As gains in tree cover are not included, the Global Forest Watch results become skewed. Vahur Puik

When new data on global forests are published, it is always a grand opportunity to update analyses of global, regional and country trends. Global Forest Watch (GFW) has done just that. As promised, we now have the first annual update of global tree cover data based on 30×30-meter satellite data. The original dataset in this series, published by Hansen et al. and covering the period 2001–2012, has been extensively used over the past 18 months. It has been accompanied by a wide-ranging debate as to the validity of shown trends and what story the observed changes over time in the remotely sensed data actually tell.

In the release blog, GFW points to a continued increase of global tree cover loss, mainly due to higher losses in the boreal region, whereas all other domains show fewer losses. Indonesia stands out with a sharp decline in losses from 2012 to 2013, which led to some policy-related commentary in this WRI article.

GFW is clear, in that the data show tree cover and tree cover losses (aka “detreecoverization”). This is a welcome improvement in terminology compared to Hansen et al.’s use of “forest loss” and “forest change” for the same type of results, which led to some confusion. Many trees (or palms or bamboo) are found outside forests, and forest stands can have temporary conditions with few or very small trees.

Further, GFW is also open about not providing any results on “tree cover gain”, as these trends are too slow and gradual to be picked up by the data update. This means that we can only look at losses and cannot, with this dataset, balance those with gains through planting or natural regeneration of trees and forests.

THE DATA: A CLOSER LOOK

GFW should be commended for providing open access to the new data, including country-by-country statistics. Myself, I could not resist the temptation to look into these numbers. Note that below, I use data for the 30 percent tree cover threshold.

I first looked at the data for my own country, Sweden, and compared it with official forest statistics from the National Forest Inventory.

The official statistics report 183,000 ha of final fellings in 2013, which is very close to the 182,618 ha of tree cover loss reported by GFW for the same year. However, GFW has not picked up tree removals in thinnings over 394,000 ha in 2013. This appears to be a limitation in the remotely sensed data as most of these thinning operations remove 10–30 percent of tree cover and constitute a major part of “tree losses” in Sweden.

Further, as gains in tree cover are not included, the GFW results become skewed. According to Swedish National Forestry Inventory statistics, the growth of all trees in 2013 was 109 Mm³, whereas removals were 86 Mm³, for a net gain of 23 Mm³ in tree growing stock for the year. In other words, about 1 percent of Swedish tree cover was removed in 2013, and the other 99 percent grew to more than compensate for the removals. Still, Sweden makes number 13 worldwide in GFW country ranking for tree cover loss.

Now, the above example was for a country that has boreal/temperate conditions, established forest management and monitoring, and very limited pressure on forests from other land uses.

So, what do the new data tell us about tropical forests?

It is normally difficult to draw conclusions on trends from a one-year update in a long time series of data. However, as the GFW data have complete coverage and are standardized over time, there is some merit in looking at 2013 figures to compare them to earlier results.

I looked at countries in tropical regions (noting that this is not a clear distinction, as countries can have a mix of tropical, subtropical and even temperate forests). When dividing these countries by the three main regions, the following totals emerge:

* GFW’s figure for tropical tree cover loss in 2013 is 8.1 Mha here, likely due to a more precise, subnational delineation of the tropical domain. This difference should not significantly affect the conclusions here.

The table above confirms that total tropical tree cover losses in 2013 of 7.8 Mha are slightly lower than (92 percent of) the average for 2001–2012. It also confirms that Latin America continues to be the region with highest tropical tree cover losses.

THE TRENDS: LOSSES AND GAINS

Two strong trends are also obvious. The losses are going down in Latin America and sharply up in Africa. Using a three-year average gives a similar result.

The trend in tropical Africa is worrying and needs further analyses.

On another account, the table above also indicates that tropical deforestation was at most 7.8 Mha in 2013. We don’t know how much of this number constitutes non-forest tree losses or temporary forest tree losses – neither of which would be classified as deforestation. On the other hand, we may assume that few deforestation events are missed in this complete-cover dataset, which puts the maximum at 7.8 Mha for 2013.

Besides being lower than the directly comparable 2001–2012 GFW data, this level is also considerably lower than that reported in the Forest Resources Assessment 2010 (13 Mha/year 2000–2010, mainly in the tropics), FRA Global Remote Sensing Survey 2012 (11 Mha/year 2000–2005 in the tropics) and the earlier FRA 2000 (14.2 Mha/year 1990–2000 in the tropics). Determining whether this is due to data differences or a real trend will require more research.

Note that none of the above numbers includes gains in forest cover from planting, restoration or natural expansion. The net forest area change is reported by FRA and stood at –5.6 Mha/year in 2000–2005 (FRA 2010).

GFW is clearly taking global forest monitoring to a new level. The biggest gain with the GFW dataset is the consistency over time and space, as well as the complete cover. The biggest drawback remains the lack of depth in the remotely sensed data, which does not allow for reliable forest classifications or consistent estimates of gradual or fractional changes in the tree cover.

Keep up the good work!