Does the Swedish Forest Area decrease?

Readers of a recent EPI-report from Yale get the impression that the Swedish Forest is not ecological vital, as the forest area in Sweden seems decreasing fast in an international comparison , where Sweden get score 14 on a scale where 100 is best and 0 worst based on an article in Science.

Below is a plain text version of a document available in a better format at

But I found that a better angle for attention was to make the bad scores a problem of countries at high latitudes, and not just Sweden, and present to the analysis at a joint Nordic-Baltic forest site.

That draw a response from the person responsible for publishing the data in Science, which EPI has used for calculating the low score of Swedish forest. He claimed that his data was not a scientific ground for the low score and that the data should not be used as EPI has done. Thus Swedish forest is not decreasing and where is no overlogging in Sweden.

Dag Lindgren 141009


Deforestation is an important global problem deserving surveillance and the Science article used as a basis is a good start of something with huge potential applications, so I am not too happy about the doubt of documentation of forest cover changes seeded in this document. But it can help both EPI and those involved in the matter of the Science article to develop a better product, and it can help Swedes to understand the reasons why international canopy cover result seemingly could be different than Swedish.

I have thought about the matter and think deeper penetration is justified and that something can be learnt, which will lead to better understanding and analyses.

Umeå 140905 Dag Lindgren ( )



EPI evaluates and publishes comparisons of countries for “ecosystem vitality” for forests and refers to a Science article. The results seem misleading and insufficiently supported. EPI refers to data in a Science article.

·         The Science article does not seem to deal with loss and gain of forest in a symmetric way. Thus changes in forest cover cannot be accurately calculated. The main author (Hansen) of the Science article seems not to support with the way EPI handles the data.

·         For countries (like Sweden) in the middle between best and worst, the “issue score” with its visual exaggeration make them appear unfairly bad in country-ranking.

·         The area of forest land in Sweden increases according to the Swedish expertise. If the Science based data are correct, they may reflect change in regeneration practice several decades ago or other changes, which should not be interpreted as recent decline in forest cover.

·         There is no good quantitative discussion about possible reasons for possible change in forest cover relevant for Sweden.

·         The calculations behind the conclusions are not transparent for me – and when probably not for anyone else. I just do not understand, and the key concepts – like percent canopy cover and gain – does not seem sufficiently precise.

·         The Science global map is quite fresh, it has uncertainties, where are doubts. It seems early to draw too wide conclusions from the Science map, before it is better understood what observations really means in real world. It is immature to apply for such controversial matters as country rankings for cases where country observations seem to contradict.

It is hard to find out exactly how EPI and Science have calculated and there are many uncertainties, e.g. exactly how canopy cover is defined and concluded, and to what degree loss and gain are symmetric. Even if the information in Science based on Landsat data on tree height and canopy cover should be correct, it is doubtful they give the most relevant information. There are important uncertainties. I am not aware of any really relevant discussion what the real causes of the low score of countries like Sweden may be. It would be better to wait spreading widely that Sweden loses its forest cover until the national expertise feel more able to understand why and until it is better understood if EPI forest score really is supported by the Science results. The results by Hansen et al 2013 could possibly be compatible with National observations and estimates, as there are differences about what is observed, but this might be better and more convincingly discussed before making international comparisons.



Environmental Performance Index (EPI) is issued annually. The current comment refers to: Hsu, A., J. Emerson, M. Levy, A. de Sherbinin, L. Johnson, O. Malik, J. Schwartz, and M. Jaiteh. (2014). The 2014 Environmental Performance Index. New Haven, CT: Yale Center for Environmental Law and Policy. This document concerns only forest. The Environmental Performance Index ranks country-level environmental performance based on scores in nine Issue Categories. Six of these issues are claimed to measure “ecosystem vitality”. One of those issues is “Forests”. EPI calculates a score for forest “ecosystem vitality” for each country. Forests scores were calculated in a new way this year. Countries are scored and ranked for assumed change in forest cover based on information in Hansen et al 2013 (M.C. Hansen et al. (2013). High-Resolution Global Maps of 21st-Century Forest Cover Change. Science. 342 (6160): 850-853.). A spectacular low score for Sweden for “ecosystem vitality” of forests – illustrated by an “EPI rose” magnifying the bad impression for Swedish forests – received my attention. I tried to understand the reasons. I found that it is very hard to follow the calculations, almost no EPI reader will understand more than that Sweden has a low score in “ecosystem vitality” connected to a considerable reduction in area with forest, and readers will believe this is a trustworthy fact as it is supported by a study in a well-reputed journal (Science). I doubt the results, and still more their relevance for “ecosystem vitality”. I am not alone with these doubts.


How does EPI derive country scores?

Country scores and ranks are the only numerical information easily – actually “too” easily – available from EPI. It takes a lot of effort and time to understand and figure out the derivation from EPIs web and the Hansen et al 2013 values, but I think I finally have got the major issues. I take it in an easy version first:

First EPI looks in an appendix to Hansen et al 2013, where the used values appears in a Table together with many other values and with entries difficult to understand. Then values are formed from two columns, one for “lossEPI” and one for “gainEPI” (I add sometimes EPI as it is an EPI decision how to interpret Science data). Values of (gainEPI – lossEPI) = column p – column n (in Table S3 below) = 6.4 – 10.5 = -4.1 (figures exemplified for Sweden) are presented in and called FORCH. Countries where FORCH is positive (and thus no forest cover reduction indicated) all get score 100. Countries where FORCH is lower than a value around – ≈8%, and thus the forest cover reduction is assumed to be larger than ≈8% get score 0. How scores for countries in between are calculated I have not bothered to understand.

EPI explains its calculation methods in . Data are said to be obtained from Table S2, but that is impossible as Table S2 does not have countries as entries, but ecozones. I assume Table S3 (see below) is meant.


Gain and loss in the Science article

It is said that Hansen (main author of Science article) thinks that gain and loss in the Science article are not symmetric. I try to read the Science article and the supplement and understand.

Citations from the Science article itself: The start of the summary is: “Quantification of global forest change has been lacking”; Opening of article itself is: “Changes in forest cover affect the”. From the opening paragraph: “Forest gain was defined as the inverse of loss, or the establishment of tree canopy from a nonforest state“.  It seems completely impossible for a reader to see, realize or even suspect that loss and gain are not symmetric from the “paper” article itself. Neither there seems to be clear reservations.

There is a supplement to the article, which can be obtained from the web on the URL given in the Science article. The base is landsat data which go through an elaborate, complicated and not transparent process. Citations: “Forest loss was defined as a stand-replacement disturbance. Results were disaggregated by reference percent tree cover stratum (e.g. >50% crown cover to ~0% crown cover) and by year…. Gain was defined as the inverse of loss, or a non-forest to forest change; longer-lived regrowing stands of tree cover that did not begin as non-forest within the study period were not mapped as forest gain. Gain was …reported as a twelve year total.”; “Outputs per pixel include annual percent tree cover, annual forest loss from 2000 to 2012, and forest gain from 2000 to 2012.”; “characterize year 2000 percent tree cover and subsequent tree cover loss and gain through 2012”.

Loss seems to a higher extent be accumulated over individual year while gain is more concidered over the whole period.

Other citations from the article: “We mapped global tree cover extent, loss, and gain for the period from 2000 to 2012 at a spatial resolution of 30 m” which I interpret as “pixels” 30*30 m.

From the supplement: “trees were defined as all vegetation taller than 5m in height.”


Does the Science article wants to evaluate change in forest cover?

The Science papers seem to try to evaluate the changes in forest cover by gain and loss. For gain the work can be considered as pioneering. But the intention seems not to get the actual result on forest cover of the combined action of loss and gain. If that had been the case the net change in forest cover 2000-2012 had been calculated in the article or supplement, but it is not. That daring step and the criticism it may raise is left to others.


How to get change in forest cover based on the material Science has compiled?

The change in forest cover 2000-2012 can be directly obtained by comparing forest cover 2000 and 2012. The area of gain and loss in that period can be analyzed. The change in forest cover could be analyzed in an article without leaving to the readers calculate the difference. No complications with both gain and loss in the same pixel or annual changes which may not be symmetric would occur. Even if the results of individual pixels is uncertain, the average change over groups of many pixels would be expected to be rather certain. The results would be similar to UNEP 1990-2006 .


EPI points finger by assigning many countries like Sweden a spectacular low score!

Sweden has a FORCH -4.1% (thus forest cover is suggested to be reduced 4.1% in 12 years). On a scale extending from 100 (best) to 0 (worst), Sweden is in the middle of assumed forest reduction (between 0 and ≈8% reduction), and would be expected to get a score around 50. If Paraguay was the bottom line (17% reduction), the fair score for Sweden would have been 75 (instead of 14). But Sweden get only 14.35 as issue score, one third of what seems logic. The misleading impression of the relative magnitude of the problem is further magnified by the presentation technique of the “EPI-rose” in the upper right corner of the country profile . The skew score of Sweden was what put me on the track of finding out what was behind, this letter would not have been written without that provocation. That Sweden actually is in (or above) the middle, no reader of the EPI reports will understand. Only 40 countries out of 137 get a score above the intuitively expected average, 50%, but the score scale is not what is intuitively expected.


Sweden gets a favorable rank in some measures according to Hansen et al 2013

In Fig S1 Sweden ranks in the best quartile for low loss share when loss is calculated as a percentage of loss + gain. Another way of expressing this (used for some examples by Hansen et al 2013) is the ratio “loss” to “gain”. I calculated that. Sweden is top-ranking (the best) among the 13 countries listed before Sweden in Table S3 (see below). Thus the bad impression of the EPI evaluation is dependent on how Hansen et al 2013 data are managed, and other management used by Hansen et al 2013, would result in a much better rank and score for Sweden.


Everything which contain IT and is not transparent can hide important unidentified errors.

So with the map of Hansen et al 2013. Airplanes crash, trains are late, power fails, information tables have errors, computers malfunction, web services do not work, systems are incompatible, it is impossible to understand what button to press, etc. etc. I get daily trouble from this large heap of imperfections. We should not trust everything which is written even when assurances that it is extremely trustworthy are made in a loud voice. Reasonable agreement with other observations, logic results expected from logic reasoning, and few seemingly contradictions should be possible additional criteria for acceptance. Against that background I suggest global forest cover change estimated with methods applied by EPI and Science is far from free from doubt.


The Landsat pixels used by Hansen et al 2013 are unreliable measures of forest cover and forests higher than 5m.

Pixel statistics given by Hansen et al 2013 are claimed to give estimates of forest cover of trees above 5m .The Science article is seemingly based on that tree height and tree cover of each tree is known. But that is not a direct observation, but an interpretation of the Landsat surveys. The Landsat measures reflected sunlight for 30*30 meter pixels, thus accumulated effect of many trees. The drastic changes at clear cutting are mostly relatively easy to identify and thus “loss” is much less problematic for Sweden than “gain”. To identify when a pixel got new forest is difficult and highly uncertain. The Landsat registration is not only depending on the situation concerning height and tree cover. The state of the atmosphere the actual day time matters. The vegetation and ground (like slope and blocks) matter besides the trees higher than 5 meters. Clouds must be correlated for. Sight through atmosphere may vary. The calibration of the sensors matter. The angle of sun matters, which depends on geographic location, date and exact time. In the far north – like Sweden – sun is far from zenith and the shadows are long, which may require special consideration for northern locations. Needles and leaves change characteristics during the season and the change is more drastic in northern countries like Sweden, where vegetation period is short. Variations in leaf area index may matter. It is a long and uncertain process between Landsat measurements of a pixel and evaluation of it.

The creation of a new forest can be seen as a gradual change, but as reflected by a Landsat pixel it will be much noise and uneven development. Not only tree height and canopy cover, but also other characteristics of the trees and ground change while trees grow during regeneration, which also may influence the Landsat observations. In the generation of clear-cuts some widely spaced mature seed trees remain to a different extent for variable time or become a part of the new stand. Pre-commercial thinning is generally practiced in Sweden and will also be reflected in a staggering and unstable behavior of Landsat observations of how regenerations develop. It seems likely that a developing pixel with young forest during some years can switch form no forest to forest to no forest to forest several times. Thus several repeated loss and gain during a period for the same pixel.

Commercial thinning may make a stand change from slightly above 50% canopy cover to slightly below. In absence of calibration to field data representing a wide range of conditions, it is impossible to make firm conclusions about the relations of Landsat observations on one hand and actual height and tree cover on the other. 30*30 land areas are not uniform. The degree of uniformity matters when then pixel information is accumulated to evaluate forest cover and uniformity may change over time.

An example of what may matter. In Sweden the development has been that clear-cuts (and thus regenerations) become smaller, get less “rectangular”, and contain more old trees or old group of trees. Thus 30*30 m pixels contain to a higher extent non-uniform mixes of mature trees and no trees or young trees. Sweden has natural regenerations, where many mature trees may remain for a decade and when may be removed. This certainly has effect on the pixels with more or less than 50% tree cover and may appear as changes of loss over time even at a constant “real” loss. With “gain” the same phenomenon may appear, but with several decades delay.


The net forest cover loss indicated by Hansen et al 2013 in many countries (examples in the top of Table S3 see below) is difficult for me to understand. In a steady state situation, losses are expected to be compensated by gains. On an individual loss pixel this gain may appear several decades after the loss, but in a steady situation the gain will come from replacement of previous losses and the forest cover on the country level would be constant. E.g. fire losses will be compensated by the gain replacing earlier fire losses which may have happened long before year 2000. Probably fire loss is decreasing in many countries because of development in fire control, and wherefore the gain replacing earlier fires is expected to be higher than late losses. Still e.g. Russia and Canada seem to show much larger loss than gain.


Hansen et al 2013 is the first effort to present globally consistent and locally relevant map data on forest cover change. I think it is very good if sufficiently safe and informative such maps could be made. The first effort cannot be expected to be perfect, but is likely to suffer from a number of unidentified problems. It is good for the development of remote sensing techniques and applications if deviations between Landsat observations and other observations are discussed with an open mind. The methods used by the Science article may be too rough at least for gain. To use the Landsat data for such specific purposes as ranking and scoring individual countries – as EPI does – seems to me yet immature. Certainly future developments will result is in more reliable evaluation based on satellite information for the good reasons listed by Hansen el al 2013.


There is much more experiences in interpretation of satellite data for forest loss than for forest gain. Hansen et al 2013 are pioneering in forest gain and net forest cover changes, while forest loss relies on more experience.

I interpret it as canopy cover including only canopy cover of trees with a height above 5 meter, shorter trees are neglected (smaller trees are just thought non-existing). I do not find that restriction on canopy elsewhere. I guess to go from light reflections measured by Landsat to conclusion about canopy cover of a part of the trees is complicated and leaves plenty of room for mistakes and uncertainties. But the exact definition of canopy cover is uncertain and not much discussed. What is actually tree cover and canopy cover? Is it the horizontal projection of the crown? Is the crown a polygon with room for the smallest branches most away from the trunk (probably not)? Or is it the size of the shadow the tree would make if illuminated from above? Or formulating it differently, the part of the ground and vegetation shorter than 5 m which would be seen if watched from above? How are dead branches considered? How does variations in leaf area index influence? What exact time are the annual measurements made? How does growth of the crown and changing characteristics of the needle/leaves during the vegetation period matter (like trees dropping their leaves)? Etc.  Etc. Questions which I do not know the answer to. It would have been better if the EPI work had been based on more classical definitions of crown cover.

In my opinion the conclusions based on Landsat for forest cover changes are currently too uncertain to be accepted as the truth in situations for cases where they seem to be contradicted by other observations, in particular for gain.


The map is available on the net and I made a very small study with it. The results are depicted in Fig 1 in the Science paper, the figure is available with a resolution better than a meter at  Fun, like with Google Earth! I can see details and identify the pixels closest to my home (Appendix 1 )! It is an extremely informative and generous contribution by Hansen et al 2013! Generally it seems to me that a fair map can be done for the world and links rather well to the real world. To better understand the relations between Landsat observations and their interpretation by Hansen et al 2013 and the connection to the real world geography, I went out and looked at the forest area closest to my home. Generally I could orientate myself among the pixels. It raises some speculations of doubt in the interpretation of data. It seems that “no forest 2000” may mean less than 75% crown coverage and not less than 50, but “gain” forest requires only >50%. It seems many pixels only a quarter of required size (15*15 m instead of 30*30). It seems a forest path besides the pixel matter. And it seems a possibility there are false “gain and loss” pixels.


In earlier statistics Sweden appears as one of the few countries, where forest cover increases

In UNEP 1990-2006 Sweden appears on the lead in increasing forest cover It seems surprising if Sweden in a few years would change from top to bottom, and that such a drastic change from increase to decrease should not be better noted and more discussed within Sweden earlier than 2014.


No country at high altitudes get a high score, indicating that latitude and not decreasing forest cover is the problem

The highest issue score of a country with forests at least as far to the north as Sweden is Russia (score 35, rank 57, slightly above the median). Norway, Lithuania, Denmark, Canada, Finland, Estonia and Latvia all come longer down on the list. No of these countries would probably subscribe to ongoing rapid deforestation and I am not aware of such a reason. The problem may be interference of results with latitude rather than real forest cover decrease.


Comment in Science says that Swedish forest area does not decrease, but rather increase

The Science article by Hansen et al 2013 is commented in Science: . The same comment is published in Swedish on the University web (Swedish University of Agricultural Sciences), presented as the universities view:  The authors claim that the forest area in Sweden does not decrease and rather seem to be increasing in the actual period, in contrast to the EPI and Science data. The authors of the comment (Fridman and Olsson) are those most relevant for the subject of the comment, the head of the national forest survey and the professor in remote sensing. This is the only comment which appeared on the Science article, although some other discussion has appeared in Science. No objection of the comment has appeared.

The Swedish forest survey is described in a recent article at  A better comparison between this two different estimation methods could be done using that article as a basis.


Can it be true that Sweden has lost “forest cover” 2000-2012?

The definition of “forest cover” EFI and Science use is different to what Swedes are accustomed to. Thus it has to be discussed if Fridman and Olsson comment is compatible with EFI. First I note that any of the causes mentioned (discussed) in the EPI documentation (land conversation from agriculture, logging, fire, decease or storm) are unlikely to be a major cause for the suggested reduction in the forest cover of Sweden.

This must be a speculative discussion. But I comment the following possible causes and if they are likely:

·         The area of clear cut has increased over time – No, the clear cut area has been constant or decreased (Fig 7.8) over time. The clear cut area was higher 1955 than now. It is a widely spread misconception that the clear felled area increases when the felled volume increases. The production per area has increased, more is harvested on the same area. In the past, expansion of the area exposed to final felling probably lead to less area, which full-filled the Landsat requirement for forest land but that is not the current situation. Forest gain from clear-cuts different times ago will replace recent losses. Some of the former clear cuts were in badly stocked degraded forests, which may not have been recognized as forests by Landsat criteria, but they should still generate gain pixels after regeneration.

·          Land converted to agriculture – No, more agriculture land has been converted to forest than the reverse way!

·          Exploitations (roads, cities etc.) – No, contributes too little to explain more than a minor part (but it contributes somewhat). And felling for exploitation is included in the constant clear felling, although it has a Long term effect as this logging will not cause a gain some decades later.

·         The Swedish forest gets younger, thus a larger part of the area has not reached the state above 50% canopy closure of trees higher than 5m. – Yes, that could be a reason for some of the observed effect. But that should not be classified as and compared with deforestation in other countries!

·         An increasing share of the Swedish forest land arise after clear-cut. – No, how could that explain an increasing loss?

·         Global warming – No, global warming results in “forest gain” as tree line raises.

·         Unusual incidents of losses in the period – No. The extraordinary storm “Gudrun” caused damage, which contributes to extraordinary loss in the period (mentioned in “Science”), but this cannot be a major quantitative explanation, although it can contribute a tenth added extraordinary loss.

·         Change in clear-cuts making them less uniform – Yes. Clear cuts get smaller with less straight borders. Individual trees and groups of trees irregularly spaced are saved more often. This means that a higher fraction of pixels are border pixels with less uniformity, and may mean that less pixels get more that 50% canopy cover even if the areas do not change. That may be interpreted by Landsat as a forest land reduction, although this is not a good interpretation.

·          Changes in forest regeneration practices over time – Yes. This seems to me the most plausible explanation, I discuss more below. But still I have difficulties to see that it could be the single explanation.

·         There are many other hypothetical explanations, but I do not know enough to expend. I have no information making any such factor a likely explanation and think I should be aware of it if such a factor was important.

No single factor seems able to quantitatively explain the loss in forest cover suggested by EPI, but it does not seem impossible that a combination of factors is a sufficient explanation. That could be further analyzed and Swedish scientists could consider what effects other used definitions of “forest” has on Sweden. Data of the type presented in Science will be more common and it is important to understand the relevance for Sweden when global maps occur.

It should be noted that the most likely explanations are unique for the period and this will lead to a higher score next decade. Quite possible Sweden will get a top ranking for the next decade if EPI does not change methods.

I should not have written this letter if it was only Sweden I got impression from, but as mentioned the forest cover reduction in some other countries by EPI methods seem also suspiciously high.

Why should EPI publish scores in a way indicating trends, if scores seem likely to change drastically from now on, provided they are correct?


Change in regeneration practices over time seem a plausible candidate for explaining some of the observed “loss”!

In an ideal steady permanent constant situation the forest area will not change over time. Loss is replaced by gain. For an individual pixel it takes often many decades till a loss has been compensated by a gain (till the new forest replacing the lost has got sufficient height and canopy coverage). But over an area gain will appear where losses were made before and – provided constant overall conditions the forest over the area will remain unchanged. This goes for fire, storm, disease and cutting, given overall conditions does not change. Even annual variations in catastrophes will be compensated and the forest will remain, but with some fluctuations. Even after clear-felling the forest may be replaced by planting or natural sources. Even if nothing is done after a clear-felling new forest usually appear, but with a longer delay than if actions for reforestation are done. The loss at felling will be compensated by later gain. For Swedish clear-cuts this is generally true. Even if a planting fails in Sweden natural forces will create a new forest. Even if the loss is caused by fire or disease or wind it will in Sweden almost certain be replaced by gain after a – sometimes long – delay. The land owner is forced by law to get a new decent forest. Sometimes the time delay between loss by logging and gain can take very long time, with the criteria applied by EPI for some of the pixels it may be four decades delay. Pre-commercial thinning will delay the time when gain occur in the loss pixels. It may never come a new forest, the most common reason is development (the clear-cut was made to give room for a road or house or something like that).  It happens that the land is converted to impediment by clear-felling, but that is probably rather rare and not a major reason for loss of forest area in current Sweden. The area of “legally defined” forest land seems to increase rather than decrease in Sweden in spite of some permanent losses due to “infrastructure development” (Fridman and Olsson 2014).

But the situation in Sweden is not permanent and steady. Things change over time. The regeneration practices change over time. The delay between loss and gain depends on these changes. If the delay increases over time an area of loss will not be compensated by a corresponding area of gain even if the loss per year is constant. This situation is not constant as the delay cannot raise forever and may sink in a later period. A possibility of such a mechanism is given, there is one factor making it likely that the delay in gain for the logging losses made around 1990-2004, which could be expected to be gain replacing the losses 2000-2012, was prolonged. That is that natural regeneration was considerable higher in that period. It takes much longer after final felling to get 5 meter trees after natural regeneration than after planting. And the regeneration will be considerable more uneven, so it takes more time to reach 50% tree cover

Figures 6.9 and 6.10

The regeneration practices may affect when and if gain occur many decades after the loss. Earlier Swedish forestry more or less regarded birch as a weed. Herbicide spraying was used in a big scale. Birch was vigorously removed at pre-commercial thinning. Today the praxis is to leave birch trees which cannot be predicted to severely outcompete conifers. This lead earlier to stands which reached the demand of canopy closure of trees more than five meters late. That can still be reflected as lower gain today. But the early practices of removing birch lead to stands which got a higher value production and better growth at a mature stage, it cannot be regarded as deforestation.

There may be other such variations connected to the regeneration practices. The use of the exotic lodgepole, which grow fast after planting, has decreased. Where may be reasons I am not aware of and it may not be known. If regeneration is done more in the south the new forest will come faster. The techniques of scarification changes. The control of animals harming regenerations (like reindeer and moose) changes.


Influence of the EFIs presentation of the forest issue

There are many ranking efforts for countries for widely different purposes, most have limited consequences. But for forestry in Sweden EFI magnified by others may have an effect I regard as undesirable. The by far most common reason for “loss” in Sweden is final felling leaving clear cuts. Swedish Foresters and Forest Scientists generally believe clear-cuts are mostly a good practice, at least for timber production. This has always been controversial, the clear-cuts are not nice and look like destruction. Green organizations, including political parties, pick up the information they can find supporting that clear cutting is a bad thing. This leads to problems and inefficient decisions even on the highest hierarchy of decisions. I consider forestry as very environmental, the forest is a renewable increasing resource. The business idea of forestry is to convert water, air (carbon dioxide actually) and sunshine to what we need. Clear-cuts means that forests with lower “vitality” are replaced by forests with higher “vitality”, thus generally can be argued to improve ecosystem vitality. The EPI presentation pushes Sweden to a forestry based on less vital trees, and in my opinion that means a worse future from the environmental point of view, without well-based reasons. Even the carbon dioxide situation will benefit from more vital trees.

However, probably the low EPI score of Sweden does not focus much attention so it is not worth as much time as I invest to argue against it. A more important reason is to get the Landsat use improved and the global maps more trustworthy to pinpoint on the problem EPI wants to address.

The EPI scoring for countries, which feel them biased, will increase the criticism about measures for deforestation globally and this may have the global effect that support for actions against deforestation will decrease even in the many situations where it is well motivated. It will also reduce the impact of the EPI scorings.

Should Sweden decide to improve its EPI rank in the coming decades, the most efficient methods seem for me to be:

·         less natural regeneration,

·         much increased use of the exotic lodgepole pine,

·         improved scarification,

·         reduction of the rein deer herds kept on unfenced land by the native population in the north,

·         intensified tree planting of already abandoned agricultural and pasture land,

·         – if technically and economically feasible, which has not been the case till now – more clonal forestry,

·         improvement of drainage at reforestation,

·         for a time perspective beyond the next decade – intensified tree breeding and seed orchard programs.

Ecologists would probably disagree on that most of these actions increase the ecosystem vitality of the Swedish forest. But I insist they do, at least if ecosystem vitality is defined as EPI does. But many ecologists would argue that EPI pushes Swedish forestry in the wrong direction. I have assumed that change in logging area is not considered.


What is Sweden?

Sweden is to my knowledge the country in the European Union with the largest forest area and the highest harvest. The standing timber volume has increased steadily for as long as known (Fig 3.11). Probably Sweden has now a higher standing volume with a faster growth than ever before. The percentage of forest cover is among the highest in EU. The annual forest growth has increased continuously and much. The area exposed to final felling has decreased considerable since 1970, but can be viewed as steady 1977-2012 (Fig 7.8). In Sweden it is in principle criminal for the land-owner with and insufficient forest cover on forest land. If the forest is exposed to final felling, the landowner is obliged to get a satisfactory new growth in a reasonable time. Sustainability is considered a fundamental pillar for forestry. Sweden often ranks rather well in international comparisons for different things and have hopefully a rather good reputation in caring for and investigating matters. In the overall EPI rank Sweden ranks 9 among 178.

Therefore Sweden’s low EPI rank for “forestry” – an issue which is more important for Sweden than most other countries and get much attention – probably is not an indication of something serious in the real world, but rather a problem for the EPI trustworthiness.


Difficulties with definitions in the Science article and EPI.

A key question is the definition and observation of tree canopy.

Landsat data are used for estimating (or rather “guessing”) change in the forest coverage of trees taller than 5 m.

FAO has 10% tree cover and ability to produce trees over 5 m as the bottom line for a forest. One may ask why a five time stricter limit should be applied by EFI than the authorized international body. Land with canopy cover 25-50% and tall trees “feels” much like a forest. It may (but probably do not) disfavor Sweden, with its slow growing and rather open forests. Even a savannah like tree cover is ecologically usually many times better than no forest in my opinion.

EPI calculates from the Science article that Sweden loses 4.1% forest cover over 13 years (4.1/12=0.34% per year), but Fridman and Olsson (2014) state that the forest cover loss based on “Science” is 0.5% per year. There is a difference of a factor 2/3 by different evaluators of the Science data. That indicates inherent difficulties to interpret the “Science” data. That different evaluators come to different results indicate that the Science study is not quite transparent.

Changes in forest status are more certain than forest status

There are many uncertainties how well the Landsat data observe forest cover and tree height, but changes in forest cover ought to be better predicted than actual forest cover. Many uncertainties will so to say level out, when changes over time are estimated for large areas. Errors which change systematically over time matters. But it is difficult to believe that e.g. the air characteristics should have changed much enough in the period to matter for the degree of uncertainty in Landsat readings concerning changes in the period 2000-2012.

Who I am and why I comment

I am a retired university professor of Forest Genetics. While browsing, my eyes fall on the EPI 2014 report with its poor performance score for “ecological Swedish forests” and suggested decline of Swedish forest cover. I tried to understand why the forests in Sweden appeared in this way and found that it is not well-supported. I found that some other scientists share this view.

Acknowledgements: This letter has been preceded by helpful discussions with others since early August 2014, among them Lars Laestadius, Håkan Olsson, Göran Ståhl and Seppo Ruotsalainen.


Figures and tables are in the better version

Leave a Reply

Your email address will not be published.