CAUSATION VERSUS CORRELATION

The real question that must be answered first is not if there has been a climate change in 3^rd and 4^th century AD central Europe, but how would sustained climate change affect the indigenous people?  The root of this question is what is the difference between causation and correlation?  If a climate change and the migrations occur simultaneously, how can one establish that there is a causal relationship or simply a correspondence between the two?

The issue of correlation versus causation is an intriguing question with only one real answer: correlation can never be completely refuted.  It is impossible to refute that two things occurred at the same time when evidence clearly states that they did.  So, the only part of the issue that remains is the degree of probability that there is a causal relationship.  The first step is to examine how a causal relationship between climate change and migrations of peoples can be made.

            If James Hutton's (1785) credo is correct that the key to the present is the key to the past, then examples in the present are of paramount importance because examining the past would be fundamentally impossible.  This concept, usually called uniformitarianism, is essential in the earth sciences for furthering our understanding of the Earth.  If scientists cannot except that the physical laws that apply in the present are the same that were applicable in the past then studies of the past would be insupportable.  This is slightly different in the fields of history and archaeology.  Historians see humans as reflection of both the past and future (Salvesen, 1992) .  Natural sciences are forced to only reflect on the past.  Historians and anthropologists realize “patterns of human activity and social change may even be determined by causes supposed to occur in the future, or by causes which never appear” (Salvesen, 1992) .  Causal effects are difficult to apply to human beings, unlike biological or chemical laws.

The droughts of the central United States in the 1930s, 1950s, 1980s, and 2000, the Great Pueblo Droughts from the 12^th century AD to 19^th century, the Lost Colony drought of 1587-9 and the Jamestown drought of 1606-12 have been evaluated for causal effects (Wigley et al., 1981, Slatter, 1979, Stahle et al., 1998) .  These studies illustrate not only the different opinions, but also the various approaches to prove or refute causal relationships between climate and human migration.

If we look at two different decades, the 1930s and the 1950s, there is clearly a significant drought occurring in the USA in these two periods as evident in both the Palmer Drought Severity Index (PDSI) and local records (Figures 1.6 and 1.7).  However, popular culture is more familiar with the Dust Bowl, the common name for the droughts of the 1930s, than with the unnamed droughts of the 1950s, even though the drought of the 1950s is arguably more widespread than the 1930s.  Why is this?  The difference can only be sorted out through the understanding of historical perspectives.
            In the 1930s, the United States was in an economic depression, as was most of the world.  The farmers of this period practiced poor land management, such as plowing dry fields and not utilizing soil conservation techniques, which was compounded by large quantities of particulate matter in the air due to these
practices.  The economic depression did not allow the government agencies (many of which did not exist until the Hoover and Roosevelt administrations) to assist the failed farms or the farmers.  In short, the conditions of the 1930s were in an economically poor and the amount of particulate matter blowing through the mid-western states was photogenic.

In the 1950s, the United States experiences an economic boom.  The post war society is very stable and many governmental institutions are in place.  Further, inventions, such as the diesel pump for the Ogallala aquifer, were available to the farmers to utilize irrigation.  The farmers also have access to Agricultural Colleges that were studying the role of drought in the region and how to overcome these obstacles.  The US government in the 1950s is capable of subsidizing the drought-effected areas and there is little migration directly due to the drought opposed to the droughts of the 1930s.

Although shorter in duration, the PDSI shows more area being affected in the 1950s than in the 1930s (Figures 1.6 and 1.7).  If we were to examine this period using only paleoclimate proxies, the 1950s droughts may appear to be the more intense of the two periods, depending on where and what the proxies were.  For example, the tree-ring chronologies around the central USA tend to react more, or show a more distinctive pattern in their ring widths, in the 1950s droughts than the 1930s droughts, especially in Oklahoma and Arkansas.  Further, there are drought periods in the late 1980s and in 2000, but these too occur in an economically prosperous time as did the 1950s droughts and are not noted as “influential”.  Arguably, these droughts could appear as more or equally important in the interpretation of the paleoclimatological data and it may further be presumed that the people who lived in the same area should have suffered accordingly when this is simply not the case.  We are capable of identifying this disparity because the historical perspective is understood.  The historical perspective needs to be understood or there is no possible way to ascertain the difference between correlation and causation.

There have been many discourses surrounding the idea of causation.  W. Meinke argued that the underlying foundation of the study of causation is the search for values, either directly or indirectly (Meinke, 1931).  Without the understanding of the values of the people in question, determining causation is nothing more than guesswork.  T. M. L. Wigley's stated, “the impact of climate fluctuations and change has long been recognized as one of the factors that demand consideration, but there has always been dispute over how much attention needs to be paid” (Wigley et al., 1981,5) .  The attention that needs to be paid concerning climate depends entirely on the values and understanding of the people in question. 

J. L. Anderson (1981) defines five possible responses of a people group to climate change: extinction, adaptation, endurance, migration, and technological advance.  The current bandwagon seemingly is to attach migration to almost every climate change period.  In the case of the 1930s and 1950s comparison, technological advance plays an important role in the difference between the two, as does adaptation.

The studies that do exist on migration-climate change are becoming more extensive and can provide a blueprint for methods that can help assist in the understanding of other similar events.  Erwin Slatter writes about the Great Pueblo droughts that occurred predominantly between the 12^th century and the 20^th century (Slatter, 1979) .  This particular study has become a blueprint of several other studies and, more importantly, several accepted assumptions that may or may not be correct.

Slatter worked with both historical and prehistoric evidence claiming the relative comparison between historical and prehistoric responses is a safe assumption to make.  This assumption is based upon Hutton and Lyell and the principle of uniformitarianism.  Uniformitarianism is the basis of dendrochronology, palynology, and every paleoclimatological methodology.  It is the concept that given the same precipitation and the same temperature the tree/ plant/ snow pack/ glacier will response in very similar ways.  However, humans do not necessarily respond this way, yet this study and similar studies suggest, if not depend, on this assumption.

Slatter repeatedly uses human response in a standardized model to suggest that all correlation between population change and climate perturbations are causal (Slatter, 1979) .  His population model is a good example of uniformitarianism.  First, the tribal populations are reconstructed in historical times and compared to the contemporaneous tree-ring master chronologies.  The population is then inferred from room counts in archaeological sites and adjusted for any possible river erosion that may have claimed any housing.  The tree-ring record and population counts are given a one-to-one relationship, meaning that if “x” population moved at “y” tree-ring value in recorded history, then they would have done the same in the past (Slatter, 1979) .  This is a purely causal, uniformitarianism methodology. 

The argument that Slatter uses is, unfortunately, skewed due to the nature of the source of the historical data that is used.  The Hopi Indians apparently have an interesting relationship between their gods and their perceptions of climate.  The Hopi were afraid of their gods leaving them in times of good, because they believed their gods would possibly leave them if they were no longer needed (Titiev, 1944) .  This fear resulted in many records that stated that there were times of famine or drought even in times of plenty.  Slatter uses this argument to filter any recorded drought where the tree-ring record did not show any.  With this historical editing, the model then worked satisfactorily.  All dendroclimatologically recorded droughts before the 1700s were then assigned as migration periods, even without any historical records.  The correlation between historical migrations and tree-ring chronologies where concluded to be a causal relationship.

If we look at the Jamestown and Lost Colonies droughts of the 16^th and 17^th centuries (Stahle et al., 1998), we can see a similar influence.  The drought that started in 1570 has historical documentation by a padre, Juan Batista de Segura, who stated that the maize over the past 7 years had come short in the harvests.  This corresponds with the Jamestown colony failure.  The tree-ring record shows smaller rings in this period that was documented by Segura and this crop failure occurred at the time that the Jamestown colony almost failed.  The indigenous people suffered 1563 - 1570, so all humans must have suffered in the two subsequent droughts, Lost Colonies and Jamestown.  Clearly, this is a use of uniformitarianism and humans.  The Stahle et al. (1998) paper does remark about further documentation that suggest that malnutrition and poor water quality were also possible problems, which could easily be exacerbated by long drought.

The problem with claiming climate-induced causal effects on humans is that it is necessary to attach, not only a predictable human response, but also the concept of sustainable yield farming.  Sustainable yield farming is theoretically achieved when the minimum amount of caloric need of a community is met by their own agricultural efforts.  This is a requirement of any climate-related causal theory.  A climate threshold can be established numerically by calculating the minimum caloric intake needed to allow the community to continue without outside assistance.  This number can then be compared to a tree-ring chronology.  Through statistical analysis, these numbers can produce a number of events that would cross this threshold and could be determined as a climate-migration causal event.  However, if endurance or climatic conditioning of humans is applied then the models do not work, because this number cannot calculate human’s ability to adapt and disregards the nutritional value of the caloric intake.  This is especially problematic in the study of the Germanic tribes because Tacitus, a 2^nd century AD Roman historian, records that the Germanic tribes were conditioned to harsh environments and even starvation (Tacitus, Germania).  This is part of the reason that climate-induced migration theories have been easily disputed.

In the study of the Germanic tribes, examining both the possibility of correlative and causal effects of climate-induced migration is necessary.  With the data available, it is not possible to eliminate either possibility.  Ellsworth Huntington originally suggested causation in 1907.  In his book, The Pulse of Asia, he suggested that climate should be a large influence on these tribes as evident by their movement patterns.  Because they all began moving around the same period and seemingly abandoned their homes for other areas, this movement appears to be symptomatic of a people who were forced to look for more agriculturally productive areas.  The Germanic tribes abandoned their traditional homelands and left their original territory to weaker tribes, invaded areas that were more supplied and heavily fortified, and headed toward areas that were considered more fertile.  It is only possible to understand these events if the values and culture of the Germanic tribes can be reconstructed.  From this and high-resolution climate reconstruction, is it possible to determine the degree that the climate changes of the 3^rd and 4^th centuries AD are causal and not just a coincidence.