Reader Comments

Post a new comment on this article

Correlation does not imply causality - comments on the article by Kauppi et al.

Posted by rojansuu on 13 Nov 2014 at 10:07 GMT

Kauppi et al. (2014) analyze the annual increment of the growing stock of Finnish forests, which has increased from 49.5 million m3 in early 1960s to 103.9 million m3 in 2008. The authors claim that 28.4 million m3 of the growth increase is caused by the climatic warming. The result is based on the assumption that the regional increment variation in Finland is caused by the variation of the effective temperature sum. A simple regression line of regional level volume growth estimates on the effective temperature sum was estimated and interpreted as causality. The estimated regression line was then used to predict temporal changes in the regional growth.

Argumentation for the causality is weak, misleading and does not accommodate the complexity of the forest growth, neither the differences of the structure of the growing stock nor the variation in site fertility in different regions in Finland. Kauppi at al. (2014) presented two arguments for the claimed causality.

First, “it has been well documented that the inter-annual variation of growth in the boreal zone correlates with changes in temperature”. A known fact is that a statistical correlation cannot be interpreted as causality. Furthermore, the somewhat weak correlation between the temperature and the diameter increment of an individual tree does not describe the relationship between growth of a forest area and temperature.

Secondly, “the slope of regression is reproduced using two independent data sets”. The data sets are not independent. Both regression lines were estimated from the forest inventories based on the growth estimates of the Finnish forests in large regions. Although the inventories are independent, the regions are the same in both inventories. In both cases, the effective temperature sum decreases significantly from South to North. Also the structure of the growing stock and site properties change across the same gradient decreasing the increment in the North. This means that there are more or less parallel trends in temperature and several other vital growth factors, such as growing stock and soil properties. This can be found empirically using same sources of inventory growth data but replacing temperature sum on the x-axis in Fig. 3 with growing stock volume per hectare. The resulting new ascending regression lines have almost identical slopes and intercepts. Although forest stand growth correlates strongly with growing stock volume, these regression lines should also not be interpreted to describe a causal relationship.

The forest growth is a complex phenomenon and it is not possible to assess the effect of any single factor without taking into account the other factors. Forest growth depends strongly on forest management and in Finland the structure of the forests is a result of the past management practices. At least three significant changes increasing the productivity of the forests and forest growth have occurred during the study period: 1) regeneration of old and low-yielding forests to fully stocked forests has been one of the major forestry trends in Finland increasing the area of young and fast growing forests, 2) selective cuttings have been replaced by thinnings from below, and 3) about 5 million hectares of peatlands have been drained in order to increase forest productivity. In addition, the possible role of nitrogen deposition, which decreases substantially from south to north, was not analyzed.

We conclude that it is not possible to estimate the effect of the climatic warming on forest growth using the Finnish National Forest Inventory data without carefully analyzing all the factors affecting the regional growth.

Erkki Tomppo, Prof., D.Sc.(Stat.). 1
Risto Ojansuu, D.Sc.(For.). 1
Antti Ihalainen, M.Sc.(For.).1
Pekka Nöjd, D.Sc.(For.). 1

1 Finnish Forest Research Institute (METLA), Vantaa, Finland

No competing interests declared.