and Quirrenbach, A. and Raab, S. and Rauth, R. and Reichherzer, P. and Reimer, A. and Reimer, O. and Remy, Q. and Renaud, M. and Rieger, F. and Rinchiuso, L. and Romoli, C. and Rowell, G. and Rudak, B. and Ruiz-Velasco, E. and Sahakian, V. and Sailer, S. and Sanchez, D. A. and Santangelo, Andrea and Sasaki, M. and Scalici, M. and Schussler, F. and Schutte, H. M. and Schwanke, U. and Schwemmer, S. and Seglar-Arroyo, M. and Senniappan, M. and Seyffert, A. S. and Shafi, N. and Shiningayamwe, K. and Simoni, R. and Sinha, A. and Sol, H. and Specovius, A. and Spencer, S. and Spir-Jacob, M. and Stawarz, L. and Sun, L. and Steenkamp, R. and Stegmann, C. and Steinmassl, S. and Steppa, C. and Takahashi, T. and Tavernier, T. and Taylor, A. M. and Terrier, R. and Tiziani, D. and Tluczykont, M. and Tomankova, L. and Trichard, C. and Tsirou, M. and Tuffs, R. and Uchiyama, Y. and van der Walt, D. J. and van Eldik, C. and van Rensburg, C. and van Soelen, B. and Vasileiadis, G. and Veh, J. and Venter, C. and Vincent, P. and Vink, J. and Volk, H. J. and Vuillaume, T. and Wadiasingh, Z. and Wagner, S. J. and Watson, J. and Werner, F. and White, R. and Wierzcholska, A. and Wong, Yu Wun and Yusafzai, A. and Zacharias, M. and Zanin, R. and Zargaryan, D. and Zdziarski, A. A. and Zech, Alraune and Zhu, S. J. and Ziegler, A. and Zorn, J. and Zouari, S. and Zywucka, N.}, title = {An extreme particle accelerator in the Galactic plane}, series = {Astronomy and astrophysics : an international weekly journal}, volume = {644}, journal = {Astronomy and astrophysics : an international weekly journal}, publisher = {EDP Sciences}, address = {Les Ulis}, organization = {HESS Collaboration}, issn = {0004-6361}, doi = {10.1051/0004-6361/202038851}, pages = {8}, year = {2020}, abstract = {The unidentified very-high-energy (VHE; E > 0.1 TeV) gamma -ray source, HESS J1826-130, was discovered with the High Energy Stereoscopic System (HESS) in the Galactic plane. The analysis of 215 h of HESS data has revealed a steady gamma -ray flux from HESS J1826-130, which appears extended with a half-width of 0.21 degrees +/- 0.02
(stat)degrees
stat degrees +/- 0.05
(sys)degrees sys degrees . The source spectrum is best fit with either a power-law function with a spectral index Gamma = 1.78 +/- 0.10(stat) +/- 0.20(sys) and an exponential cut-off at 15.2
(+5.5)(-3.2) -3.2+5.5 TeV, or a broken power-law with Gamma (1) = 1.96 +/- 0.06(stat) +/- 0.20(sys), Gamma (2) = 3.59 +/- 0.69(stat) +/- 0.20(sys) for energies below and above E-br = 11.2 +/- 2.7 TeV, respectively. The VHE flux from HESS J1826-130 is contaminated by the extended emission of the bright, nearby pulsar wind nebula, HESS J1825-137, particularly at the low end of the energy spectrum. Leptonic scenarios for the origin of HESS J1826-130 VHE emission related to PSR J1826-1256 are confronted by our spectral and morphological analysis. In a hadronic framework, taking into account the properties of dense gas regions surrounding HESS J1826-130, the source spectrum would imply an astrophysical object capable of accelerating the parent particle population up to greater than or similar to 200 TeV. Our results are also discussed in a multiwavelength context, accounting for both the presence of nearby supernova remnants, molecular clouds, and counterparts detected in radio, X-rays, and TeV energies.}, language = {en} } @article{NakoudiStachlewskaRitter2021, author = {Nakoudi, Konstantina and Stachlewska, Iwona S. and Ritter, Christoph}, title = {An extended lidar-based cirrus cloud retrieval scheme}, series = {Optics express : the international electronic journal of optics / Optica}, volume = {29}, journal = {Optics express : the international electronic journal of optics / Optica}, number = {6}, publisher = {Optical Society of America}, address = {Washington}, issn = {1094-4087}, doi = {10.1364/OE.414770}, pages = {8553 -- 8580}, year = {2021}, abstract = {Accurate and precise characterization of cirrus cloud geometrical and optical properties is essential for better constraining their radiative footprint. A lidar-based retrieval scheme is proposed here, with its performance assessed on fine spatio-temporal observations over the Arctic site of Ny-Alesund, Svalbard. Two contributions related to cirrus geometrical (dynamic Wavelet Covariance Transform (WCT)) and optical properties (constrained Klett) are reported. The dynamic WCT rendered cirrus detection more robust, especially for thin cirrus layers that frequently remained undetected by the classical WCT method. Regarding optical characterization, we developed an iterative scheme for determining the cirrus lidar ratio (LRci) that is a crucial parameter for aerosol - cloud discrimination. Building upon the Klett-Fernald method, the LRci was constrained by an additional reference value. In established methods, such as the double-ended Klett, an aerosol-free reference value is applied. In the proposed constrained Klett, however, the reference value was approximated from cloud-free or low cloud optical depth (COD up to 0.2) profiles and proved to agree with independent Raman estimates. For optically thin cirrus, the constrained Klett inherent uncertainties reached 50\% (60-74\%) in terms of COD (LRci). However, for opaque cirrus COD (LRci) uncertainties were lower than 10\% (15\%). The detection method discrepancies (dynamic versus static WCT) had a higher impact on the optical properties of low COD layers (up to 90\%) compared to optically thicker ones (less than 10\%). The constrained Klett presented high agreement with two established retrievals. For an exemplary cirrus cloud, the constrained Klett estimated the COD355 (LRci355) at 0.28 +/- 0.17 (29 +/- 4 sr), the double-ended Klett at 0.27 +/- 0.15 (32 +/- 4 sr) and the Raman retrievals at 0.22 +/- 0.12 (26 +/- 11 sr). Our approach to determine the necessary reference value can also be applied in established methods and increase their accuracy. In contrast, the classical aerosol-free assumption led to 44 sr LRci overestimation in optically thin layers and 2-8 sr in thicker ones. The multiple scattering effect was corrected using Eloranta (1998) and accounted for 50-60\% extinction underestimation near the cloud base and 20-30\% within the cirrus layers.}, language = {en} } @article{Grebenkov2022, author = {Grebenkov, Denis S.}, title = {An encounter-based approach for restricted diffusion with a gradient drift}, series = {Journal of physics : A, Mathematical and theoretical}, volume = {55}, journal = {Journal of physics : A, Mathematical and theoretical}, number = {4}, publisher = {IOP Publ. Ltd.}, address = {Bristol}, issn = {1751-8113}, doi = {10.1088/1751-8121/ac411a}, pages = {34}, year = {2022}, abstract = {We develop an encounter-based approach for describing restricted diffusion with a gradient drift toward a partially reactive boundary. For this purpose, we introduce an extension of the Dirichlet-to-Neumann operator and use its eigenbasis to derive a spectral decomposition for the full propagator, i.e. the joint probability density function for the particle position and its boundary local time. This is the central quantity that determines various characteristics of diffusion-influenced reactions such as conventional propagators, survival probability, first-passage time distribution, boundary local time distribution, and reaction rate. As an illustration, we investigate the impact of a constant drift onto the boundary local time for restricted diffusion on an interval. More generally, this approach accesses how external forces may influence the statistics of encounters of a diffusing particle with the reactive boundary.}, language = {en} } @article{SmirnovShpritsAllisonetal.2022, author = {Smirnov, Artem and Shprits, Yuri and Allison, Hayley and Aseev, Nikita and Drozdov, Alexander and Kollmann, Peter and Wang, Dedong and Saikin, Anthony}, title = {An empirical model of the equatorial electron pitch angle distributions in earth's outer radiation belt}, series = {Space Weather: the International Journal of Research and Applications}, volume = {20}, journal = {Space Weather: the International Journal of Research and Applications}, number = {9}, publisher = {American Geophysical Union}, address = {Washington, DC}, issn = {1542-7390}, doi = {10.1029/2022SW003053}, pages = {17}, year = {2022}, abstract = {In this study, we present an empirical model of the equatorial electron pitch angle distributions (PADs) in the outer radiation belt based on the full data set collected by the Magnetic Electron Ion Spectrometer (MagEIS) instrument onboard the Van Allen Probes in 2012-2019. The PADs are fitted with a combination of the first, third and fifth sine harmonics. The resulting equation resolves all PAD types found in the outer radiation belt (pancake, flat-top, butterfly and cap PADs) and can be analytically integrated to derive omnidirectional flux. We introduce a two-step modeling procedure that for the first time ensures a continuous dependence on L, magnetic local time and activity, parametrized by the solar wind dynamic pressure. We propose two methods to reconstruct equatorial electron flux using the model. The first approach requires two uni-directional flux observations and is applicable to low-PA data. The second method can be used to reconstruct the full equatorial PADs from a single uni- or omnidirectional measurement at off-equatorial latitudes. The model can be used for converting the long-term data sets of electron fluxes to phase space density in terms of adiabatic invariants, for physics-based modeling in the form of boundary conditions, and for data assimilation purposes.}, language = {en} } @phdthesis{Bastian2023, author = {Bastian, Martin}, title = {An emergent machine learning approach for seasonal cyclone activity forecasts}, school = {Universit{\"a}t Potsdam}, pages = {135}, year = {2023}, abstract = {Seasonal forecasts are of great interest in many areas. Knowing the amount of precipitation for the upcoming season in regions of water scarcity would facilitate a better water management. If farmers knew the weather conditions of the upcoming summer at sowing time, they could select those cereal species that are best adapted to these conditions. This would allow farmers to improve the harvest and potentially even reduce the amount of pesticides used. However, the undoubted advantages of seasonal forecasts are often opposed by their high degree of uncertainty. The great challenge of generating seasonal forecasts with lead times of several months mainly originates from the chaotic nature of the earth system. In a chaotic system, even tiny differences in the initial conditions can lead to strong deviations in the system's state in the long run. In this dissertation we propose an emergent machine learning approach for seasonal forecasting, called the AnlgModel. The AnlgModel combines the analogue method with myopic feature selection and bootstrapping. To benchmark the abilities of the AnlgModel we apply it to seasonal cyclone activity forecasts in the North Atlantic and Northwest Pacific. The AnlgModel demonstrates competitive hindcast skills with two operational forecasts and even outperforms these for long lead times. In the second chapter we comprehend the forecasting strategy of the Anlg-Model. We thereby analyse the analogue selection process for the 2017 North Atlantic and the 2018 Northwest Pacific seasonal cyclone activity. The analysis shows that those climate indices which are known to influence the seasonal cyclone activity, such as the Ni{\~n}o 3.4 SST, are correctly represented among the selected analogues. Furthermore the selected analogues reflect large-scale climate patterns that were identified by expert reports as being determinative for these particular seasons. In the third chapter we analyse the features that are used by the AnlgModel for its predictions. We therefore inspect the feature relevance (FR). The FR patterns learned by the AnlgModel show a high congruence with the predictor regions used by the operational forecasts. However, the AnlgModel also discovered new features, such as the SST anomaly in the Gulf of Guinea during November. This SST pattern exhibits a remarkably high predictive potential for the upcoming Atlantic hurricane activity. In the final chapter we investigate potential mechanisms, that link two of these regions with high feature relevance to the Atlantic hurricane activity. We mainly focus on ocean surface transport. The ocean surface flow paths are calculated using Lagrangian particle analysis. We demonstrate that the FR patterns in the region of the Canary islands do not correspond with ocean surface transport. It is instead likely that these FR patterns fingerprint a wind transport of latent heat. The second region to be studied is situated in the Gulf of Guinea. Our analysis shows that the FR patterns seen there do fingerprint ocean surface transport. However, our simulations also show that at least one other mechanism is involved in linking the Gulf of Guinea SST anomaly in November to the hurricane activity of the upcoming season. In this work the AnlgModel does not only demonstrate its outstanding forecast skills but also shows its capabilities as research tool for detecting oceanic and atmospheric mechanisms.}, language = {en} } @article{CiemerRehmKurthsetal.2020, author = {Ciemer, Catrin and Rehm, Lars and Kurths, J{\"u}rgen and Donner, Reik Volker and Winkelmann, Ricarda and Boers, Niklas}, title = {An early-warning indicator for Amazon droughts exclusively based on tropical Atlantic sea surface temperatures}, series = {Environmental Research Letters}, volume = {15}, journal = {Environmental Research Letters}, number = {9}, publisher = {IOP - Institute of Physics Publishing}, address = {Bristol}, pages = {10}, year = {2020}, abstract = {Droughts in tropical South America have an imminent and severe impact on the Amazon rainforest and affect the livelihoods of millions of people. Extremely dry conditions in Amazonia have been previously linked to sea surface temperature (SST) anomalies in the adjacent tropical oceans. Although the sources and impacts of such droughts have been widely studied, establishing reliable multi-year lead statistical forecasts of their occurrence is still an ongoing challenge. Here, we further investigate the relationship between SST and rainfall anomalies using a complex network approach. We identify four ocean regions which exhibit the strongest overall SST correlations with central Amazon rainfall, including two particularly prominent regions in the northern and southern tropical Atlantic. Based on the time-dependent correlation between SST anomalies in these two regions alone, we establish a new early-warning method for droughts in the central Amazon basin and demonstrate its robustness in hindcasting past major drought events with lead-times up to 18 months.}, language = {en} } @misc{CiemerRehmKurthsetal.2020, author = {Ciemer, Catrin and Rehm, Lars and Kurths, J{\"u}rgen and Donner, Reik Volker and Winkelmann, Ricarda and Boers, Niklas}, title = {An early-warning indicator for Amazon droughts exclusively based on tropical Atlantic sea surface temperatures}, series = {Postprints der Universit{\"a}t Potsdam : Mathematisch-Naturwissenschaftliche Reihe}, journal = {Postprints der Universit{\"a}t Potsdam : Mathematisch-Naturwissenschaftliche Reihe}, number = {9}, issn = {1866-8372}, doi = {10.25932/publishup-52586}, url = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-525863}, pages = {12}, year = {2020}, abstract = {Droughts in tropical South America have an imminent and severe impact on the Amazon rainforest and affect the livelihoods of millions of people. Extremely dry conditions in Amazonia have been previously linked to sea surface temperature (SST) anomalies in the adjacent tropical oceans. Although the sources and impacts of such droughts have been widely studied, establishing reliable multi-year lead statistical forecasts of their occurrence is still an ongoing challenge. Here, we further investigate the relationship between SST and rainfall anomalies using a complex network approach. We identify four ocean regions which exhibit the strongest overall SST correlations with central Amazon rainfall, including two particularly prominent regions in the northern and southern tropical Atlantic. Based on the time-dependent correlation between SST anomalies in these two regions alone, we establish a new early-warning method for droughts in the central Amazon basin and demonstrate its robustness in hindcasting past major drought events with lead-times up to 18 months.}, language = {en} }