Atmosphere-Ocean

ISSN: 0705-5900 (Print) 1480-9214 (Online) Journal homepage: http://www.tandfonline.com/loi/tato20

Stained Glass and Climate Change: How are they Connected? C. T. Simmons & L. A. Mysak To cite this article: C. T. Simmons & L. A. Mysak (2012) Stained Glass and Climate Change: How are they Connected?, Atmosphere-Ocean, 50:2, 219-240, DOI: 10.1080/07055900.2012.667387 To link to this article: http://dx.doi.org/10.1080/07055900.2012.667387

Published online: 02 Apr 2012.

Submit your article to this journal

Article views: 255

View related articles

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tato20 Download by: [37.44.207.66]

Date: 25 January 2017, At: 06:17

Stained Glass and Climate Change: How are they Connected? C.T. Simmons* and L.A. Mysak Atmospheric and Oceanic Sciences, McGill University, Montréal, Quebec, Canada [Original manuscript received 18 April 2011; accepted 16 May 2011]

As expressions of regional architecture, sacred (Christian) Gothic structures often possess several adaptations to their prevailing climate regime. The late medieval (Gothic) period in northern Europe is also, according to the evidence presented here, marked by a transition from warm and sunny to cooler and cloudier conditions. It is within the context of this climate change that we consider one of the most important features in Gothic churches—interior daylighting—during the transitional period (the thirteenth to the end of the fifteenth centuries) between the Medieval Warm Period (MWP) and the Little Ice Age (LIA). This paper seeks to determine whether increasingly cloudy conditions over northern continental Europe, in part due to a shift in North Atlantic Oscillation (NAO) phase, may have influenced the use of more white glass in the fourteenth century. To the best of our knowledge, this is the first time an extensive daylighting dataset has been collected in medieval sacred interiors. From an analysis of these illuminance and luminance data collected in European churches and cathedrals, we find that high-translucency glazing is associated with limited lighting gains compared to full-colour programs under sunny conditions but substantial lighting improvements (up to an order of magnitude) for cloudy conditions. Additionally, we find that backlighting from direct sunlight produces significant obscuration of some of the iconographical glass when interiors are dominated by high-translucency glazing, further suggesting that these interiors are not ideal for sunny conditions.

ABSTRACT

[Traduit par la rédaction] Étant donné que les cathédrales gothiques (chrétiennes) sont des expressions de l’architecture régionale, plusieurs adaptations au climat de l’époque y ont souvent été apportées. À la fin du Moyen Âge (période du gothique), le climat chaud et ensoleillé en Europe du nord continentale a fait place à un climat plus froid et plus nuageux, d’après les preuves que nous présentons ici. C’est donc dans la perspective de ce changement climatique que nous nous penchons sur l’un des éléments les plus importants de l’architecture des églises gothique, l’éclairage naturel intérieur, durant la transition (du XIIIe siècle à la fin du XVe siècle) entre la période chaude médiévale (MWP) et le petit âge glaciaire (LIA). Dans le présent article, nous voulons notamment évaluer si l’utilisation de plus en plus fréquente du vitrail blanc au XIVe siècle s’expliquerait par les conditions plus nuageuses en Europe du nord continentale, attribuables en partie à un changement dans l’indice d‘oscillation nord-atlantique (NAO). À notre connaissance, c’est la première fois qu’une série élaborée de données a été recueillie sur l’éclairage à l’intérieur des cathédrales gothiques. L’analyse des données sur l’éclairement lumineux et la luminance lumineuse dans les églises et les cathédrales d’Europe nous permet de constater que le verre très translucide présente peu d’avantages comparativement au verre plein coloré dans des conditions ensoleillées, mais qu’il améliore considérablement l’éclairage dans des conditions nuageuses (jusqu’à 10 fois). De plus, nous constatons que l’éclairage en contre-jour produit par l’ensoleillement direct obscurcit une partie des pièces de verre ornées d’icônes lorsque le verre très translucide domine à l’intérieur, ce qui confirme une fois de plus qu’il ne s’agit pas d’un aménagement idéal pour les conditions ensoleillées. RÉSUMÉ

KEYWORDS climate change; stained glass; medieval art; cloudiness; interior daylighting; Renaissance; Little Ice Age; Medieval Warm Period

1 Introduction: The stained glass-climate connection Lighting, architecture and climate have always been inextricably linked in the design of church interiors. Light serves to communicate the aesthetics of the sacred architectural space and also acts as the primary illuminator of wall paintings, mosaics, and narrative capitals (column heads) in medieval churches. Daylight climatology, along with latitude, season and time of day, determines how much daylight is present and where in the sky the most light is available. As a result,

*

climate strongly governs how natural lighting is directed into the church. Because daylight and candlelight (the latter often constrained by economics) were the primary sources of interior illumination, medieval churches had to use natural lighting adequately with respect to the prevailing climate to attain both aesthetic and functional needs. Given the intricacy with which windows were designed and arranged, daylighting concerns were also clearly of utmost concern to Gothic architects and planners (Janson, 2001).

Corresponding author’s email: [email protected] ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

220 / C.T. Simmons and L.A. Mysak Window size provides an obvious constraint on the amount of lighting available for the interior. Early Christian and Romanesque churches in northern Europe built before 1150 possessed windows similar to those in Mediterranean churches (a discussion of technical terms and a labelled plan of a church interior are provided in Appendix A). Before the invention of sophisticated forms of buttressing, small window sizes were necessary in order to maintain the structural stability of the tall, thick walls of many major churches. Even so, a general increase in window size can be observed with increasing latitude during the Romanesque period (Moore, 1985), which reflects the latitudinal gradient of daylight distribution in Europe (see Fig. 1). Later, during the Gothic era (starting around 1150), external buttressing allowed walls in Gothic churches to become thinner and windows to expand across much of the wall space. Upon lifting the limitations in window sizes, light became more readily available for church interiors. However, the transition to the Gothic style and larger windows also occurred in conjunction with an important shift in the colour of window glazing, which did not always afford a dramatic increase in interior daylight (Grodecki, 1983). Instead of the bright, higher translucency glass (whites, yellows and light blues) that dominated many Romanesque glazed windows, Gothic apertures were often largely filled with richly coloured glass that restricted interior lighting (Grodecki, 1983). The relative transmissivity differences between windows from different eras has been recently studied in Simmons and Mysak (2010) and quantitatively confirms Grodecki’s observations on the increasing colour saturation of stained glass in the first half of the thirteenth century. Therefore, the substantial expansion of window coverage at the beginning of the Gothic era may have been associated with a notable decrease in glazing transmission, which would have limited the lighting gains associated with the birth of Gothic architecture. Furthermore, the Cistercian order mandated (in the late twelfth and thirteenth centuries) the use of ornamental grisaille glass over coloured glass in their institutions, and in many cases early Gothic Cistercian abbey (monastic) churches possessed deliberately smaller apertures than their non-Cistercian counterparts (Wachs, 1964). This indicates that medieval architects had a sophisticated sense of how much daylight should enter an interior, and they restricted it by altering both glazing transmission and/or window sizes. As the New Style evolved in France during the thirteenth century, window sizes expanded considerably, turning churches into virtual glass cages. However, full-colour programs (Chartres, Le Mans) or mixed grisaille/full-colour programs (Reims, Auxerre) continued to be applied, limiting overall glazing transmission. In some places (e.g., Sainte Chapelle, Tours and Le Mans), mid-thirteenth century stained glass projects reached new levels of colour saturation. Then, in an abrupt change in aesthetic from these earlier, colourrich campaigns, more white glass was used in the late thirteenth and fourteenth centuries (Grodecki and Brisac, 1985), a period generally regarded as the beginning of the Little Ice Age (LIA). Sacred architecture projects finally became

largely white glass cages (e.g., Abbey Church of St-Ouen and Évreux Cathedral). Whiter glasses remained dominant in new construction until high-transmissivity enamelled and flashed glass afforded a return to brighter full-colour interiors during the Northern Renaissance. The permanent nature of the change to high-transmissivity glass after a persistent tradition of much darker colour programs suggests dissatisfaction with the earlier aesthetic, perhaps associated with a recognition of cloud cover changes. Even outside France, new projects in the Holy Roman Empire eventually abandoned their largely coloured windows for silver stained, geometric grisailles and half-white grisaille canopy windows in the fourteenth century (Sherrill, 1927). Similarly, evidence from fourteenth century paintings of church interiors from Flanders demonstrates a similar reliance on translucent white glass. Some of the most illustrative examples are paintings by van Eyck from the first half of the fifteenth century: the Madonna in the Church (Gemäldegalerie, Berlin), the Annunciation (National Gallery, Washington D.C.), the Dresden Triptych (Gemäldegalerie, Dresden), and the Madonna with Canon van der Paele (Groeninge Museum, Bruges). These paintings emphasize directional lighting and portray both Romanesque and Gothic interiors with a strong reliance on windows with highly transparent, clear glass (both circular crowns and diamond-shaped quarries). Farther to the north, English cathedrals (outside Canterbury) generally never fully endorsed the largely colour-dominated programs of France and appeared to mix a substantial number of grisailles into their increasingly large early thirteenth century windows (Lincoln, Salisbury, York) (Morgan, 1983; Marks, 1993). Furthermore, when colours were used in these programs, such as at Lincoln (see Fig. 2), they were often less saturated and lighter than those in France, which emphasized darker blues and reds (Arnold, 1925). Thus, England maintained a grisaille-dominated Gothic interior lighting aesthetic, likely due in part to predominately cloudy conditions (M. Lillich, personal communication, 2008). By contrast, France only converted to grisailles very gradually over the course of the second half of the thirteenth century and more permanently by the end of the thirteenth century and beginning of the fourteenth century. Providing a further variation from French and English models, both the Gothic style and stained glass were slower to be adopted in Mediterranean cathedrals, and when stained glass was used it was often richly coloured (such as at Assisi, León, Siena and Toledo). Windows also continued to be severely restricted in size and coverage in southern Europe during the Renaissance, despite the new possibilities for window expansion. In addition, when northern architectural models were directly adopted, their relatively oversized apertures were often filled with richly coloured glass (León and Toledo), which continued to limit interior lighting. Thus, Mediterranean cathedrals maintained small windows, and when stained glass was adopted it was often richly coloured, likely in response to the sunnier climate of this region.

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 221

Fig. 1 Daylighting information (Meteosat, 2011) obtained from algorithms applied to images from the European Space Agency’s geosynchronous Meteorology Satellite (METEOSAT) taken every 30 minutes between sunrise and sunset from 1996–2000, derived from calculations of global horizontal irradiance. The frequency of clear skies during (a) the entire year and (b) winter months (December, January, February and March) is shown. In addition, the frequency of global (including both direct sunlight and diffuse (scattered) illumination) horizontal illuminance for (c) the entire year above 20 klx and (d) the December, January, February and March period above 10 klx is mapped. Finally, the frequency of diffuse (scattered) horizontal illuminance for (e) the entire year above 20 klx and (f) the December, January, February and March period above 10 klx is portrayed. ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

222 / C.T. Simmons and L.A. Mysak

Fig. 2 Geographic locations (a) of major medieval stained glass glazing campaigns and their (b) yearly mean hourly global illuminance and (c) December, January, February and March mean hourly global illuminance. Mediterranean aesthetics (smaller windows and full-colour/mixed stained glass) are represented by warm colours (reds, pinks and oranges). The transitional region that experienced the grisaille revolution (an increasing preference for lighter coloured grisaille-dominated windows in the late thirteenth and fourteenth centuries) are represented by shades of green. Finally, British locations, which generally always favoured very large windows and/or lighter coloured palettes and grisailles, are represented by shades of blue.

Fig. 3 Night cloudiness as determined by comet discovery records from Egypt, Europe and China (between 30° and 50°N); vertical scale unknown. Data before 900 are dominated by records from China and those after 900 are taken mostly from European sightings (Link (1958) with figure reproduced from Lamb (1985) by permission of Princeton University Press). ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 223 Within this context, French cathedrals at first sustained a Mediterranean-style aesthetic with larger windows but darker glass. Then, more than a century after the development of Gothic style, French cathedrals switched to an opposing “cloudy” interior lighting strategy allowing the maximum amount of illumination possible. Therefore, two different trends existed in glazing transmission in the French Gothic style, first opposing and then embracing the increased light levels allowed by greater aperture size (Grodecki and Brisac, 1985). Contrary to changing aesthetic tastes in northern continental Europe, Mediterranean cathedrals maintained a consistent dark aesthetic, with small windows and/or coloured glazing adequate for sunny conditions (see Fig. 2). English cathedrals similarly persisted in their tradition of large windows filled predominately with white glass, compatible with a cloudy aesthetic and limited daylighting (Fig. 2). In addition, consistent with this impression of consistently cloudy aesthetics in the British Isles and sunny aesthetics in southern Europe, a survey of paintings (post 1400) from both regions provided in Neuberger (1970) indicates that English paintings were more likely to depict hazy or overcast skies (especially low clouds) and Italian artists more likely to portray partly cloudy (convective and mid-level clouds) or clear, blue skies. With France and southern Germany lying on the cloud cover gradient between the two extremes (Figs 1 and 2), the transition in France from full-colour to white-dominated glass suggests that sacred interiors designed for a sunny, more Mediterranean-style lighting aesthetic were being replaced by interiors dominated by whiter glass, better adapted to greater lighting under cloudy conditions. As a result, medieval architects may have been subconsciously documenting not just a change in aesthetic but also a climate transition associated with an increase in cloudiness. After this transition there was a persistence of white-dominated glazing programs for nearly two centuries until the innovation of enamel glass and the greater use of flashed glass, which provided high-transmissivity coloured programs in the fifteenth and sixteenth centuries. In this study we wish to test the hypothesis that the grisaille revolution, the transition from the predominately full-colour windows to the largely white-dominated (grisaille and quarry) programs in Gothic France at the end of the thirteenth century, may have been partially driven by an increasing dissatisfaction with the quality and quantity of interior lighting available under the full-colour programs when cloudy conditions prevailed. First, in Section 2, we investigate the climate proxy evidence for increasingly cloudy conditions in the thirteenth and fourteenth centuries during the transition to the LIA, due in part to changing North Atlantic Oscillation (NAO) patterns. Section 3 discusses the collection of lighting data in Europe’s churches and our methods of analysis. Results from this project are presented in Section 4, which first explores the daylight factors and illuminance values (in terms of the SI unit lux (lx), a measure of watts per square metre weighted according to the human eye’s response to

visible radiation) obtained from full-colour (Section 4a) and white-dominated (Section 4b) interiors. Then in Section 4c the implications of church window glazing techniques on architectural form illumination are provided through an analysis of luminance data (measured in SI units of candela per square metre, cd m−2). A discussion of the broader implications of our results and the conclusions are given in Section 5. 2 Evidence of a transition to cloudier conditions in northern continental Europe between the twelfth and fifteenth centuries In order to investigate the changes of medieval sacred interior lighting aesthetic, it is first necessary to determine through proxies how external illumination and cloudiness evolved during the Romanesque and Gothic eras in different parts of Europe. This task has not been undertaken before because few historical records pertaining to the weather are available for the Middle Ages. Cloudiness is also a difficult parameter to document without the help of satellite imagery or a welldeveloped regional network of cloud observations (see the International Satellite Cloud Climatology Project (Warren et al., 2007; ISCCP, 2008). It should thus be noted that there are no proxies that can give a direct measure of cloudiness during the Middle Ages, and regional cloudcover patterns are often difficult to resolve. However, there are some proxy records from which we can make inferences about medieval cloudiness. For example, humidity and precipitation proxies (from ombrotrophic bogs, speleothems and tree rings) may be closely related to cloudiness; precipitation is particularly well-correlated to local nimbostratus occurrence and the proximity of the storm track to northern Europe. In addition, broad-scale temperature changes (as inferred from tree rings, boreholes, speleothems and historical records) can give indications of a mean shift in the position of the jet stream and associated cloud-producing disturbances over Europe (Yan et al., 1997). Finally, atmospheric circulation patterns and related modes of variability (such as the NAO) can be inferred from a combination of different proxies that are used to estimate mean jet-stream orientations, which can be further correlated to cloud patterns. A more in-depth discussion of these patterns is provided in Simmons (2008). In general, low frequency (multidecadal to centennial-scale) changes in European climate are considered most important for this investigation, as they are likely to have had the most impact on perceptions of interior lighting programs, which were sometimes themselves carried out over decades. a Cloud Proxies and Related Indicators One of the few attempts to establish a historical record of cloudiness is provided by Link (1958), which is discussed extensively in Lamb (1985). Link indicates that comet records after 900, dominated by European sightings, indicate a minimum in nighttime cloudiness during the central Middle

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

224 / C.T. Simmons and L.A. Mysak Ages and significantly greater cloudiness during the years 1100–1300 (Fig. 3). More recently, another illumination proxy (related to localized cloudiness) has been discovered by analyzing the δ13C content (dependent on sunshine for photosynthesis) of planktonic foraminifera such as G. ruber found in shallow water sediment cores (Castagnoli et al., 2002). Such an analysis of a core taken in the Gulf of Taranto (southern Italy) suggests a cloud cover minimum during the Medieval Warm Period (MWP) around 1000–1200 and a slight increase in cloudiness during the LIA. It appears that the transition from the MWP to the LIA was not as evident in this part of the Mediterranean basin in terms of daylighting climatology (especially in comparison to the large change in the transition to the Early Medieval Cold Epoch), which is a possible explanation for the relatively continuous daylighting philosophy used in Mediterranean churches throughout the medieval period and into the Renaissance. The near constancy of the Mediterranean climate in terms of temperature from the MWP to the LIA has also been noted in Frisia et al. (2005), Serre-Bachet (1994), Alexandre (1987), Hughes and Diaz (1994) and Chapron et al. (2002). This contrasts with the cooling and rainier climate of northern Europe during this time as reported by McDermott et al. (1999, 2001), Pfister et al. (1998) and Niggemann et al. (2003). Similar to the planktonic foraminifera record, the (sunshinedependent) δ13C content in Scots pine trees in regions of low moisture stress can be used as a proxy for growing-season cloudiness. Recent papers (Young et al., 2010, 2011; Gagen et al. 2011) have produced such time series indicating July– August cloudiness in northern Scandinavia during the past millennium. While this region is far removed from the cathedrals of interest in this study, Young et al. (2011) associates the prolonged period (late fifteenth century to early nineteenth century) of sunny and cool summers during the LIA to broad North Atlantic circulation patterns and a southward displacement of the mean summer storm track over Europe. Alternatively, Young et al. (2011) suggest that a relatively rare combination of conditions in northern Scandinavia (cloudy and cool summers) is likely related to a circulation pattern that is associated with persistent spring and summer rains responsible for prominent famines in northwestern Europe. Most notable are the French and English famines of the fourteenth century (including the Great European Famine of 1315– 1322), a period (1300–1400) characterized by persistently cloudy and cold summers in northern Scandinavia (Young et al., 2011) and well above-average spring rainfall in central Europe (Büntgen et al., 2011).

b Precipitation Proxies While direct cloudiness proxies are rare, indirect cloudiness proxies can be obtained from information on storm tracks and precipitation patterns. Therefore, proxies for high and low precipitation anomalies can also be used as an indicator of the prevalence of precipitation-producing clouds such as

nimbostratus and cumulonimbus and their associated storm systems. Available precipitation-related climate proxies that are available include peat bog stratigraphy, speleothems, tree rings in arid regions, and glacial mass balance fluctuations (see Barber et al., 2004b). Peat analyses of ombrotrophic (exclusively rain-fed) bogs in northwestern Europe provide a direct link to long-term precipitation minus evaporation (henceforth referred to as effective precipitation) variability (Barber et al., 2004a). Peat humification, macrofossil counts and testate amoebae analysis are applied to peat cores. These features are dependent on the prevalence of wet-indicating rather than dry-indicating species that contribute to the bog’s underlying peat layers. Further, because these proxies are derived from plants, they are likely more closely correlated to growing season effective precipitation rather than winter effective precipitation (Barber et al., 2004b). Thus, low bog wetness would likely be related to dry and/or warm conditions during the growing season, although Charman (2010) suggests that precipitation is the dominant influence on bog records for much of the late Holocene because of large bog wetness variability even during times of little temperature change. Barber et al. (2004b) further considers the polar front, and thus the prevailing storm track, likely the predominant influence on bog surface wetness. Peat bog climate reconstructions are often constrained to parts of northwestern Europe, because ombrotrophic bogs require a maritime climate. Therefore, most studies have been limited to Denmark, northern Germany and the British Isles (Haslam, 1987). A variety of peat core analyses have been performed for these regions (see summary in Barber et al., 2004a). Most studies, such as Aaby (1976) in Denmark, Baker et al. (1999) in northwestern Scotland, Barber et al. (2004a) in Denmark and northern Germany, Langdon et al. (2003) in southeastern Scotland, and Chiverrell (2001) in northern England, among others, have demonstrated shifts to wetter/ cooler conditions during the thirteenth or fourteenth centuries (a similar teleconnection pattern is evident for the Early Medieval Cold Epoch, around 1400 years before present (approximately 550), coincident with the transitions to cloudier conditions described in Link (1958) and Castagnoli et al. (2002). Farther to the south in the Jura mountains, French pre-Alps, and the Swiss Plateau, Magny (2004) demonstrates periods of high lake levels beginning around 1250–1350 and continuing after 1394. Combining peat analyses for Ireland and comparing to the lake-level data in Magny (2004), Blundell et al. (2008) also concluded that a significant regional transition from warm/dry to cool/wet conditions occurred around the year 1300. The peat records appear to indicate a southward shift of the polar front during the LIA summer months to a mean position where more synoptic-scale storms affect the British Isles and locations farther south in northern Europe, bringing a combination of cooler air and more saturated conditions to these regions during the thirteenth and fourteenth centuries. This may explain why many Dutch and English artists portrayed

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 225 mostly cloudy landscapes during the LIA (Neuberger, 1970; Lamb, 1985). On the other hand, during the earlier MWP, warmer and drier summer conditions appear to have been observed relatively consistently across many bogs (especially in the eleventh century), suggesting that the climatological summer storm track shifted away from the British Isles, likely either having been weakened or displaced northward. There are also periods (approximately 1090–1230) when northern Great Britain experienced greater bog surface wetness in sync with winter preciptiation reconstructions from Scotland (see Section 2c) while data suggests that regions to the south experienced relatively warm (Lamb, 1965; Yan et al., 1997) and dry (Yan et al., 1998; Büntgen et al., 2010) summers during this period. A shifted or weakened warm season storm track during the MWP is further supported by a recent study of oak tree rings in Kassel, central Germany (Büntgen et al., 2010), which provides a relatively higher-resolution (and lower-latitude) picture of European climate than peat bogs in Demark and the British Isles. Oak trees in central Germany are sensitive to drought, and Büntgen et al. (2010) used the tree-ring record to reconstruct a proxy drought index (Büntgen et al.; their Fig. 7) for the past millennium for north-central Europe. This index suggests that the eleventh and twelfth centuries were both anomalously stable and very dry during the summer, with relatively little high-frequency climate variability. However, precipitation gradually increased starting in the late twelfth century, and summer soil moisture reached an alltime maximum in the third quarter of the thirteenth century and again in the third quarter of the fourteenth century, with greater high-frequency variability. Furthermore, this drought index demonstrates strongly positive and statistically significant correlations (r = 0.4–0.5) with June-July-AugustSeptember cloud cover over France and the southern half of Great Britain (regions of distinct interest in this study) during the calibration period (1950–2000). This indirectly suggests that during the 1000–1200 period there may have been relatively little summer cloud cover in northern France, with greater cloudiness during the relatively more humid period from 1250–1400. A follow-up study (Büntgen et al., 2011) used tree-ring width from Central European firs (including a separate analysis for northeastern France) to determine spring (April–May–June) precipitation anomalies during the past milliennium; these records similarly showed the period from 1000 to 1250 to be anomalously drier than average, with the years 1300–1400 and 1550–1825 being wetter than average. In addition to peat core, fir and oak records, a high-resolution, historically based precipitation proxy was developed for medieval Europe by Yan et al. (1998) for the period 800–1400 (see Fig. 4a). Unlike the aforementioned proxies, which focus on warm-season soil moisture, Yan et al. (1998) take into account all four seasons and address documented precipitation extremes more exclusively. The time series showing year-round trends indicates that the first half of the thirteenth century was anomalously dry, and this dry period was even more pronounced for northwestern

continental Europe (W region in Fig. 4b) than on the panEuropean scale shown in Fig. 4a. Then there was a rapid transition to much wetter conditions by the end of the thirteenth century and a second substantial increase to wetter conditions in the fourteenth century (Fig. 4a). A seasonal comparison of the precipitation proxies for different regions of Europe (Yan et al., 1998, their Fig. 10) agrees qualitatively with Büntgen et al. (2010) for the summer (June, July and August) season in north-central Europe, showing two precipitation maxima at the end of both the thirteenth and fourteenth centuries with a brief dry spell between them in the early fourteenth century. For northern France and the Benelux region (the area of most interest in this paper), this trend is seen in both summer and spring (March, April and May) precipitation (at slightly earlier dates than in central Europe) during the thirteenth and fourteenth centuries, with significant multidecadal fluctuations. The spring series for the twelfth, thirteenth and late fourteenth centuries in Yan et al. (1998; their Fig. 10b) also agrees well with the fir tree proxy evidence for the same season and region (Büntgen et al. 2011; their Fig. 5). A telling dry spell in the late twelfth century and early thirteenth century (especially pronounced in the winter (December, January and February), spring and summer) is visible in every season in the documentary index for northern France and the Low Countries. Furthermore, the separate winter precipitation reconstruction shows this dry spell to be followed by a gradual, continuous increase in precipitation from 1200 until the end of the record in 1426 (Yan et al., 1998; their Fig. 10a). In fact, the documentary evidence from Yan et al. (1998) suggests that regional winter precipitation in northern France and the Low Countries during the entire fourteenth and early fifteenth centuries surpasses that of any other time during the middle ages (except perhaps a brief period in the midninth century). This contrasts with the cold-season (September to June) England–Wales wetness index (partly reliant on temperature–precipitation correlations) in Lamb (1965), which broadly infers high precipitation in the twelfth and thirteenth centuries and rapidly decreasing precipitation after 1300. Decreasing cold-season precipitation in central Great Britain and increasing precipitation in France and the Low Countries suggest a southward displacement of the coldseason storm track and likely relates to the North Atlantic Oscillation (Graham et al., 2007, 2011). c NAO Proxies and Winter Cloudiness The NAO index, defined by the normalized sea level pressure (SLP) difference between the Azores (or alternatively Gibraltar or Lisbon) and Iceland (i.e., PAzores - PIceland), is another quantity that can be used to track cloud cover over the European continent and climatic variability on long time scales. The NAO is one of the dominant modes of climatic variability in the northern hemisphere; it is most strongly correlated with climate parameters (pressure, precipitation, temperature, etc.) in the winter months, but its signal exists year-round (Qian et al., 2000). According to Hurrell (1995), a high (positive) NAO index during winter is associated with a stronger zonal

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

226 / C.T. Simmons and L.A. Mysak

Fig. 4 Historically based precipitation proxy developed in Yan et al. (1995) represented by a 30-year running annual mean for (a) continental Europe as a whole (solid line) and (b) independent analysis regions. The two curves marked 1 and 2 in (a) were derived by Yan et al. (1995; reproduced by permission) by applying two statistical methods relating meteorological station data to the narrative accounts. These curves are averaged to produce the resulting mean precipitation trend (solid curve in (a)). In the vertical axis in (a), positive values indicate the number of seasons (yearly quarters) experiencing wetter than normal conditions, and negative values indicate the number of seasons experiencing drier than normal conditions. The vertical axis in (b) represents (in an offset manner—see next sentence) the number of seasons (yearly quarters) of the historical precipitation dataset for different regions. A constant of 3 is added to all these values in the M (Mediterranean) region of present-day northern/central Italy and southern France, 6 to values in the S (Switzerland and Burgundy) region, 9 to values in the C (central Europe, including most of modern-day Germany and Austria) region, and 12 to values in the W (western continental Europe, encompassing the present-day northern half of France, Belgium, and the Netherlands) region. Figure provided courtesy of Gaston Demarée.

jet (concentrated between Iceland and Scandinavia), southwesterly flow across Europe, greater precipitation in the northern British Isles and Scandinavia, and less precipitation in the Mediterranean and much of France and central Europe. When the NAO index is low (negative), cyclonic activity in the northeastern Atlantic basin during winter decreases; the subpolar jet weakens and shifts south; temperatures decrease across much of Europe; and precipitation often increases in

the Mediterranean basin and central Europe (Hurrell, 1995; Trigo et al., 2002). Illustrative of this, Tiree station in northwestern Scotland and Bergen, Norway, have NAOprecipitation correlations of 0.68 and 0.77, respectively, whereas Paris and Frankfurt (an axis of interest for this study) both have an NAO-precipitation correlation coefficient of −0.18, and the more southerly Lyon has a correlation of −0.37 (Hurrell, 1995).

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 227

Fig. 5 Correlation coeffients (r) for December, January and February (1971– 1996) nimbostratus observation anomalies with the NAM index on a 2.5 × 2.5 grid of Europe, reproduced from Warren et al. (2007) by permission of the American Meteorological Society.

Thus, the NAO index has often been used as a measure of mean latitudinal variation in storm track and may indicate changes in the winter cloud cover climatology of western Europe, which also follows a latitudinal gradient (Fontoynont, 2002; Meerkötter et al., 2004) that is especially pronounced in winter (compare Figs 1e to 1f). It should also be noted that the five-year contemporary dataset represented in Figs 1 and 2 was collected during primarily positive NAO years, and within this context winter daylighting data are given in Figs 1b, 1d, 1f and 2c. In general, NAO and cloud cover relationships for winter are also particularly relevant to architectural studies, as our results indicate that the interior lighting differences between cloudy and sunny conditions are often most visible during the winter. Fortunately, a variety of studies have already been performed on European cloudiness and its relationship to the NAO. Sizov (1997) concluded that decreases (7–10%) in cloudiness in the 40°–50° latitude belt in Europe were often associated with increased values of the NAO index. The area of decreasing cloudiness corresponds broadly to our area of interest in the evolution of the northern Gothic interior lighting aesthetic. In addition, Previdi and Vernon (2007) show that cloud cover (measured by decreased shortwave radiation and increased longwave radiation incident at the surface, see their Fig. 7) for high NAO index months is particularly substantial over the British Isles and western Norway. Also in their study, the cloud cover appears to drop off strongly in southern England and northern France for positive NAO indices. Chaboureau and Claud (2006) provide an exclusive analysis of cloud cover variance in the Mediterranean basin, which reveals that the total number of Mediterranean cloud systems per day decreases with increasing NAO index (particularly in the northeastern and northwestern Mediterranean basin). Warren et al. (2007) provide a more pan-European perspective on the correlations between winter NAO and cloudiness. They compiled over two decades of observational cloud

reports and correlated them with the Northern Annular Mode (NAM) of sea level pressure variability (the first Empirical Orthogonal Function (EOF) of northern hemisphere sea level pressure, analogous to the NAO). Their study demonstrates a very strong negative NAM-cloud (nimbostratus) correlation across much of continental Europe, particularly in Spain, southern France and Italy, as shown in Fig. 5. Correspondingly, Eleftheratos et al. (2007) show the greatest positive NAO-cirriform cloud cover correlations to the north of the British Isles, co-located and to the northwest of the areas demonstrating the greatest nimbostratus correlations in Warren et al. (2007). This area helps define the geographic position of a typical high-NAO cloud band. Altocumulus clouds (a mid-level cloud indicator) show much stronger negative correlations (r = −0.4 to −0.8) over much of continental Europe, a trend that is strongest in northern and central France and weaker in the Mediterranean basin and over the British Isles (Ryan Eastman, personal communication, 2007; Simmons, 2008). Even a small increase in cloudiness may provide significant correlations in the Mediterranean, whereas in northern Europe, where cloud cover is climatologically prevalent, the correlations may represent a larger, more perceivable change in total cloudiness. Using data from the ISCCP, Huang et al. (2006, their Fig. 11a) also provide evidence of positive cloud cover anomalies over western Europe during negative NAO regimes. While cloud cover databases are not yet extensive enough to rigorously test decadal–multidecadal NAO correlations to cloud cover variations, Norris and Wild (2007) and Chiacchio and Wild (2010) have shown that cloud cover has very strong low-frequency correlations to the NAO index, even in the summer, for a 5-year running average of most cloud types. Furthermore, the clouds with the highest correlations are mid- and low-level stratus clouds, which are most closely associated with a Commission Internationale de l’Éclairage (CIE) standard overcast sky that is used for analysis in our study (see Section 3). Norris and Wild (2007) further suggest that prolonged dry periods (such as occurred in the early thirteenth century; see Fig. 4) may be associated with this NAO-related variability. Further confirmation of these trends is provided by European sunshine duration measurements. In particular, while this parameter is strongly influenced by changing aerosol and pollution emissions and residence times, data from 1938 to 2004 in Europe (Sanchez-Lorenzo et al., 2008) nevertheless suggest that winter sunshine duration in a region stretching from Belgium and north-central France to the extreme northern coast of Spain is positively (r = 0.52) and significantly (99%) correlated to the NAM (calculated using North Atlantic sea-level pressure only), signifying sunnier conditions in this region during positive NAO regimes. In addition, a strong positive correlation between sunshine duration and the NAM index was obtained for Spain and Mediterranean France (r = 0.56), as expected from Warren et al. (2007) and Chaboureau and Claud (2006), while in northern England and Scotland the NAM is negatively correlated (r = −0.30) with sunshine

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

228 / C.T. Simmons and L.A. Mysak duration. Given the indicators presented above, we would suspect that the greatest cloud cover over France and central Europe occurs during negative or neutral NAO and NAM regimes and is constrained more to the British Isles during strong positive NAO regimes. A series of proxy records related to the NAO are given in Fig. 6. The one-thousand year precipitation record over northwestern Scotland presented in Proctor et al. (2000, 2002) was the first proxy to be used as a general indication of NAO phase (Fig. 6, second panel from top). From this, we note that the NAO appears to have multidecadal to centennial cycles. In the thirteenth century, in particular, there was a particularly high positive-phase NAO regime (corresponding to less winter cloud cover over continental Europe). This was followed by a dramatic transition to lower NAO indices (less winter precipitation in Scotland) around 1300. As discussed in Section 2b, bog wetness proxies in the British Isles show increasing warm-season precipitation during the 1300–1500 period at the same time as drier winter conditions in the speleothem record from Scotland, and this suggests a storm track over continental Europe in winter and over the British Isles in summer (Charman et al., 2011). Furthermore, an active warm-season storm track corresponds well with the findings of Trouet et al. (2012), which provides evidence that proxies and documentary records indicating greater LIA storminess in the British Isles and the Low Countries (during a prolonged period of more negative NAO indices) are likely associated with spring, summer and autumnal storm systems. In addition, the precipitation in the record from Scotland not only indicates substantial variation in NAO sign over the course of the LIA but also suggests that LIA NAO indices were, on average, significantly lower than in the twelfth and thirteenth centuries. A solar mechanism for this is also given in Shindell et al. (2001); the lower NAO indices detected during the Oort minimum (1010–1050) in solar flux also suggest a relationship with solar processes. After the Oort minimum, solar irradiation was much higher in the 1100– 1250 period than in the subsequent four centuries (with a strong decline between 1250 and 1450) (Bard et al., 2000), and in a model study of this period, Swingedouw et al. (2011) have recently suggested that higher long-term solar irradiances forced a low-frequency positive NAO with a lag of 40 years through a series of teleconnections with Pacific tropical SSTs (see also Graham et al., 2007). A reconstruction of Baltic deep-water temperature provided in Brutckner and Mackensen (2006) extends back to the MWP and further suggests a strong shift in NAO index around the end of the thirteenth century (Fig. 6, second panel from bottom). The authors use a positive correlation between current Skagerrak basin deep water temperature and the NAO index, along with the existing benthic foraminifera δ18O isotopic record of deep water temperature (Fig. 6, bottom), to reference past NAO trends. While Brutckner and Mackensen (2006) conclude that no multidecadal variation in the NAO is evident from their dataset, particularly

because of the rapid oscillations of the LIA pattern after 1600, the trend shows the same signature as Proctor et al. (2002) for the thirteenth to the sixteenth centuries, with a strongly positive NAO phase prevalent in the thirteenth century and a rapid decrease thereafter in the fourteenth century, with the index remaining relatively low until 1600. Recently, Trouet et al. (2009) combined proxies (Scotland winter precipitation minus Morocco winter precipitation) to provide a reconstruction of the NAO index for the past 947 years (Fig. 6, top). This new record confirms a consistently strong positive NAO during the MWP, with a trend toward more and more negative index values between 1250 and 1450. Another mechanism that may be useful in analyzing longterm NAO variability is the sea-ice extent near Iceland. Mysak and Venegas (1998) confirmed that increased sea-ice coverage in the Barents Sea corresponded, on annual time scales, with substantial decreases in the NAO index. Greater sea-ice coverage cuts off latent heat fluxes from the ocean to the Icelandic Low, increasing surface pressure near Iceland and forcing the Icelandic Low southward and/or weakening it. Historical records presented in Ogilvie (1984) and Ogilvie (1991) appear to indicate a relatively calm climate in Iceland during the tenth to the twelfth centuries, with reports of inclement weather becoming more frequent at the end of the late twelfth and thirteenth centuries and direct reports of extensive and semipermanent ice pack in this region by 1364 (Ogilvie, 1984). New proxy methods, based on shallow water sediment cores of diatoms containing the chemical compound IP25 (which is more prevalent in diatoms associated with sea ice), were presented in Belt et al. (2006) and Massé et al. (2008) and provide details on medieval sea-ice extent in Iceland and the location of polar fronts. These reconstructions generally agree with interpretations by Ogilvie and the data seen in the Baltic Sea and NAO proxies using the Scottish data, demonstrating a significant, rapid increase in sea ice off the north coast of Iceland starting in 1300 (Fig. 6, bottom) after a long period of relatively little ice and warm sea surface temperatures (Sicre et al., 2008). Sea-ice presence at the beginning of the LIA was also detected well south of Iceland (57°27.09′N, 27°54.53′ W) in another core (Sicre et al., 2011). This fast transition to an icier North Atlantic occurs in tandem with the rapid, long-term shift toward lower NAO indices around 1300 (Trouet et al., 2009) and may be related to a centennial-scale oscillation in the thermohaline circulation and solar radiation (Bond et al., 2001; Sicre et al., 2008, 2011; Trouet et al., 2012). Finally, we note that strong correlations between glacial extent and sea-ice coverage have been documented in Iceland, and variations in Icelandic glaciers can be linked to an NAO signal (Mackintosh et al., 2002). Glacial advances have been recorded in Iceland since 1200, and this may provide another proxy for the NAO index (Gudmundsson, 1997). It has also been suggested by Nesje and Dahl (2003) that the rapid expansion of Norwegian glaciers during the MWP could not be attributed to decreasing temperature

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 229

Fig. 6 Proxy records and reconstructions that relate to trends in the NAO index during the Middle Ages. (a) the reconstructed index in Trouet et al. (2009), calculated as the difference in precipitation between Morocco and Scotland. Below, the reconstructed northwestern Scotland winter precipitation record from Proctor et al. (2000) (second panel from top) and Skagerrak deep water temperatures from Brutckner and Mackensen (2006) are represented (second panel from bottom). The bottom panel portrays IP25 relative abundances (a measure of late-winter sea ice presence) from a core north of Iceland from Massé et al. (2008).

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

230 / C.T. Simmons and L.A. Mysak alone and may be linked to higher precipitation under a sustained positive NAO pattern. Given the above changes in the NAO and the relationship between cloud cover and precipitation, we propose that the Gothic era is suitable for analyzing architectural responses to likely increases in precipitation and cloud cover across continental northern Europe during the MWP-LIA transition. Furthermore, Figs 1 and 2 suggest that northern and central France and southern Germany bracket a region of several sharp daylighting gradients over relatively short distances (see Clermont Ferrand to Bourges/Paris, Paris to Rouen, and Strasbourg to Cologne in Fig. 2a), which are generally reflective of cloud cover increases toward the north. The difference in global illumination derived from Fig. 2 between regions describing the roughly permanent Mediterranean coloured preferences in stained glass (such as Lyon and Clermont Ferrand) and regions experiencing increasing preferences for light-coloured glasses (Bourges/Paris) are subtle but afford a nearly definitive transition from the art historical perspective. The magnitude of the climate changes during the MWP-LIA transition in Europe are great and permanent enough to suggest that these subtle daylighting thresholds would have been pushed southward during the fourteenth century. The corresponding shift in cloud cover climatology perhaps partly encouraged major northern Gothic churches and cathedrals to permanently adopt more white or higher-transparency glass in new or continuing programs in order to produce an aesthetic better suited for illumination under cloudy conditions. Therefore, this paper next evaluates daylighting performance in early and late Gothic programs under both cloudy and sunny conditions to help define the climatic contribution to the relationship between stained glass and architecture. 3 Methodology A variety of church interiors retaining a large proportion of their original stained glass were analyzed as part of this study. Most of these are in France because many of the best original glazing programs in France are in an easily accessible region around Paris (Picardie, Normandie, Sarthe, Loire and Champagne). This allowed us to collect data by taking day trips out of Paris under strategically chosen weather conditions. This region also corresponds to the greatest cloud cover gradient in Europe according to Meerkötter et al. (2004) and Fontoynont (2002); thus, it is a region where changes in circulation patterns would likely have the most notable effect on cloud cover climatology. We employed four Extech 407026 illuminance meters (see Simmons (2008) for details). Our method of data collection was proposed to us by a committee of lighting experts at the National Research Council Canada (NRC) in Ottawa led by Dr. Jennifer Veitch. We were advised to take ambient illuminance, using the SI unit lux, at regular grid points in various interiors for both overcast and sunny periods. Most interior lighting illuminance measurements were taken at a standard tripod height of

125 cm. Exterior illuminance measurements were also taken by an outside observer toward the sky (horizontally) and in the axial directions of the church, from a position with a quasi-hemispherical view of the sky. From these data, horizontal daylight factors, a frequently used standard for architectural lighting research, were calculated when the prevailing sky conditions were completely cloud covered with little or no brightening toward the sun (Robbins, 1986). Overcast skies allow the calculation of both horizontal daylight factors, which are particularly stable proportions under the symmetric CIE standard overcast sky luminance distribution regardless of cloud type or height (Darula and Kittler, 2002). For a uniform, overcast sky, incoming illuminances are roughly equal in all directions for a particular inclination angle at any given moment, and these were verified by our outside observer using the illuminance meters under cloudy conditions. The horizontal daylight factor (HDF) is defined as a percentage by (Betman, 2005), HDF = 100


 Ed,i,h , Ed,h

(1)

where Ed,i,h is the interior horizontal (zenith, or ceilingdirected) diffuse illumination (the only illumination available under completely overcast skies is diffuse) and Ed,h is the unobstructed (hemispherical sky-view) measure of the exterior horizontal diffuse illuminance (again, under overcast skies the total exterior illuminance is entirely diffuse). Daylight factors should be constant under any uniformly overcast sky, so that regardless of the season or thickness of the cloud cover, given any exterior horizontal illuminance value the corresponding interior illuminances can be determined if the daylight factors are known. This provides a particularly useful method for comparing different interiors using the same standard. Interior measurements were repeated when possible and averaged together to ensure daylight factor consistency. In addition to an evaluation of wide-angle ambient lighting afforded by illuminance measurements, luminance surveys were conducted in several cathedral interiors to help identify visual focal points, bright spots and illumination of the architectural form and aesthetic of the building. We estimated luminance values from a photo series taken using a Canon Digital Rebel XTi 10.1 MP SLR camera (recommended by PierreFélix Breton, a lighting and daylight simulation specialist), from which we created High Dynamic Range (HDR) images and associated luminance estimate profiles. In taking the photo series, we approximately doubled the exposure time step between each individual photo from 1/4000 s to 30 s, as discussed in Moeck and Anaokar (2006) and Beltrán and Mogo (2005), while keeping the f-stop, ISO (film speed or granular light sensitivity), and white balance constant in the manual adjustment mode of the camera. We most often used the settings of ISO 100, f-stop 4.0, and daylight white balance, which were recommended to us by Dugyu Cetegen, an HDR luminance specialist at the NRC, with a lens focal

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 231 length of 28 mm. To avoid jiggling, we took the images remotely using Canon© EOS Utility software on a computer connected to the camera through a USB cable. The photos were taken under either (stable) cloudy or mostly clear-sky conditions so that the lighting was less dynamic. This allows exterior luminance distributions to be reasonably estimated using CIE sky standards. The images were later assembled into Photosphere© (Ward, 2007). The dynamic range of each HDR image was verified through an analysis of red-green-blue (RGB) values of the brightest and darkest incorporated photo, according to the settings recommended by Reinhard et al. (2005). The luminance calculation methodology and algorithms are thoroughly discussed by Inanici (2006) and Moeck and Anaokar (2006). No corrections for vignetting were applied, so some light fall off can be expected around the edges of the HDR image. However, a fish eye (hemispherical-view) lens was not used (instead we commonly used a 28 mm focal length with a bayonnet-type lens hood), so the 20% maximum underestimates seen in Inanici (2006) are not expected. 4 Results The results focus on the daylight performance of glazed Gothic churches and cathedrals containing original glass from the late twelfth to the sixteenth century. These were identified using records of the state of preservation of medieval stained glass obtained from the Corpus Vitrearum Medii Aevi volumes, which provide a detailed reference for nearly every stained glass window in France, England, Germany and other participating countries. Because there are only a few particularly well-preserved programs from any one period, our results are presented chronologically as a collection of case studies that demonstrate a range in possible interior lighting values for each style of architecture and glazing. Daylight factor results (Figs 7 and 8) are presented in cathedral plans in thousandths of a percent; however, the last significant digit is in the hundredths place. Because of instrumental errors and different states of preservation and maintenance of the original glazing, the actual numerical value of daylight factors and illuminance/luminance measurements in any one location is not always representative of the original illumination of the cathedral. However, the range of values seen in a particular type of interior over many points provides a good signal of the original lighting. The intricacies of the state of preservation and restoration of the stained glass windows are dealt with in Simmons (2008) and in other art historical studies; hence, they are not treated here. a Coloured Programs of the Thirteenth Century Chartres Cathedral provides one of the best expressions of the early, richly coloured aesthetic of all of the sacred interiors we analyzed. Its glazing program, retaining 70–90% of its Medieval glass (Lautier, 1994), was largely realized during one period in the evolution of Gothic architecture, and it is an excellent example of the full-colour/mixed tradition of late

Fig. 7 (a) Horizontal daylight factors (multiplied by one thousand) averaged from two separate experiments in the nave and side aisles of Chartres Cathedral. Window numbers are indicated along the outside of the cathedral plan. (b) Horizontal daylight factors for the choir and ambulatory of Tours Cathedral, where the numbers are an average based on two rounds of highly precise measurements. The two figures are not on the same scale.

twelfth and early thirteenth century glazing. Under cloudy conditions (with the cathedral doors shut), the daylight factors calculated for the nave and side aisles (defined in Fig. A1 of the Appendix) were extraordinarily low (see Fig. 7a), varying largely between 0.02% and 0.05%. These values also correspond almost exactly with the range of daylight factors in Bourges Cathedral’s ambulatory (defined in Fig. A1), which are between 0.02% and 0.05% in the outer ambulatory bays and largely between 0.05% and 0.10% in the chapels. Le Mans, Tours (Fig. 7b), and Angers possess roughly equivalent daylight factors in their choirs, again between 0.02% and 0.05%. Therefore, for a variety of different architectural orientations and

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

232 / C.T. Simmons and L.A. Mysak TABLE 1. Illuminance profile of various cathedral interiors under sunny and cloudy conditions (Horizontal Daylight Factors). For sunny illuminances, winter measurements refer to observations carried out between 1–30 January 2008; spring measurements were obtained from 1 May–20 June 2007; and summer readings are from 20 June–7 July 2008. Horizontal Daylight Factors

Sunny Illuminances

Full-Colour Interiors

0.02–0.05% (Angers, Bourges, Chartres, Le Mans, Strasbourg, Tours)

25–50 lx (Le Mans Choir, summer and winter) 10–60 lx (Bourges Ambulatory, summer and winter) 2–10 lx (Chartres Nave, spring/summer) 15 lx (Strasbourg Nave, spring) 8–30 lx (León, winter)

Grisaille Revolution Interiors

0.05–0.06% (Cologne Choir) 0.10–0.20% (Évreux Choir)

34 lx (Beauvais Choir, summer) 45–54 lx (St-Serevin Nave, winter) 39 lx (Évreux Nave, summer)

0.10–0.17% (Bourges Nave) 0.15–0.40% (Évreux Ambulatory) 0.08–0.35% (Chartres South Ambulatory) Renaissance Interiors

0.18–0.23% (Troyes Nave) 0.16–0.25% (Bourges Ambulatory Chapels) 0.20–0.40% (Évreux Lady Chapel)

Fig. 8 (a) Horizontal daylight factors (multiplied by one thousand) in Cologne Cathedral, based on one round of measurements, (b) Horizontal daylight factors (multiplied by one thousand) in Évreux Cathedral, calculated from three rounds of measurements. The two figures are not on the same scale.

contaminations from modern glazing, interior illumination at the surface level in early full-colour interiors appears to be remarkably consistent. On the other hand, clear-sky illumination in full-colour interiors is highly dependent on time of day and season. In Chartres, early morning summer and winter sunlight measurements were not obtained, but spring and summer midday illuminances (see Table 1) ranged from 5 to 10 lx and 2 to 4 lx, respectively. Clearly, early summer solar lighting represents little improvement over typical illumination expected for

0.08–0.22% (St-Gervais-StProtais, Paris, Nave/S. Ambulatory) 0.10–0.20% (St-Romain, Rouen, crossing and transepts) 0.33–0.43% (St-Étienne-duMont Choir and Ambulatory, Paris) 0.20–0.45% (St-Ouen Nave, upper limit) 0.07–0.15% (St-Serevin, Nave) 0.30–1.00% (St-Pantaléon and St-Nicolas, Troyes, upper limit)

39 lx (Troyes Nave, summer) 140–200 lx (Bourges Ambulatory Chapels, winter) 80–202 lx (St-GervaisSt-Protais Chapels, winter) 20–131 lx (St-GervaisSt-Protais Nave/Choir, upper limit) 238 lx (St-Nizier Choir, Troyes, upper limit)

cloudy conditions, although this may well be an issue related to the orientation of the cathedral. In Bourges, the greatest solar infiltration occurred in the winter, producing illuminances of 30–60 lx over a broad area of the south ambulatory in the late morning (see Simmons, 2008). This is in contrast to illuminance values in the summer, where at the same time of day much of the ambulatory possessed values between 10 and 20 lx and only a very constrained area of greater illuminances was present. This winter illumination pattern can be expected on summer mornings with similar low solar angles. Given these results, one major conclusion is that in winter, when interior illuminances are higher in architectural spaces like the ambulatory, there would be a marked contrast

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 233 between sunny and cloudy conditions at most times of day, whereas in summer the differences would be less pronounced (except for the morning or evening hours). The deeper winter penetration of sunlight in the cathedral also provides higher illuminance levels on the side of the cathedral facing away from the sun in Bourges. Therefore, direct solar lighting (particularly low sun angles) provides enhanced illumination over much of the cathedral compared to cloudy-weather lighting. Underneath high spaces such as the choir, similar illuminances are also seen in full-colour interiors. The interior illuminances of Le Mans and Tours cathedrals range from 25–50 lx in the choir year-round, reaching higher levels only under very high solar elevations during the summer midday. b The Grisaille Revolution and Interior Illumination The effects of the grisaille revolution are highlighted by presenting daylight factors in the cathedrals of Cologne and Évreux. The inner choir of Cologne (Fig. 8a) possesses horizontal daylight factors ranging from 0.055% to 0.058%. By contrast, in the full-colour interior seen in the inner choir of Tours, which is further removed from clear modern glass in the nave and transepts affecting the western choir, the daylight factors range from 0.025% under the hemicycle vault convergence (under the eastern apse) to 0.049% at the edge of higher values associated with modern illumination (see Fig. 7b). This indicates that Cologne has as much as twice the illumination seen in the inner heart of the choir at Tours, in spite of the difference in geometry between the two cathedrals that favours greater illumination in Tours (a more compact construction) than in Cologne. These results therefore imply that the grisaille clerestory provides a notable improvement in interior lighting, as much as double the low-level illumination seen in the full-colour aesthetic. The increase is even greater in cathedrals with lower vault elevations, such as Évreux, which also possesses a very well-preserved and nearly completely intact glazing program from the late Gothic period (Callias Bey et al., 2001). The results (Fig. 8b) indicate that the heavily white-dominated glazing of Évreux yields daylight factors in the ambulatory ranging mainly from 0.14% to 0.32%, and averaging slightly lower in the choir between 0.08% and 0.20% (thus, several times brighter than the similarly sized choir at Tours). Comparable values in the range of 0.10% to 0.30% were obtained in the south ambulatory of Chartres, characterized by a mix of thirteenth century coloured glass and later grisailles and quarries. The ambulatory of Évreux appears to be generally brighter than most locations in the choir due to the closer proximity of low-elevation windows and the lower height of the ambulatory vaults where internal reflections occur. However, in all cases the daylight factor values throughout much of the interior are much larger than those calculated for full-colour programs, which often only reached values of 0.02% to 0.05%. Thus, at several points the interior illumination in Évreux is an order of magnitude larger than the interior illumination in Bourges and parts of the north ambulatory of Chartres. We conclude that virtually all post-grisaille

revolution programs belonged to an aesthetic tradition that affords a radical and decisive break from twelfth and thirteenth century norms.

c The Formalist Approach: The Effect of Grisailles on Architectural Lighting The formalist argument provides an often-cited explanation for the increasing preference for grisailles and white-dominated programs in France and Germany at the end of the thirteenth century and beginning of the fourteenth century. This hypothesis suggests that grisailles were selected to better illuminate the architectural forms of the interior structure according to changing aesthetic and theological preferences (Grodecki and Brisac, 1985; Lillich, 1994). With increasing architectural complexity in the fifteenth and later centuries, high translucency glazing continued to be important for providing adequate illumination of these intricate architectural details. A direct causal relationship is hard to establish, especially considering that some Rayonnant interiors with elaborate architectural decoration continued to employ coloured windows (for example, the choirs of Le Mans and perhaps at Amiens). Therefore, it is useful to analyze the performance of coloured and uncoloured Rayonnant programs to demonstrate their respective ability to illuminate architectural forms. One of the best summer solar lighting comparisons between the two types of interiors was obtained from the examples of Évreux and Le Mans around midday. Both cathedrals receive the greatest direct sunlight at the surface near the altar (in the eastern choir) during this time, but they have different orientations (Le Mans’ apse is pointed to the south-southeast and Évreux’s apse is directed due east). As such, Évreux Cathedral had a greater number of directly sunlit windows than Le Mans at the time of our measurements. Figures 9a and 9b show typical luminances from sunny conditions in the early afternoon in Évreux Cathedral. At the clerestory level, the actual luminance levels do not appear to differ substantially from Le Mans’s interior (Fig. 9d). The vault luminances in both interiors are between 10 and 15 cd m−2, with Le Mans’ values skewed only slightly lower than Évreux’s, despite Évreux’s more sunlit interior. The spandrel luminances (those on the wall space between adjoining arches) on the side of the cathedral receiving sunlight are actually lower in Évreux’s example than in Le Mans. Évreux did, however, demonstrate brighter points than Le Mans on the inner choir walls opposite the windows with direct sunshine (compare Figs 9a, 9c and 9d). In Évreux, pillars at the clerestory level have streaks of luminance averaging 30 cd m−2, whereas in Le Mans similar streaks average 20 cd m−2. In another example from Évreux the spandrels below the triforium average just above 40 cd m−2, whereas in Le Mans the choir spandrels are correspondingly lower, at around 25 cd m−2. Similarly, the low-level column bases in Le Mans have luminances of 20–30 cd m−2 and in Évreux, 30–40 cd m−2. To summarize these observations on the summer solar illumination of architectural forms, the full-colour interior has just

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

234 / C.T. Simmons and L.A. Mysak

Fig. 9 Clear-sky luminance profile (0–50 cd m−2) for (a) the choir clerestory of Évreux Cathedral at 1132 GMT 24 June 2008, (b) the lower choir of Évreux Cathedral at 1138 GMT 24 June 2008, and (c) the lower choir of Le Mans Cathedral at 1014 GMT 30 June 2008, (d) and the choir clerestory of Le Mans at 1007 GMT 30 June 2008.

over half the luminance for some limited surfaces in the lower choir opposite the sunlit windows and more than half the luminance for other surfaces when compared to the grisailled interior. Similar values were obtained for winter measurements (see Simmons, 2008); for example, a comparison of Beauvais’ and Le Mans’ measurements indicate that surfaces opposite windows receiving direct sunshine in Beauvais were often illuminated by more extensive bright patches than corresponding surfaces in Le Mans. However, outside of these bright patches, much of the rest of Beauvais’ interior maintained luminances comparable to those in Le Mans. Therefore, for sunny conditions Le Mans affords a more evenly lit interior daylighting aesthetic and a more even diffusion of sunlight through the interior, with the grisailled programs providing limited lighting gains in comparison. Furthermore, windows opposite those receiving direct sunlight in grisailled interiors are often strongly backlit (obscured

by the strong sunshine) at low inclination sun angles, rendering a large part of the coloured iconographical program illegible. The same degree of backlight is not observed in Le Mans even at very low sun angles, likely because of the greater diffusion of sunlight associated with full-colour glass. Thus, the lighting gains on architectural forms associated with grisailled programs appears to be minimal for sunny conditions, and reflected sunlight can have a significant effect on the iconographical function of band windows. However, for overcast conditions the formalist argument appears to have greater merit. A cloudy weather luminance map is unfortunately not available for Le Mans, but with daylight factors similar to those found in the inner choir of Tours, we can probably expect comparable lighting performances and vault luminances for the two interiors. A collection of three choirs for which exterior horizontal illuminances (indicated in parentheses) were similar at the time of observation are

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 235

Fig. 10 The choir overcast luminance profiles (0–5 cd m−2) of (a) Tours Cathedral at 955 GMT 29 January 2008, (b) St-Ouen in Rouen at 1051 GMT 26 January, 2008, and (c) Cologne Cathedral at 1211 GMT 14 January 2008. In (c), the white-painted vault luminance values have been multiplied by a factor of 0.764 to account for the higher reflectance of the white paint compared to the original limestone surface, assuming that the paint and limestone have a similar reflective diffusivity.

presented here (Fig. 10): Tours (8610 lx), St-Ouen (7540 lx), and Cologne (9600 lx). Overcast, slightly foggy exterior conditions dominated the measurements for Tours and St-Ouen, whereas the measurements at Cologne were taken under generally overcast stratus conditions without fog. Although Cologne, unlike St-Ouen and Tours, possesses painted vaults, the reflectance of the white paint was corrected to represent that of the interior’s limestone by downward calibrating pixels associated with the white paint by 76.4% of the original value to bring its reflectance down to that of the limestone. The effects of the whiter glass on the illumination of the architecture appear to be quite significant, especially for the well-preserved grisailles of Cologne; here the vaults are illuminated over 5 cd m−2. By contrast, the vaults of Tours Cathedral are illuminated predominately under 1 cd m−2 outside of the presence of the two band windows in the choir. The anomalously bright patch on the vaults near Tours’ Cathedral choir is associated with window 105, a band window which has modern replacement glass in its upper levels (Grodecki et al., 1981). More modern glass is present in the clerestory of St-Ouen than in the other two examples, but many of the windows have transmissions comparable to the original glass (see Simmons and Mysak, 2010). The contrast in lighting between different forms appears to be less significant in Tours under cloudy conditions and Le Mans under sunny conditions, which produces a more even illumination of the architecture. Thus, glaziers and architects appear to have used not

just greater illumination but illumination contrasts to accentuate forms. Also, in St-Ouen, the inner vaults possess luminance values across broad surfaces that are more than double those of the vaults in the inner choir of Tours. In summary, while sunny conditions in grisaille interiors were only observed to double the luminance on forms for a few surfaces of the interior (compared to full-colour programs like Le Mans), cloudy conditions in grisaille interiors generally produce more than double the luminance of most forms in the interior. We now discuss later Renaissance glazing by focusing on the Troyes Cathedral clerestory. The upper levels of the cathedral, which as in other examples are further removed from the aisle-level contaminations of modern grisailles, provides the best environment to analyze the lighting afforded by the interior’s Renaissance windows. With equivalent window dimensions and stone reflectances (confirmed using a reflectance standard) between the choir and nave, we expect glazing transmission to be the primary control on the luminance values received on interior surfaces. In both cases, the choir measurements were taken first, and the declining sun at the time suggests that the background exterior luminance distribution for the choir measurements (under a thirteenth century full-colour aesthetic) should be brighter than for the photo series taken afterward in the nave if both sides of the cathedral possessed glass with the same glazing transmission. Furthermore, the greater contamination lighting from the

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

236 / C.T. Simmons and L.A. Mysak

Fig. 11 Luminance profiles (cd m−2) obtained from four HDR images taken on 9 January 2008 in the choir ((a) and (c), respectively) and nave ((b) and (d), respectively) of Troyes Cathedral. The window numbers given below are those provided in Callias Bey et al. (1992). North choir HDR in (a), taken at a mean time of 1510 GMT , is centred on window 209, whereas the north nave in (b), taken at a mean time of 1516 GMT , is centred on window 232. Both (a) and (b) are on a luminance scale of 0–1.5 cd m−2. In (c), the south choir HDR photo series, taken at a mean time of 1528 GMT , is centred on window 208, whereas in (d) the south nave centred on window 231 is taken at a mean time of 1533 GMT . Both (c) and (d) are on a luminance scale of 0–0.7 cd m−2.

ambulatory windows would also favour greater lighting of forms in the choir over the nave. However, Fig. 11 reveals that, even with the preferential disposition of the choir to have greater lighting, the nave possesses noticeably higher luminances over almost all surfaces compared to those of the choir. Much of the triforium tracery and piers of the nave in particular are more than double the luminances seen for similar points in the choir. Therefore, it appears that the more richly coloured Renaissance glazing traditions common in France in the fifteenth and sixteenth centuries continued to provide greater illumination to forms on a level comparable to that seen during the grisaille revolution. This, in turn, isolates the grisaille revolution during the turn of the fourteenth century as a single, permanent shift in interior lighting aesthetic.

5 Discussion and conclusions Using an illuminance meter and luminance estimates determined from high dynamic range images, we have provided a profile of interior illumination in several Gothic cathedrals retaining much of their original stained glass (as dated in Corpus Vitrearum volumes). These measurements were performed during different seasons under different types of skies, with an emphasis on overcast versus mostly clear conditions. These data, summarized in Table 1, indicate that the interior illumination in locations with lighting dominated by

richly coloured glass (typical of early and mid-thirteenth century interiors in northern Europe) is extraordinarily low, with daylight factors ranging from 0.02% to 0.05% for most measurements in Angers, Bourges, Chartres, Le Mans, Strasbourg, and Tours cathedrals (Table 1, Full-Colour Interiors). Interiors with more grisaille windows and whiter glass from the late thirteenth century and fourteenth century, such as Cologne and Évreux, retain daylight factors that are double or more (Table 1, Grisaille Revolution Interiors). In addition, the coloured Renaissance programs of the fifteenth and sixteenth centuries (e.g., St-Serevin and Troyes) also provide high daylight factors (generally above 0.1%; see Table 1, Renaissance Interiors), which are, like grisaille-dominated programs at Évreux, at times as much as an order of magnitude greater than the illumination in full-colour interiors. Thus, the increase in ambient lighting in the white-glazed structures under overcast skies is enough to provide conditions that are often as bright as full-colour interiors under sunny conditions. The extremes between cloudy and sunny lighting are naturally more marked in the winter, when low sun angles provide much greater interior illumination in certain parts of the church under clear skies, whereas winter interior illuminances are very low under cloudy conditions. During the summer, exterior illuminances under cloudy skies are higher (horizontal exterior illuminances as high as 30000–40000 lx) for most overcast conditions than during the winter (rarely above 20000 lx), thus providing less of a contrast between solar

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 237 and overcast lighting in summer than in winter. Therefore, changes in the winter cloud cover climatology (between the autumnal and vernal equinoxes) may potentially have a greater effect on the perception of solar and diffuse interior lighting than under summer lighting. Therefore, if cloud cover (or lack thereof) was indeed a design factor, as suggested by the contrasting preferred aesthetics of Mediterranean and British glazing, winter cloudiness would likely be the most important consideration. This is due to both climatologically greater cloud cover during winter and the fact that winter illumination produces the largest extremes between solar and overcast illumination lighting. Thus, the variability in cloud cover as it relates to the winter NAO is deemed pertinent to our study, and a significant transition in the NAO between the thirteenth and fourteenth centuries, associated with likely cloudier conditions over continental Europe, may be relevant to concurrent experimentation with glazing styles. While the lack of a lag between the grisaille revolution and the cloud cover changes expected from other proxies may seem problematic, the cloud cover shift may well have partially encouraged the development of high transmissivity quarries, the widespread adoption of silver stain and the eventual conversion of German glass to more translucent glass over the course of the fourteenth century. Furthermore, the climatic shift may have helped ensure continuity in the high translucency tradition throughout the late Gothic period. Our luminance measurements further suggest that solar lighting in full-colour interiors appears to provide an even illumination that is often much brighter than cloudy illumination for post grisaille-revolution interiors. Despite some gains in lighting under sunny conditions for grisailled interiors, much of the architectural space retains luminances equivalent to or only slightly greater than those seen in corresponding sections of full-colour interiors under sunny conditions. Furthermore, when provided with a clear sky and low sun angles (mornings, evenings, and all day during winter), reflected sunlight becomes particularly severe in white-dominated programs and often obscures windows opposite those receiving direct sunlight. By contrast, under cloudy conditions, as demonstrated for Cologne and Troyes cathedrals, the illumination in grisaille revolution interiors becomes more even (as seen in Le Mans for clear-sky conditions), and the luminance values seen on many of the architectural forms are as much as double or more those for full-colour interiors. The very nature of the overcast sky, with its largely symmetric luminance distribution, produces more substantial gains in lighting across broad sections of the interior for post-grisaille revolution programs under cloudy conditions. These gains are much greater than those seen under sunny conditions when compared to coloured interiors. Therefore, the formalist argument for the grisaille revolution is most viable when evaluating the performance of full-colour and post-grisaille revolution interiors under cloudy conditions. This suggests, along with the extensive and often debilitating backlighting seen for white-

dominated interiors under sunny skies, that grisaille glass may have indeed been ideally adopted for cloudier conditions. Therefore, we conclude that increasingly cloudy conditions over continental Europe during the thirteenth and fourteenth centuries may have influenced the aesthetic outlook and architectural design of many cathedrals, supporting the inclusion of more white glass during a period of remarkable experimentation in architectural design. While this study suggests that there is some correlation between the use of white glass and cloud cover variability, it does not conclude that cloud cover and cloud climatology changes are the primary driving factor in medieval stained glass aesthetics. Others, such as economic concerns, changes in religious philosophy (Lillich, 2001), and even broad, illogical trends in aesthetic tastes or perception (Lillich, 1994) may also be more important, as discussed further in Simmons (2007, 2008). Given the marked adaptations of architecture (and window size) for even small changes in climate, as seen when comparing the more northern architecture of the Loire Valley to that just a few hundred kilometers to the south, we suspect that changes in the climate system may be seen through human adaptations in architecture. Therefore, the rapid climatic transition occurring during the Gothic period may have breached a threshold that required, either consciously or subconsciously, human concessions to the climate.

Acknowledgements This research has been largely supported through the Richard H. Tomlinson Fellowship awarded to CTS for graduate studies at McGill University. The financial support through a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant awarded to LAM is also gratefully acknowledged. In addition, Dr. Jennifer Veitch, Chantal Arsenault, Duygu Cetegen, Dr. Christoph Reinhart and others from the National Research Council Canada’s Institute for Research in Construction (Ottawa) are acknowledged for their close advisory role in the technical execution of this research. Furthermore, Teresa Simmons, Robert Simmons and Cathérine Thauer are thanked for their help in data collection. For permissions and additional support, we gladly acknowledge the Services Departmentaux de l’Architecture et du Patrimoine, Alain de Maistre, Msgr. Claude Cesbron, Gilles Fresson, Père Jacques Legoux, Curé Jean-Michel Bodin, Juan Pedro Sánchez Gamero, Père Roland Frat, Dr. Ulrike Brinkmann and others. In addition, Stephen Warren and Ryan Eastman from the University of Washington, Seattle, USA, are also gratefully recognized for providing us further insight into the correlations between the Northern Annular Mode and cloud cover. We particularly thank Dr. Gaston Demarée for his enlightening discourse and providing us with his historical dataset on European precipitation patterns during the Middle Ages, which we show in Fig. 4. Finally, Dr. Guillaume Massé is gratefully acknowledged for providing the raw data for the bottom panel in Fig. 6.

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

238 / C.T. Simmons and L.A. Mysak Appendix: Definitions The architectural regions of a church are shown in Fig. A1. Other technical and historical terms used in this paper are given below. Band Window – a stained glass window composed of a grisaille (uncoloured glass) background pierced by one or more horizontal bands of coloured figures. Clerestory – the top (upper) level windows and architecture of a church or cathedral. Enamelled stained glass – stained glass painted with a thin layer of coloured vitreous enamels, or ground glass substrate. Flashed glass – the application of a very thin coloured glass layer on another thicker piece of glass (usually uncoloured), often accomplished through dipping a piece of glass into a coloured glass substrate. Gothic – a general term applied to the architectural style characterized by bent arches, flying buttresses and ribbed vaults, developed in France starting in the mid-twelfth century. Gothic architecture can be subdivided into several periods of increasing architectural complexity (with approximate dates): early/ transitional Gothic (1140–1190), high Gothic (1190–1240), Rayonnant Gothic (1240–1300) and late Gothic (1300–1550). Grisailles – any “white” glass (for the purposes of this paper), often with a silvery, greenish, or greyish hue (hence the name, grisaille), in the twelfth to fourteenth centuries. This whitish glass became generally more transparent by the fifteenth and sixteenth centuries. Quarry – a term, from the French carré, often applied to the larger, diamond or square-shaped panes of white (grisaille) glass frequently used in the fourteenth century. Late thirteenth or early fourteenth century quarry prototypes are often called bulged quarries. Romanesque – an architectural style, characterized strongly by halfcircle arches, developed for churches roughly between 1000 and 1190 in northern France (and through the early fourteenth century in the Holy Roman Empire). Most twelfth century glass is considered “Romanesque” following Grodecki (1983). Renaissance glass – a generalized term used here for fifteenth and sixteenth century glass. Silver stain – staining of glass through the application of silver oxide during the melting phase, usually used to produce various shades of highly transmissive yellow or orange in the fourteenth century and thereafter. Triforium – the mid-levels of a church, above the aisle-level arcade and below the clerestory.

Fig. A1 A plan of Amiens Cathedral with the distinct architectural regions of the church defined.

References 1976. Cyclic climatic variations in climate over the past 5500 years reflected in raised bogs. Nature, 263: 281–284. ALEXANDRE, P. 1987. Le Climat en Europe au Moyen Âge. Paris: École des Hautes Études en Sciences Sociales. ANYHERE SOFTWARE . 2007. [Computer software]. Retrieved from http:// www.anyhere.com/ ARNOLD, H. 1925. Stained glass of the middle ages in England and France. London: A. & C. Black. BAKER, A.; C. CASELDINE and M. GILMOUR. 1999. Stalagmite luminescence and peat humification records of palaeomoisture for the last 2500 years. Earth Planet. Sci. Lett. 165: 157–162. BARBER, K.E.; F.M. CHAMBERS and D. MADDY. 2004a. Late Holocene climatic history of northern Germany and Denmark: peat macrofossil investigations at Dosenmoor, Schleswig-Holstein, and Svanemose, Jutland. Boreas, 33: 132–144. AABY, B.

and A.F. LOTTER. 2004b. Atlantic to Urals — the Holocene climatic record of mid-latitude Europe. In R.W. Battarbee, F. Gasse and C.E. Stickley (Eds), Past Climate Variability through Europe and Africa (pp. 417–442). Dordrecht, Netherlands: Kluwer Academic Publishers. BARD, E.; G. RAISBECK, F. YIOU and C. JOUZEL. 2000. Solar irradiance during the last 1200 years based on cosmogenic nuclides. Tellus, 52 (B): 985–992. BELT, S.T.; G. MASSÉ, S.J. ROWLAND, M. POULIN, C. MICHEL and B. LEBLANC. 2006. A novel chemical fossil of palaeo sea ice: IP25. Org. Geochem. 38: 16–27. BELTRÁN, L.O. and B.M. MOGO. 2005. Assessment of luminance distribution using HDR photography. Proceedings of the ISES Solar World Congress, Orlando, FL, 7–12 August, 2005. BETMAN, E. 2005. Daylight calculations using constant luminance curves. Renew. Energy, 30: 241–257. BARBER, K.; B. ZOLITSCHKA, P. TARASOV

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie

Stained Glass and Climate Changes: How Are They Connected? / 239 and K.E. BARBER. 2008. Multiproxy late Holocene peat records from Ireland: Towards a regional palaeoclimate curve. J. Quat. Sci. 23: 59–71.

and W. TEGEL. 2011. Combined dendro-documentary evidence of Central European hydroclimatic springtime extremes over the last millennium. Quat. Sci. Rev. 30: 3947–3959. CALLIAS BEY , M .; V . CHAUSSÉ and L. DE FINANCE . 1992. Les vitraux de Champagne-Ardenne. Corpus Vitrearum-France, Rencensement des vitraux anciens de la France, IV (Paris). CALLIAS BEY, M.; V. CHAUSSÉ, F. GATOUILLAT and M. HÉROLD. 2001. Les vitraux de Haute-Normandie. Paris: Corpus Vitrearum-France, Recensement des vitraux anciens de la France IV. CASTAGNOLI, C.C.; G. BONINO, C. TARICCO and S.M. BERNASCONI. 2002. Solar radiation variability in the last 1400 years recorded in the carbon isotope ratio of a Mediterranean sea core. Adv. Space Res. 29: 1989–1994. CHABOUREAU, J.P. and C. CLAUD. 2006. Satellite-based climatology of Mediterranean cloud systems and their association with large-scale circulation. J. Geophys. Res. 111 (D1): D01102.1–D01102.10, doi:10.1029/2005JD006460 CHAPRON, E.; M. DESMET, T. DE PUTTER, M-F. LOUTRE, C. BECK and J-F. DECONINCK. 2002. Climatic variability in the northwestern Alps, France, as evidenced by 600 years of terrigenous sedimentation in Lake Le Bourget. The Holocene, 12: 177–185. CHARMAN, D.J. 2010. Centennial climate variability in the British Isles during the mid-late Holocene. Quat. Sci. Rev. 29: 1539–1554. CHARMAN, D.J.; V. HOHL, A. BLUNDELL, F. MITCHELL, J. NEWBERRY and P. OKSANEN. 2011. A 1000-year reconstruction of summer precipitation from Ireland: Calibration of a peat-based palaeoclimate record. Quatern. Int. In press: doi: 10.1016/j.quaint.2011.12.011. CHIACCHIO, M. and M. WILD. 2010. Influence of NAO and clouds on longterm seasonal variations of surface solar radiation in Europe. J. Geophys. Res. 115: D00D22, doi:10.1029/2009JD012182 CHIVERRELL, R. 2001. A proxy record of late Holocene climate change from May Moss, northeast England. J. Quat. Sci. 16: 9–29. DARULA, S. and R. KITTLER. 2002. CIE general sky standard defining luminance distributions. Proceedings of the bi-annual IBPSA conference, IBPSA, Canada. ELEFTHERATOS, K.; C.S. ZEREFOS, P. ZANIS, D.S. BALIS, G. TSELIOUDIS and K. GIERENS. 2007. A study on natural and manmade global interannual fluctuations of cirrus cloud cover for the period 1984–2004. Atmos. Chem. Phys. 7: 2631–2642. FONTOYNONT, M. 2002. Perceived performance of daylighting systems: lighting efficacy and agreeableness. Sol. Energy, 73: 83–94. FRISIA, S.; A. BORSATO, C. SPÖTL, I.M. VILLA and F. CUCCHI. 2005. Climate variability in the SE Alps of Italy over the past 17 000 years reconstructed from a stalagmite record. Boreas, 34: 445–455.

AMMANN, D. FLEITMANN, K.M. COBB and 2011. Support for global climate reorganization during the “Medieval Climate Anomaly”. Clim. Dyn. 37: 1217–1245. GRODECKI, L. 1983. Le vitrail roman (2nd ed.). Fribourg: Fribourg Office du livre. GRODECKI, L.; F. PERROT and J. TARALON. 1981. Les Vitraux du centre et des pays de la Loire. Paris: Corpus Vitrearum-France, Recensement des vitraux anciens de la France II. GRODECKI, L. and C. BRISAC. 1985. Gothic stained glass, 1200–1300. Ithaca, NY: Cornell University Press. GUDMUNDSSON, H.J. 1997. A review of the Holocene environmental history of Iceland. Quat. Sci. Rev. 16: 81–92. HASLAM , C.J . 1987. Late Holocene peat stratigraphy and climatic change - a macrofossil investigation from raised mires of north-western Europe. Ph.D. thesis, University of Southampton. HUANG, J.; M. JI, K. HIGUCHI and A. SHABBAR. 2006. Temporal structures of the North Atlantic Oscillation and its impact on the regional climate variability. Adv. Atmos. Sci. 23: 23–32. HUGHES, M.K. and H.F. DIAZ. 1994. Was there a ‘medieval warm period,’ and if so, where and when? Clim. Change, 26: 109–142. HURRELL, J.W. 1995. Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science, 269: 676–679. INANICI, M.N. 2006. Evaluation of high dynamic range photography as a luminance data acquisition system. Light Res. Technolog. 38: 123–134. ISCCP (INTERNATIONAL SATELLITE CLOUD CLIMATOLOGY PROJECT) . 2008. ISCCP 25th symposium: Website highlights. Retrieved from http:// isccp.giss.nasa.gov/ JANSON, A. 2001. History of Art (5th ed.). New York: Harry N. Abrams. LAMB, H.H. 1965. The early medieval warm epoch and its sequel. Palaeogeogr. Palaeoclimatol. Palaeoecol. 1: 13–37. LAMB, H. 1985. Climatic History and the Future. Princeton, NJ: Princeton University Press. LANGDON, P.G.; K.E. BARBER and P.D. HUGHES. 2003. A 7500-year peatbased palaeoclimatic reconstruction and evidence for an 1100-year cyclicity in bog surface wetness from Temple Hill Moss, Pentland Hills, southeast Scotland. Quat. Sci. Rev. 22: 259–274. LAUTIER, C. 1994. Les restaurations anciennes dans les vitraux de la nef de la cathédrale de Chartres. Actes du XVIe colloque du comité technique international du Corpus Vitrearum. News Letter Corpus Vitrearum, 45: 4–14. LILLICH, M. 1994. The armor of light : stained glass in western France, 1250–1325. Berkeley, CA: University of California Press. LILLICH, M. 2001. Studies in Medieval Stained Glass and Monasticism. London: Pindar Press. LINK , F. 1958. Kometen, Sonnentätigkeit und Klimaschwankungen. Die Sterne, 34: 129–140. MACKINTOSH, A.N.; A.J. DUGMORE and A.L. HUBBARD. 2002. Holocene climatic changes in Iceland: evidence from modelling glacier length fluctuations at Solheimajokull. Quat. Int. 91: 39–52. MAGNY, M. 2004. Holocene climate variability as reflected by mid-European lake-level fluctuations and its probable impact on prehistoric human settlements. Quat. Int. 113: 65–79. MARKS, R. 1993. Stained glass in England during the Middle Ages. Toronto: University of Toronto Press. MASSÉ, G.; S.J. ROWLAND, M.A. SICRE, J. JACOB, E. JANSEN and S.T. BELT. 2008. Abrupt climate changes for Iceland during the last millennium: Evidence from high resolution sea ice reconstructions. Earth Planet. Sci. Lett. 269: 564–568.

GAGEN, M.H.; E. ZORITA, D. MCCARROLL, G.H.F. YOUNG, H. GRUDD,

MCDERMOTT, F.; S. FRISIA, Y. HUANG, A. LONGINELLI, S. BARUCH, T.H.E. HEATON,

and A.J. KIRCHHEFER. 2011. Cloud response to summer temperatures in Fennoscandia over the last thousand years. Geophys Res. Lett. 38: L05701, doi:05710.01029/02010GL046216

C.J. HAWKESWORTH, A. BORSATO, E. KEPPENS, I.J. FAIRCHILD, K. VAN DER

BLUNDELL, A.; D.J. CHARMAN

BOND, G.; B. KROMER, J. BEER, R. MUSCHELER, M.N. EVANS, W. SHOWERS,

and G. BONANI. 2001. Persistent solar influence on north Atlantic climate during the Holocene. Science, 294: 2130–2136. BRUTCKNER, S. and A. MACKENSEN. 2006. Deep-water renewal in the Skagerrak during the last 1200 years triggered by the North Atlantic Oscillation: evidence from benthic foraminiferal δ18O. The Holocene, 16: 331–340. S. HOFFMANN, R. LOTTI-BOND, I. HAJDAS

BÜNTGEN, U.; V. TROUET, D. FRANK, H.H. LEUSCHNER, D. FRIEDRICHS,

and J. ESPER. 2010. Tree-ring indicators of German summer drought over the last millennium. Quat. Sci. Rev. 29: 1005–1016.

J. LUTERBACHER

BÜNTGEN, U.; R. BRÁZDIL, K-E. HEUSSNER, J. HOFMANN, R. KONTIC, T. KYNCL, C. PFISTER, K. CHROMÁ

R. JALKANEN, N.J. LOADER, I. ROBERTSON

GRAHAM, N.E.; M.K. HUGHES, C.M. AMMANN, K.M. COBB, M.P. HOERLING, D. J. KENNETT, J.P. KENNETT, B. REIN, L. STOTT, P.E. WIGAND and T. XU. 2007. Tropical Pacific – mid-latitude teleconnections in medieval times. Clim. Change, 83: 241–285.

GRAHAM,

N.E.;

C.M.

J. LUTERBACHER.

BORG, S. VERHEYDEN and E. SELMO. 1999. Holocene climate variability in Europe: Evidence from δ18O, textural and extension-rate variations in three speleothems. Quat. Sci. Rev. 18: 1021–1038. MCDERMOTT, F.; D.P. MATTEY and C.J. HAWKESWORTH. 2001. Centennialscale Holocene climate variability revealed by a high-resolution speleothem δ18O record from SW Ireland. Science, 294: 1328–1331.

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 Canadian Meteorological and Oceanographic Society

240 / C.T. Simmons and L.A. Mysak and H. MANNSTEIN. 2004. A 14-year European cloud climatology from NOAA/AVHRR data in comparison to surface observations. Geophys. Res. Lett. 31: L15103.1– L15103.4, doi:10.1029/2004GL020098 METEOSAT . 2011. S@tel-Light: The European database of daylight and solar radiation [Data]. Retrieved from http://www.satel-light.com/ MOECK, M. and S. ANAOKAR. 2006. Illuminance analysis from high dynamic range images. J. Illum. Eng. Soc. 2: 211–228. MOORE, F. 1985. Concepts and practice of architectural daylighting. New York: Van Nostrand Reinhold. MORGAN, N.J. 1983. The medieval painted glass of Lincoln Cathedral. London: Corpus Vitrearum Great Britain Occasional Paper III. MYSAK, L.A. and S.A. VENEGAS. 1998. Decadal climate oscillations in the Arctic: A new feedback loop for atmosphere-ice-ocean interactions. Geophys. Res. Lett. 25: 3607–3610. NESJE, A. and S.O. DAHL. 2003. The ‘Little Ice Age’—only temperature? The Holocene, 13: 139–145. NEUBERGER, H. 1970. Climate in art. Weather, 25: 46–56. NIGGEMANN, S.; A. MANGINI, D.K. RICHTER and G. WURTH. 2003. A paleoclimate record of the last 17,600 years in stalagmites from the B7 cave, Sauerland, Germany. Quat. Sci. Rev. 22: 555–567. NORRIS, J.R. and M. WILD. 2007. Trends in aerosol radiative effects over Europe inferred from observed cloud cover, solar “dimming,” and solar “brightening”. J. Geophys. Res. 112: D08214, doi:10.1029/2006JD007794 OGILVIE, A.E.J. 1984. The past climate and sea-ice record from Iceland, Part 1: Data to A.D. 1780. Clim. Change, 6: 131–152. OGILVIE, A.E.J. 1991. Climatic changes in Iceland AD c. 865 to 1598. Acta Archaeologica, 61: 233–251. PFISTER, C.; J. LUTERBACHER, G. SCHWARZ-ZANETTI and M. WEGMANN. 1998. Winter air temperature variations in western Europe during the Early and High Middle Ages (AD 750–1300). The Holocene, 8: 535–552. PREVIDI, M. and D.E. VERNON. 2007. North Atlantic cloud cover response to the North Atlantic Oscillation and relationship to surface temperature changes. J. Geophys Res. 112 (D7): D07104, doi:10.1029/2006JD007516 PROCTOR, C.J.; A. BAKER, W.L. BARNES and M.A. GILMOUR. 2000. A thousand year speleothem proxy record of North Atlantic climate from Scotland. Clim. Dyn. 16: 815–820. PROCTOR, C.J.; A. BAKER and W.L. BARNES. 2002. A three thousand year record of North Atlantic climate. Clim. Dyn. 19: 449–454. QIAN, B.; J. CORTE-REAL and H. XU. 2000. Is the North Atlantic Oscillation the most important atmospheric pattern for precipitation in Europe? J. Geophys. Res. 105 (D9): 11901–11910, doi:10.1029/2000JD900102 REINHARD, E.; G. WARD, S. PATTANAIK and P. DEBEVEC. 2005. High dynamic range imaging: acquisition, display, and image-based lighting. San Francisco: Morgan Kaufmann. ROBBINS, C. 1986. Daylighting: design and analysis. New York: Van Nostrand Reinhold. SANCHEZ-LORENZO, A.; J. CALBÓ and J. MARTIN-VIDE. 2008. Spatial and temporal trends in sunshine duration over western Europe (1938–2004). J. Clim. 21: 6089–6098. SERRE-BACHET, F. 1994. Middle ages temperature reconstructions in Europe, a focus on northeastern Italy. Clim. Change, 26: 213–224. SHERRILL, C.H. 1927. Stained glass tours in Germany, Austria and the Rhine lands. London: John Lane. SHINDELL, D.T.; G.A. SCHMIDT, M.E. MANN, D. RIND and A. WAPLE. 2001. Solar forcing of regional climate change during the Maunder Minimum. Science, 294: 2149–2152. MEERKÖTTER, R.; C. KÖNIG, P. BISSOLLI, G. GESELL

SICRE, M.A.; J. JACOB, U. EZAT, S. ROUSSE, C. KISSEL, P. YIOU, J. EIRIKSSON, K.L. KNUDSEN, E. JANSEN and J.L. TURON.

2008. Decadal variability of sea

surface temperatures off North Iceland over the last 2000 years. Earth Planet. Sci. Lett. 268: 137–142. SICRE, M.-A.; I.R. HALL, J. MIGNOT, M. KHODRI, U. EZAT, M.-X. TRUONG, J. EIRÍKSSON and K.-L. KNUDSEN. 2011. Sea surface temperature variability in the subpolar Atlantic over the last two millennia. Paleoceanography, 26: PA4218, doi:10.1029/2011PA002169. SIMMONS, C.T. 2007. Let there be light: stained glass and the late medieval climate transition. Montréal, QC: Centre for Climate and Global Change Research Report 2007-1 (available www.esmg.mcgill.ca/Chris. htm). SIMMONS, C.T. 2008. Fiat lux: climatic considerations in medieval stained glass aesthetics. MSc Thesis, McGill University, Montreal (available from www.esmg.mcgill.ca/Chris.htm). SIMMONS, C.T. and L.A. MYSAK. 2010. The transmissive properties of Medieval and Renaissance stained glass in European churches. Arch. Sci. Rev. 53: 251–274. SIZOV, A.A. 1997. The response of cloudiness in the Atlantic-European section to the interannual variability of the North Atlantic Oscillations. Phys. Oceanogr. 8: 261–267. SWINGEDOUW, D.; L. TERRAY, C. CASSOU, A. VOLDOIRE, D. SALAS-MELIA

and J. SERVONNAT. 2011. Natural forcing of climate during the last Millennium: fingerprint of solar variability. Clim. Dyn. 36: 1349–1364. TRIGO, R.M.; T.J. OSBORN and J.M. CORTE-REAL. 2002. The North Atlantic Oscillation influence on Europe: climate impacts and associated physical mechanisms. Clim. Res. 20: 9–17. TROUET, V.; J. ESPER, N.E. GRAHAM, A. BAKER, J.D. SCOURSE and D. C. FRANK. 2009. Persistent positive North Atlantic Oscillation mode dominated the Medieval Climate Anomaly. Science, 324: 78–80. TROUET, V.; J.D. SCOURSE and C.C. RAIBLE. 2012. North Atlantic storminess and Atlantic Meridional Overturning Circulation during the last millennium: Reconciling contradictory proxy records of NAO variability. Glob. Planet. Change, 84–85: 48–55. WACHS, W.C. 1964. A historical geography of medieval church architecture. Published Thesis, University of Cincinnati. WARD , G . 2007. Photosphere. Retrieved from www.anyhere.com WARREN, S.G.; R.M. EASTMAN and C.J. HAHN. 2007. A survey of changes in cloud cover and cloud types over land from surface observations, 1971–96. J. Clim. 20: 717–737. YAN, Z.; P. ALEXANDRE and G. DEMARÉE. 1995. Some climatic scenarios in continental western Europe based on a dataset of medieval narrative sources, 708-1426 AD. In International Conference on Past, Present and Future Climate (Proceedings of the SILMU conference held in Helsinki, 22-25.8.1995), Helsinki 193–194. YAN, Z.; P. ALEXANDRE and G. DEMARÉE. 1997. Narrative warm/cold variations in continental western Europe, AD 708-1426. Sci China (D), 40: 509–517. YAN, Z.; G. DEMARÉE and P. ALEXANDRE. 1998. Wet vs. dry variations in continental western Europe, A.D. 708-1426, as deduced from medieval narrative sources. Workshop on Calibration of Historical Data for Reconstruction of Climate Variations, Barcelona, Spain, 6–8 July 1998. pp. 12–13. YOUNG, G.H.F.; D. MCCARROLL, N.J. LOADER and A.J. KIRCHHEFER. 2010. A 500-year record of summer near-ground solar radiation from tree-ring stable carbon isotopes. The Holocene, 20: 315–324. YOUNG,

G.H.F.;

D.

MCCARROLL,

N.J.

LOADER,

M.H.

GAGEN,

A.

and J.C. DEMMLER. 2011. Changes in atmospheric circulation and the Arctic Oscillation preserved within a millennial length reconstruction of summer cloud cover from northern Fennoscandia. Clim. Dyn. doi:10.1007/s00382-011-1246-3

J. KIRCHHEFER

ATMOSPHERE-OCEAN 50 (2) 2012, 219–240 http://dx.doi.org/10.1080/07055900.2012.667387 La Société canadienne de météorologie et d’océanographie