1. Introduction and goal

Environmental history intersects both the social and natural sciences. It is a meeting place where narratives about historical populations--their quarrels, movements, cognitions, etc.--are connected to stories about the historical ecology of their environmental context. Although the study of past climates is an old pursuit, integrated research on the interactions that make up historical population ecology is relatively new and efforts integrating climatic factors at a regional or local, landscape level even more scarce. The French Project is an interdisciplinary research effort aimed at understanding the multiscalar, evolutionary and historical development of the southern Burgundian landscape. From the beginning of its life, this longtitudinal research program has been interested in the role climate played in the mutual, dialectical interactions that exist between historical population and their ecology. Climate-culture relations were traced back to beyond the Roman Climatic Optimum, fueling understanding of the link between state hierarchization and eventual collapse in the face of climatic change. As Crumley notes (1994), Roman patterns of settlement and land use contrasted with the Celtic polities as they were especially suited to the mediterreanized climate of Europe from 300 B.C. to 300 A.D. During this time, continental Europe experienced a markedly warmer and drier climate, while reducing the frequency of extreme climatic events. In the context of these long-term, climatic changes, the dynamical and flexible organizational relationships that underlie Celtic (agro)pastoral societies and allowed them to thrive in the face of unstable European climate have commonly been obscured in cultural evolutionary theory. The lesson learned is the importance of including climate into the historical equation--complex and multiscalar--that shaped the physical landscape as we encounter it today. Burgundy proved to be the ideal place to study the effects of climate heterogeneity.

Described as the "European climatic broker between east and west, north and south" (Crumley & green, 1987, p. 28), the varied historical climatic conditions affecting Burgundy make it a particularly interesting place to study landscape evolution. The role of climate in the meshwork of factors that interacted to shape the present landscape are presumed to be the result of an incidental, but self-organizing sequence of multiple causal events, creating critical historical conjunctures that play out the process of history Lennihan 1982; De Landa, 1999). However, it appears that the closer one gets to the phenomenological live world of humans and their perceived cultural landscapes, the more obscured and unknown the influence of climatic factor becomes. When cultural evolution is conceived as the result of a complex history of localities interacting amongst and within, microclimatic studies at small spatial and temporal scales are needed to inform us about the contextual environment. Such studies remain relatively rare compared to macro-climatic studies over longer time periods, and this seems mostly due to the lack of integrated, and detailed local data sets. Furthermore, climate modellers seem to often ignore the historical influence of humans on climate regulation, while tending to prefer modern high-tech data over fuzzy historical evidence. Fortunately, links between macro and micro-climatic scales are increasingly being to counter such biases (Gunn & Crumley, 1991; Gunn et al. 1995; Ierland et al., 1996). From these studies it can be learned that frequency of climatic change, duration, and rapidity are important factors that impact human adaptability. As Crumley exemplifies (1994b),

In the remainder of this paper, five topics will be briefly discussed. First, the climatic situation in Burgundy will briefly be sketched. Second, the data obtained from Switzerland will be assessed for their association with Burgundy. Third, wine harvest data are reported on. Fourth, a (yet failed) attempt at climatic analysis of treering data (dendrochronology) will be described. Finally, some words will be said about the river discharge water obtained from Mechet Creek. In addition to this, an extensive climatic annotated bibliograpahy will be provided for further references. Together, these sections provide an initial gateway into the world of Burgundian climatic history.
 
 

2. Climatic Investigation: Burgundian dialectics

In search for an environment characterized by heterogeneity, the research team found itself attracted to the unique climatic conditions which form the Burgundian landscape. Located in East-Central France, Burgundy is situated in the intersection of our climatic zones. This is illustrated in the "Figure 5" below.

Because of this situation, Crumley and Green (1987) suggest that "Burgundy ensures the transition between the rhythm of Mediterranean precipitation and summer dryness (as at Marseille) and the style of northern Alsace in the peak of summer (Strasbourg)" (p. 28). In general, the dominance of oceanic and Mediterranean wind patterns brings moisture to Burgundy, while dominating continental winds are low in humidity. Low temperatures in winter months are due to the dominance of higher-latitude (oceanic and continental) upper air currents as the jet stream shifts south, while in summer Mediterranean and mountainous regions dominate when equatorial superheating occurs. Thus, the shifting location of the Westerlies (jet stream) is important in determining the general regime. In addition, differential heating by the sun of the earth's surface strengthens or weakens the three systems relative to one another to produce frontal (daily, weekly), seasonal, and yearly variations. Other factors also impact the mix, including orbital movements, volcanism, anthropogenic trace gasses, etc (Crumley, 1994b). The influences of such macroclimatic circumstances create an astonishingly variable weather pattern at lower scales of interest.

When one traces this pattern back through time, the region's enduring central position between north and south is reflected in its dialectical history (Guinot, 1987; Crumley, 1994). For example, Crumley has drawn up a map of the shifts in ecotonal positions from 1200 B.C. to the late Holocene. This can be seen in the "Figures 8.2-8.4 below.

From this map, the macro-climatic situation of Burgundy is sketched. The last of the four pictures shows the ecotonal border as it has been shifting during the Late Holocene. The focus of the climate data in this paper is set within this climatic regime. Today, the transition between southern and northern climatic regimes can be seen at the crest of the Montagne Noir, at the southern edge of the Massif Central, where the vegetation changes in less than a meter from Mediterranean to temperate flora.

3. Switzerland: a visit to Dr. Pfister as a first step

To asses the specific climate of Burgundy, it appeared that a goldmine of historical climatology was located 100 miles east of Burgundy in the historical laboratory of Dr. Christian Pfister--Department for Economic, Social and Environmental History, Institute of History, University of Berne, Switzerland --who has made it his life-quest to publish on European historical climate, including many detailed analyses of the history of climate in Switzerland. We obtained a climatic database that was published online (Pfister, 1993; ftp://medias.meteo.fr/paleo/historical/switzerland/), and includes the Graduated Indices [+3 ,-3] for both temperature and precipitation from 1525 to 1989. Recently, Pfister published a book called Wetternachhersage: 500 Jahre Klimatvariationen und Naturkatastrophen (Climate retrodiction: 500 years of climate variations and natural catastrophes, 1999: see http://www.cx.unibe.ch/hist/fru/fru-wett.htm#abstract ; Davis -- CALL NUMBER: QC989.S88 P46 1999). In this book Pfister reports on all pronounced anomalies back to 1500 and an oversight of ground pressure fields in Europe that give a location of anticyclones and cyclones and the direction and origin of the dominant air flows. This book is somewhat of a remarkable book in its depth. As Pfister himself describes it:

Table 1. Switzerland and Geugnon correlations in temperature and precipitation over 1949-1989
 

Month	Temperature  Switzerland (1949-1989)	Precipitation  Switzerland  (1949-1989)	Temperature  Basel  (1949-1998)	Precipitation  Basel   (1949-1998)
January	0.83	0.72	0.89	0.71
February	0.88	0.73	0.96	0.8
March	0.91	0.77	0.92	0.63
April	0.91	0.70	0.90	0.64
May	0.91	0.65	0.92	0.49
June	0.83	0.59	0.87	0.56
July	0.88	0.58	0.93	0.59
August	0.76	0.65	0.87	0.53
September	0.88	0.69	0.93	0.63
October	0.88	0.67	0.91	0.72
November	0.79	0.73	0.86	0.79
December	0.90	0.75	0.92	0.74

All significant p<.01 unless otherwise stated

From Table 1 it appears that the relationship of association between both Swiss data sets and the Geugnon (Burgundy, Saone-et-Loire, France), is quite high, for temperature, but lower for precipitation. For ease of interpretation this relationship is graphed in Graph 1 below. This graph shows that for precipitation the correlation was especially low in the summer months (May, June, July, August), while for temperature is seems to be lowest for August. This might suggest an enhanced southern influence in Saone et Loire compared to Switzerland.

This graph illustrates again that the relationship is close, but not perfect. Further, Basel has less precipitation in the winter than the average Based on these data though, it seems warranted to suggest that the data for Switzerland as a whole are not significantly different from those of Basel over the second half of the 20^th century. Based on this, it seems that the conclusion can be drawn that the Swiss data on pronounced anomalies back to 1500 provided by Pfister are useful but not perfect proximates for the climatic situation in Burgundy--including Geugnon--as well.

The different data sets of Pfister have been show in the appendix. Appendix1 shows the index for temperature Switzerland 1525-1989 and Appendix 2 shows the index of precipitation. These two appendixes contain the Graduated Indices GI [+3 ,-3] for both temperature and precipitation from 1525 to 1989. The first column contains the year, followed by the monthly temperature indices. Then there are five columns with indices for each season and the year. Precipitation indices are provided in the same order. The Graduated Index GI [+3,-3] is bound on a high density of data and on the availability of quasi continuous proxy data from historical and natural archives that are calibrated within the period of instrumental observations. The reference period is 1901-1960. Winter temperatures are estimated from hydrological indicators such as the proportion of the number of snowfalls relative to that of rainy days, the duration of snow-cover and the freezing of lakes in the Alpine borderland and observed signs of vegetative activity. For the spring and summer season the focus is on biological indicators such as dendro-climatic data, phenological observations, para-phenological indicators such as grape or vine harvest dates, as well as on continuous quantitative data on the volume of vine harvests, but also on snow-falls in the Alps. Precipitation is estimated from the number of rainy days obtained from weather diaries and from evidence on floods and low water tables of large rivers and lakes.

Monthly, seasonal and yearly indices may take the following values: + /3 for very warm or wet, cold or dry anomalies respectively. Anomalies are defined as <1st duodecile or 11th duodecile according to the 1901-60 distribution for temperature. The precipitation index is based upon several instrumental series of both sums of precipitation and number of days with precipitation 0.3mm (see Pfister 1984 for details) + /-2 for warm or wet, cold or dry months, respectively + /-1 for months with temperature or precipitation above or below average 0 for "average" months or data not available On a seasonal level the GI is defined as the average of the monthly GI, which yields gradations of 0.3 between -3 and +3. Anomalies correspond to graduated indices of = + 2.3 and <=- 2.3, respectively. On a yearly level the GI is defined as the average of the monthly GI. The entire documentation which underlies the indices is titled CLIMHIST-CH, and is available on paper or microfiche from METEOTEST, Fabrikstr. 29 a, CH 3012 Bern Switzerland.

• Data: Swiss temperature 1525-1989 based on Pfister (1993) in Excell
• Data: Swiss precipitation 1525-1989 based on Pfister (1993) in Excell

• Data: Geugnon, France, precipitation (mm.) and temperature (celcius) 1949-1998 from Meteo France, and
• Data: Basel, Switzerland, precipitation (mm.) and temperature (celcius) 1900-1998 from Meteo Switzerland (SMA-MeteoSchweiz, Kundendienst Bodendaten). These data sets are not shown in the Appendix.

• Decennial number of monthly temperature anomalies per decade 1496-1995 in Southern Central Europe
• Anomalies according to type and period
• Anomalies according to type and month ("kalt-trocken"=cold-dry, "feucht"=humid)
• Amount of anomalies in Winter 1496-1995
• Amount of anomalies in Spring 1496-1995
• Amount of anomalies in Summer 1496-1995
• Amount of anomalies in Autumn 1496-1995

• Highwater of the Rhein at Basel since 1496 ("Uberschwemmungen"=flood, "schwere"=heavy)
• Extremely highwater of the Rhein at Basel: 1500-1994
• Winterstorms (western winds from October to March) 1500-1994

• General representation of the phenomenon of the Northatlantic oscilation and possible influence on Alp region (+: "cooler northwest atlantic, Island deep and Azoren high, strong western winds, warmer in the alps/ higher pressure", "schwacher"=weaker, "tiefer"=deeper).

Excellent records of grape harvests have been kept for the wine growing regions of France. All other factors being equal, late harvest dates are indicative of a vine growth period (March-April to September-October) during which average temperatures were mostly cold. Early harvests, on the contrary, indicate relatively high average temperatures during the seasonal intervals. In 1981 Emmanuel Le Roy Ladurie and Micheline Baulant published a synthesis of all 1981 known harvest data series for the vineyards of northern and central France (Paris, Burgundy, France-Comte), Switzerland, Alsace, and Rhineland. Together the series represents a weighted average of 103 local sites.

• Data: Wine series in Excell
• Graph of wine series

For the wine data in Europe, it appeared that the archival vineyard series for Dijon are the most complete series that exist. Unfortunately though, these data could not be obtained from the published series by Ladurie and Baulant, since they were synthesized together with many other wine data sources and averaged out. Because Ladurie published a graphed version of the Dijon data in an earlier book--"Histoire du climat depuis l'an mil", 1967--it was however possible to see the relative difference between some sites that compromise the total synthesis. The scanned in picture is presented below:

Appendix 4 shows the large version of this data. It can be seen that the variations are mostly differing in order of magnitude, while the general pattern between Dijon and the Average (lowest time series) is similar. Ladurie and Baulant remark that the harvest in Salins took place on average 12.2 days later than Dijon. It seems however recommended that the original Dijon data be obtained, since these differences may be further accentuated depending on other factors (since in reality, these other factors are not held constant). Monsieur Ladurie informed us kindly that he did not have the raw data, and that it was Monsieur Baulant who possibly still had them. Unfortunately, Monsieur Baulant has not been found yet. Ladurie and Baulant did report a series of correlations of the Cote d'Or Region (Dijon, Volnay, Beaune) with some other regions. These can be seen in Table 2 below:

Table 2. Correlation coefficients of beginning grape harvests series in Cote-d'Or (Dijon, Volnay, Beaune) and other wine growing regions by time periods
 

1790-1879: "a number of excellent correlations"	Correlation Coefficient
Franche-Comte	0.82
Seine, Sein-et-Oise	0.82
Marne, Seine-et-Marne	0.83
Meuse, Meurthe-et-Moselle, Vosges	0.87
Aube, Yonne	0.92
Sarthre, Loire-et-Cher	0.83
Germany	0.89
1690-1810: general poorer than 1790-1879, except:
Aube, Yonne	0.82
1590-1710: general poorer than1790-1879, except:
Franche-Comte	0.87
Seine, Seine-et-Oise	0.80
1590-1710: example of relatively low correlation
Switzerland	0.58

Ladurie and Baulant mostly provide the "excellent correlations", but not much is mentioned about the nature of the low correlations. It also appears that the earlier the period, the less strong the correlations. It is not clear if this is due to technological, climatological, or measurement changes. Bell (1981) addresses this "general retardation" by referring back to Ladurie (1967), who apparently suggested that it is "a sign of viticultural and not of climatic revolution… [because] in order for the delay to take on meteorological significance and indicate a long term climatic movement, it has to occur in all vineyards at one."--and the latter is not the case. A statistical comparison of the 103-series grape harvest data provided in 1981 and earlier data from Ladurie (1967) for different locals, a series of six stations in the central-septentrional (Dijon, Salins, Bourges, Lausanne, Aubonne, Lavaux), shows a correlation of 0.95. We have therefore discarded the 1967 dataset.

Working with the vineyard data, Bell divided her data into two temporal regimes to overcome this problem. She suggested that based on trial calculations the data synthesis of the 103 locations reported on shows a viticultural transition around 1700. Using the harvest data and further including data on vine yield and wine quality, Bell, using a method of cumulative deviation, identifies clusters of years tending to above-average and below average spring-summer temperatures. According to her, Pfister (1981) has shown that harvest date, vine yield, and wine quality indicate temperature variations of the early, mid-, and late summer, respectively, of record. The results of Bell's analysis are shown below in Table 3.

Table 3: Clusters of Years Tending to Above-Average and Below-Average Spring-Summer Temperatures (after Bell, 1981).
 

Warmer Intervals			Cooler Intervals
Apr-June	June-Aug	Aug-Sept	Apr-June	June-Aug	Aug-Sept
Early harvest dates	High vine yields	Good wine quality	Late harvest dates	Low vine yields	Poor wine quality
					1454-60
		1473-84
			1488-93	--	1488-92
| 1500-04*		| 1493-1510		--
| 1515-24		| 1517-23
|		|		--	1526-29
| 1530-40		| 1531-41
|		|	1541-44	1542-44
| 1545-62	1545-59	| 1543-53
			| 1563-70		| 1559-66
			| 1573-82	| 1576-79	| 1568-74
			| 1585-87	| 1585-1602**	| 1577-1602**
			| 1591-97
			| 1600-01
		1610-20
			1616-35	1618-29	--
| 1636-39	| 1634-38	--
|	|		1640-43
|	|		1648-50	1647-52
| 1651-71	| 1653-72
|	|		1672-75	1673-75
| 1676-86	| 1677-84
			1687-1703	1685-1701
			1713-1716	1701-17
1718-37	1718-30
			1740-45	1740-50
1757-62	1760-65
			1767-77	1766-73
1778-84	1774-84
				1786-90
	1791-98
				1799-1801
	1803-08
			| 1812-24	1813-21
	--		| 1835-38	--
			| 1850-56
1857-70***

*     =The more prolonged anomalies of one sign, with minor interruptions of the opposite sign, are marked by vertical lines.
**   =The period of first documented damage from growing Alpine glaciers of little ice age in 1585-1602
*** =The start of the major alpine glacial recession and the end of little ice age is 1857-1870.
 

5. Treering data: moving closer to Burgundy?

5.1 Notes on methodology

The goal of the attempt however was to move the climate observation to Burgundy itself, despite the high correlations with the Switzerland data. In the context of the French Project, it is assumed that both regional and landscape level microclimatic influences are at play in the historical time series. Fortunately, G. Lambert C. Lavier and Y.Trenard at the University of Becancon have compiled a data base of dendrochronological data of trees in Burgundy, called the "Burgundy 29". This database has been available online at (ftp://ftp.ngdc.noaa.gov/paleo/treering/chronologies/asciifiles/europe/fran029.crn).

• Data: Burgundy 29 extracted file, incl. sample size, 681 A.D - 1991 A.D. (G. Lambert C. Lavier and Y.Trenard)

Based on this information, deductions can be made on departures in climate. The above graph however can only be seen as an unrefined representation of climate. The extent to which climatic factors affect the rate of growth differs among ring width chronologies so that there is variation in the relative importance to growth of temperature and precipitation from one month to the next throughout the year. Fritts and al. (1981) suggest that climate for given year t may include factors such as temperature, precipitation, sunshine, and wind, and these vary in their importance to growth throughout autumn, winter, spring, and summer. The growth layer is formed in spring and summer and may be directly affected by climatic factors at that time. Stem growth is dormant during autumn and winter, but high and low values of climatic factors can have important effects on soil moisture recharge, the making and storage of food in the tree, and the growth or roots. In addition, Fritts et al. point out, there are many lagging effects of climate which appear in the growth ring of the succeeding years t+1 through t+k. The climate in the preceding year controls the formation of leaves and roots along with ring width which is in turn related to the efficiency of the tree in the following years. This produces an "autocorrelation" in the chronologies: narrow rings tend to be followed by narrow rings, even though the climate of the second year was favorable to growth. Fritts (1979) suggests that lack of response can amount from several months to 15 or more years between climatic input and corresponding tree ring output. This process is illustrated by Fritts et al. below:

Fritts (1973) suggests that the degree of the relationship between climatic factors and ring-width variation dependent upon the ecological amplitude of the species, the proximity of the sampled trees to their environmental limits, and the range of variability in the limiting factors affecting growth. The goal for the dendrochronologist is to estimate the ringwidth variation due to climatic factors--the signal--and the ringwidth variation due to noise. In moist sites and mild climates such as Burgundy, the growth-controlling processes will be less limited by conditions associated with variations in climate relatively to stress sites, but they will instead be limited by nonclimatic conditions such as shading by neighbors, lack of soil minerals, size of crown, and tree age. In such cases ring widths will vary widely from one tree to the next and only a small amount of variation will be common to all trees of a given site. This common variation, although it is less precisely defined, represents potential information in the macroclimate. Thus, the "signal-to-noise ratio" is small in this circumstance (and large in stress sites).

Response functions refer to the weights or coefficients used to estimate tree ring growth from variables from climate. The weights describe how the tree "responds" to climate." In the context of the French Project, what is needed is however the opposite: the estimation of climate based on tree ring data. The transfer function refers to a different set of weights used to estimate climate from ring-width values. In this way, growth records are transferred into reconstructions of climate (retrodiction is a less appropriate term than reconstruction when tree ring data are used). Both response and transfer functions are created through "calibration." Calibration is the process by which the instrumental treering reading is compared to known weather conditions. The purpose of this comparison is to understand the relationship between the ring growth (as dependent variable) and the climatic indices (as independent variables), often monthly precipitation or temperature, and to create a statistical regression function with which the climatic situation can be reconstructed. The success of such calibration is measured as the percentage variance accounted for. This success if often described as "goodness of fit," which provides an indication of the strength in the signal data set as well as the adequacy of the model in describing the relationship. Fritts sketch of the model of statistical calibration can be seen below.

Important data considerations in this process are sample size, and the use of an independent set of data for verification. Sample size is an important factor determining the goodness of fit. Growth data from a single tree may include a signal of climate accounting to only 25% of the total variance. If several trees are sampled and the mean growth is obtained, the variation of signal which is the same in every tree is not lost: some of the noise is cancelled. Thus, the consequential "Law of Large Numbers" simply states that as the numbers of individual trees sampled in a mean chronology increases, and as the number of radii used from each individual tree increases, the amount of variability due to noise in the mean chronology will decrease because there is less error in the means of the larger sample. Fritts (1976) mentions that an increase from 5 to 10 trees will lead to a far greater increase in signal and reduction of noise than from 15 to 20 trees. For the Bourgogne 29 master sample size does provide some problems at earlier years, where sample size is under ten for the period AD681-AD924, and the years AD 1091-AD1197. The sample sizes are plotted below in Fig. 2

Verification of the transfer function obtained through calibration is done with a different set of independent data. Fritts argues that verification is as important as the reconstruction, especially when complex calibration questions are utilized and the size of the dependent data set is limited.

5.2 Results: trying to be a statistician

Despite all the promising talk by Fritts and his colleges who report squares of multiple correlation coefficients of around 0.80, for this paper the attempts to create a transfer function remained a failure. This is undoubtedly a result of a lack of statistical sophistication on behave of the author. The challenge taken up was to construct a transfer function that would suggest climate and temperature from the tree ring data. Factor analysis was used to reduce the months of data without success (factor analysis transforms correlated variables into a new set of orthogonal or uncorrelated variables called eigenvectors or principal components). Many suggestions made by Fritts (1976) were unsuccessfully followed. Figure 3 below illustrates the challenge.

From this graph it can be seen that the data are non-linear and chaotic. The multiple implications of delay in response, the interaction of temperature and precipitation, and the lack of other data made this exercise quite a challenging one--or even too much so. It appeared that for the period 1950-1989 the correlation between the precipitation and temperature data of Geugnon with the treering indices were low to nonexistent: 0.262 for average yearly temperature and -0.097 for yearly precipitation. When decomposed by months, it appeared that the temperature of October and November had low significant correlations (0.31 and 0.32 respectively), and not at all for precipitation in any of the months. In an attempt to further look at this, it appeared that for the Swiss data over the period 1525-1986, the precipitation of summer months correlated significantly but low at r=0.32 (p<.00) with tree ring growth. Although this analysis is surely not finished, in the end it appeared that the only results were the meager findings of a transfer function of tree ring growth explaining 0.114 squared R for combined precipitation in March through July (precipitation = 176.78 +100.7 * Treering indices; constant p<.00, Beta p<0.05).

For temperature treerings indices explained R-square of 0.18 for the months October and November together (temperature = 21.816 +3.758 * Treering indices; constant p<.00, Beta p<0.0). For temperature treerings indices explained R-square of 0.18 for the months October and November together (temperature = 21.816 +3.758 * Treering indices; constant p<.00, Beta p<0.0). I graphed the estimated and real values for the period 1950-1989 below in Figure 4 and Figure 5.

These statistical equations show the normalizing effect of linear regression in the case of the precipitation indices, and a substantial overestimation of temperature. This is probably a mistake made somewhere. Most importantly however remains the observation that not much--if anything at all--of the variance of the 1950-1989 climate data is explained by these equations based on tree ring data.

6. Water: Mechet Creek and Lac du Bouchet

Gunn has for several years now been investigating the use of river discharge flow for assessing precipitation (Gunn & Crumley, 1991; Gunn & Folan, n.s.). From this iterative research it has been suggested that some rivers respond strongly to changes in atmospheric temperatures, while others do not. One of these rivers that do seem to respond is Mechet Creek, in Saone-et-Loire, located about 30 miles north of Geugnon. Discharge was measured as monthly averages of weekly readings at the D296 bridge 2.5 km southwest of Monthelon. A gauge on the downstream side of the bridge which forms a rectangular discharge 4m. wide. The bottom of the gauge is 0.6 meter above the bottom of the stream.

• Data: Water dicharge file in excell

Table 4. Correlations between Mechet Creek Discharge data and Geugnon precipitation 1969-1986.
 

Month	Correlation
February	0.797793*
March	0.676472*
April	0.068692
May	0.309235
June	0.404755
July	0.422053

*=significant p<.01

From this it appears that precipitation in Geugnon only correlates with the river discharge of the Mechet in February and March. For the other months this relationship disappears. This effect might be due to the coming of summer and shift in climatic conditions between the two sites, plus the higher elevation of the Mechet, but remains somewhat unexpected. The relationship is even less pronounced for Basel.

A final source of climatic data are the lake levels available from Lac Du Bouchet (Bonifay and Creer, 1987), which is an extensively studied volcanic crater lake 200 km south Mechet Creek. These lake core analyses however are also very crude in their measurements, and not necessarily subtle enough for the temporal scale sought for here. In their conclusion, Gunn and Crumley do suggest that these lake levels are in reasonable correspondence with the retrodiction model based on the global energy balance and volcanic index. It seems thus warranted to further take into account the use of all these indexes when dendrochronological data and other data sources are combined in a statistical reconstruction.

7. Conclusion

The difference between retrodiction and reconstruction is an important one. While it seems possible to create a statistical model based on different indexes, further retrodiction in the past is not the same as reconstructing actual condition based on historical evidence. It is the latter which is the goal here, based on empirical evidence. The task ahead is to better make use of non-linear statistics and understand dendrochronological methodology to make use of these data. For this, it might be helpful to obtain more information on the data set itself. Another suggestion would be to try to find other data sets from MeteoFrance with which to calibrate the tree rings. However, the lack of results seem more due to statistical flaws than data flaws. In the meantime the comprehensive data sets provided by Pfister and colleges on Switzerland, the vineyard data obtained by Ladurie and Baulant, and the raw treering data still show trends that can to a certain extent approximate climatic history as it was played out in Burgundy. For now then, "micro-climate" remains at the "meso-level."
 
 

8. References