Here, we present a high resolution marine El Niño flood record from Peru. A period of extreme drought without strong flooding occurred from A.D. 800–1250. Anomalous precipitation patterns characterized the entire Indo-Pacific ENSO domain, with dry events in the northern Arabian Sea and the mid-latitudes of both Americas, coinciding with wet periods in the Atlantic Cariaco Basin. The occurrence of contemporaneous moisture anomalies in other archives in the ENSO region highlights the role of El Niño strength in global climate evolution during the late Medieval period when temperature reconstructions show a rather heterogeneous pattern.
 A reconstructed time series of global temperature evolution during the last 2000 years [Mann and Jones, 2003] provides background for understanding and evaluating the magnitude and processes of ongoing global climate change. However, discrepancies exist between the Mann curve and alternative time series for the Medieval period. Most notably the global Mann curve has no temperature optimum, whereas the Esper et al.  reconstruction shows northern hemisphere temperatures almost as high as those of the 20th century. El Niño Southern Oscillation (ENSO) functions as a major heat engine and it influences large parts of the earth's surface (Figure 1). Therefore extreme and persistent anomalies of this climate system have a large potential to influence the mean climate of large regions. In this paper, we illustrate the presence of a major Holocene El Niño anomaly during late Medieval times (A.D. 800–1250) in a high-resolution flood record off the coast of Peru. We show that this precipitation anomaly also occurred in other high-resolution records throughout the ENSO domain [e.g., Abbot et al., 1997; Binford et al., 1997; De Putter et al., 1998; Fleitmann et al., 2003; Haug et al., 2001; Laird et al., 1996, 1998; Stine, 1994, 1998; Verschuren et al., 2000; von Rad et al., 1999].
 Interannual climate variability along and off coastal Peru is dominated by ENSO (Figure 1). In the hyperarid coastal deserts, heavy winter rainfalls only occur during times of El Niño warm water anomalies and are particularly intense during years of maximum El Niño strength [Philander, 1990]. Fine-grained lithics are eroded by intense, storm-induced runoff and are washed onto the shelf by local rivers. They are dispersed northward by the Peru Current and its southward undercurrents. Dust-transported particles only form a minor constituent of the outer shelf sediments [Scheidegger and Krissek, 1982]. Grain size variations along the core are small, and 90–95% of the lithic grains are in the <45 μm fraction. Coarser grains are mostly embedded in fecal pellets. High bioproductivity in surface waters and subsequent decay of organic matter results in strong oxygen-minimum conditions from 50–650 m water depth which, in the absence of bottom-dwelling organisms, favor the preservation of laminated diatomaceous and diatom bearing oozes. These sediments constitute the basis for reconstructing past hydroclimatic conditions.
2. Data and Methods
 Sediments for this study were recovered from a piston core (18.82 m) and its accompanying overlapping trigger core (0.91 m) from a water depth of 184 m. The sediments were obtained from a sheltered basin situated on the edge of the Peruvian shelf 80 km west of Lima/Peru (Figure 1; 12°03′S, 77°39.8′W) during cruise SONNE 147 [Kudrass et al., 2000]. The organic carbon fraction of sediment was used for AMS-14C radiocarbon analyses, as calcareous remains are rare in this area of high bioproductivity. The uppermost sediments were slightly disturbed during coring. Lead and caesium isotopes were therefore used as age control on the uppermost undisturbed sediments, which were found to date at A.D. 1920.
 The detailed chronology of the Holocene part of the core (down to 8 m depth) is based on 38 AMS 14C dates (Figure 2a). In the age calibration process we used a constant reservoir correction of 800 years for all dates (for details, see auxiliary materials). Therefore dating uncertainties due to changing reservoir ages [e.g., Staubwasser et al., 2002] have to be considered. The age of water may change, when the mixture of water from different sources occurs. This can be expected when oceanic circulation (e.g., upwelling intensity) varies.
 The mid-Holocene section (from 4900–7800 cal. BP) is represented by only 20 cm of sediment. Radiographs revealed several erosional unconformities in this middle portion of the core as has formerly been reported from similar time intervals in other cores of the region [Suess et al., 1987]. No hiatus was identified in the radiographs and large thin sections from the upper core portion down to 2.7 m depth, whose chronology of the last 2500 years is based on 8 AMS 14C dates and lead and caesium isotope profiles of the topmost sediments (see auxiliary materials).
 A proxy for the lithic sediment fraction can be found in reflectance spectra of the split cores. The spectra were acquired with a Gretag Spectrolino, which allows a sample resolution of 2 mm along the core. This small sample resolution results in a high temporal resolution of proxy data (Figure 2b). Sediment brightness was measured in 36 discrete wavelength intervals, each 10 nm wide between wavelengths of 380–730 nm. The spectra show photosynthesis pigment absorption bands [see Rein, 2003] and a systematic change of reflectance between 570 and 630 nm wavelength which is caused by the lithic fraction in the core (see auxiliary materials) [Rein, 2003]. A simple ratio of the reflectance at 570 nm over the reflectance at 630 nm provides a proxy for lithic concentrations.
3. Medieval ENSO Anomalies in the Tropical Eastern Pacific
 A major Holocene anomaly in flux of lithic components from the continent onto the shelf occurred during late Medieval period (Figure 2c). Lithic concentrations (Figure 2c) were very low for about 450 years during the Medieval climatic anomaly (MCA [Stine, 1994]) from A.D. 800 to 1250. All known terrestrial deposits of El Niño mega-floods [Magillian and Goldstein, 2001; Wells, 1990] precede or follow the medieval anomaly in our marine records and none of the El Niño mega-floods known from the continent date within the marine anomaly. The major post-anomaly peak (M/N in Figure 3) could be the result of the catastrophic Miraflores/Nyamlap flood around 1300 A.D [Kosok, 1965].
 From an Ecuadorian lake record where moderate to strong El Niño floods are recorded [Moy et al., 2002], a minimum of such events is reported during the upper Medieval period. Short coral records provide latest views on central Pacific El Niño activity. The oldest of five time windows covers part of the tenth century (A.D. 928–961) and Cobb and colleagues [Cobb et al., 2003] postulate significantly weaker El Niños than during the later time-windows (1149–1220, 1317–1464, 1635–1703 and 1886–1998).
4. Teleconnections to Anomalies in Remote Archives
 The climatic teleconnections of ENSO anomalies during the MCA [Stine, 1994] are clearly visible in other geoarchives. Varve thickness, corrected for turbidites and sporadic white-event layers, on the continental slope off Pakistan (SO-90-56 KA [von Rad et al., 1999]) is controlled predominantly by river discharge [Lückge et al., 2001, 2002; Staubwasser and Sirocko, 2002]. The catchment on the coast of Makran (Hingol River) [Lückge et al., 2001] receives a maximum of rainfall (240 mm) during boreal winter with incursion of cyclones from the eastern Mediterranean Sea [Roberts and Wright, 1993]. Indian summer monsoon rainfall is less than 100 mm in the western and less than 200 mm in the remote eastern part of the catchment [Rao, 1981]. Thus varve thickness data at Site SO-90_56KA are used as a proxy for winter-rainfall and not a proxy for the SW-monsoon.
 Although the time axes of the Arabian Sea and Peru records seem to be biased by up to 200 years within the MCA, the offset can be explained by dating errors arising from the varying age of surface water [e.g., Staubwasser et al., 2002]. We observe increasing bias of ages going into the anomaly and decreasing bias after the anomaly (Figure 3). Less are the maximum dating differences of about 100 years which separate the Peru MCA from a wet phase in the Cariaco Basin. Precipitation in this part of Venezuela varies with seasonal and long-term changes of the position of the ITCZ as well as with the state of ENSO, showing a tendency towards more rainfall during non- or weaker El Niño years [Haug et al., 2001]. Consistent with the apparent dearth of ENSO events documented elsewhere, the titanium precipitation record of the Cariaco Basin indicates wetter conditions during the late Medieval time.
 Extreme long-lasting droughts that peaked coincident with those in the Peru record around A.D. 1160, are reported from several archives in the western USA and Southern Patagonia [Stine, 1994]. Dry periods also occurred in the tropical Andes [Abbot et al., 1997; Binford et al., 1997], Oman [Fleitmann et al., 2003] and eastern Africa [De Putter et al., 1998; Verschuren et al., 2000]. Hints that these droughts are not only coinciding events but related to El Niño anomalies come from the high-resolution Moon Lake (North Dakota/USA) salinity record [Laird et al., 1996]. Salinity changes indicate drier and wetter periods whose timing is very similar to those of precipitation minima and maxima in Peru (Figure 3). The bias of the time axes between both chronologies seems to be less than 50 years.
 The similarities in precipitation proxies dating back to 2500 years before the present in tropical/subtropical archives from both the northern and southern hemisphere cannot be explained by coincidence, but demand a common explanation. ENSO anomalies seem to provide that explanation. Charles and colleagues [Charles et al., 1997] demonstrated that the proxies in coral records from the monsoon region in the western Indian and in the Pacific ENSO region were closely related during the last century. The interannual impact of ENSO variability in the entire Indo-Pacific domain is widely accepted [Hoerling and Kumar, 2000] and the impact of ENSO variability on global climate is well established [Cane and Clement, 1999]. Increasingly, data supports the notion that ENSO influences the tropical Atlantic [Latif and Grötzner, 2000] as well as the interdecadal and secular change of North Atlantic climate [Allan, 2000; Hoerling et al., 2001]. The occurrence of a Medieval climatic anomaly (A.D. 800–1250) with persistently weak El Niños may therefore assist the interpretation of some of the regional discrepancies in thermal reconstructions of Medieval times.
 We acknowledge two very constructive anonymous reviews and A. Vink. This work was supported by the German Bundesministerium für Bildung und Forschung (03G0147A, 03G0147C) and KIHZ (Klima in Historischen Zeiten, 01LG9911).