Three Reasons FAO’s New Forest Numbers Don’t Add Up – Kenneth MacDicken and Colleagues Disagree

In an earlier blog post, I wrote that the UN Food and Agricultural Organization (FAO)’s Global Forest Resources Assessment (FRA) provides a useful compilation of official national statistics on forests. But when it comes to aggregating national data and drawing conclusions about global trends, I offered three reasons why the numbers reported by the FAO should be viewed with a healthy dose of skepticism.

Today, Kenneth MacDicken and colleagues disagree. Their response is reprinted in full below.  MacDicken is a Senior Forestry Officer at the FAO and Team Leader of the Global Forest Resources Assessment.

We read with interest the 21 Sept 2015 blog by Jonah Busch.  While we appreciate the attention drawn to issues related to forest area change, we believe that the blog presents a mixture of useful analyses and serious errors.

1. Bad numbers in, bad numbers out.  No argument with the principle - and certainly there are some bad numbers in FRA 2015. However, just as for IPCC reporting, FRA 2015 includes a tier system that helps users understand data reliability.  Globally, only 11% of forest area is reported in the lowest quality tier - 89% is in the top 2 tiers.  Certainly some countries report less reliable numbers than others - which is the case for all global statistics.  The blog article suggests that countries play games with data - an unsubstantiated claim that is itself inflamatory. It also suggests that governments use their "own methods and definitions". This is true for national data collection, but it is not true for reporting to FRA as national data are generally converted by countries to comply with agreed  global definitions.

Busch focuses on data for deforestation in Indonesia as an example. However, while these data are available, the response rate from countries was low and data quality poor - as noted in FRA reporting.  The appropriate values for forest area change are gains and losses of total or natural forest area over time.  Indonesia actually reported an annual forest area loss of 685,000 ha for the period 2005-2010 - not the 500 ha per year "deforestation" number cited in the blog. 

By using the wrong data for forest area change Dr. Busch suggests FRA data cannot be relied upon.  We disagree. We believe good analysis should be a prerequisite for publishing critiques - even in the blogosphere. 

2. Rejigging the past.  It seems odd to suggest that learning should not be continuous.  There has been a substantial increase in the number of forest inventories, most combining remote sensing and field plots. Inventory and remote sensing technologies and applications have improved substantially in recent decades - should not these improvements be utilized? 

When there are changes to previously reported values, countries generally document the reasons for change - these can be found in the 18,000 pages of country reports that countries submitted for FRA 2015. The change in forest area is a very small proportion of global forest area - such that relatively minor adjustments due to new inventories, aerial imagery or remote sensing can make significant differences in change rates.  

3. Findings contradicted by satellite data.  Interesting assertion - but unfortunately quite superficial.  Over 70% of all forest inventories conducted or completed since 2010 include remote sensing - so to suggest that countries report without the use of remote sensing data is incorrect.  Comparison of the University of Maryland tree cover data with forest land use change is fraught with difficulties  The UMD data mixes oil palm, fruit trees, coffee, rubber, etc. with forest and reflects only cover change - not necessarily forest change, a fact that is noted on the Global Forest Watch site.  It also has a user accuracy of 80% for the tropics (Hansen et al., 2013). Among other definitional differences, the FRA excludes agricultural tree/shrub species and includes forest lands that are temporarily cleared and re-established with forest cover. Using satellite data alone is also ‘one-sided’ - tree cover loss is reported much more quickly than gain.

The assertion that remote sensing estimates of tree cover are superior to country-reported forest area is at best spurious (Coulston et al., 2014).  The UMD data suggest that Canada, Sweden and the United States are among the top 20 deforestation countries in the world - which is far from the truth using anything but short-term tree cover loss as a standard.  Forest area change analyses using the same satellite imagery over the same location and time can produce forest area change numbers that are very different from one another (Beuchle et al., 2015).  Using both field plots and satellite data provides a substantially more accurate assesment of forest area change than remote sensing alone (Reams et al, 2010).  Focusing on tree cover loss data may also mean a misdirection of effort away from where the problems are most acute.  

In the concluding paragraph of the blog the author suggests that FAO’s global statistics on forest cover change are not up to the task.  Here we also disagree.  Firstly, the FRA reports forest area - not tree cover.  A clear understanding of definitions is a first step in understanding forest area change. Secondly, the blog overlooks the fact that in most cases country-based reporting is more accurate and more responsible than other sources.

Can the FRA improve? Certainly.  Is it the best source of long-term global forest resource data?  Absolutely.  Do the numbers add up - not always.  Nevertheless, FRA provides the best current information on direction of forest area change and changes in  management and use. Remote sensing studies of tree cover change can be a useful supplement to the FRA data but the limitations of these studies also need to be understood. Rather than immediately leaping to criticism, those seeking to influence forest-related policies are advised to understand the value and implications of different data sets.  FRA data suggest that supporting economic development in the tropics can reduce rates of forest loss and increase forest area. Effective actions to help poorer tropical countries use existing agricultural land more efficiently and manage forests sustainably will slow the rate of forest loss in the tropics. 

Remi D'Annunzio, France
Örjan Jonsson, FAO, Italy
Rodney Keenan, University of Melbourne, Australia
Rodel Lasco, World Agroforestry Centre, Philippines
Kenneth MacDicken, FAO, Italy
Satoru Miura, University of Tokyo, Japan
Jeffrey Sayer, James Cook University, Australia
Sean Sloan, James Cook University, Australia
E. Somanathan, Indian Statistical Institute, India

Beuchle et al., 2015. Applied Geography 58, 116-127.
Coulston et al., 2014. Forestry 87, 267-276.
Hansen et al., 2013. DOI: 10.1126/science.1244693
Reams et al., 2010. PNAS 107, E145.

Sincere thanks to Dr. MacDicken and colleagues for reading my blog post and taking the time to respond publicly to the concerns I raised.  Not every large international agency would take the time to do this, so I commend their transparency and public engagement.  And it seems we largely agree on a common set of facts, although we have different interpretations of the implications of some those facts. In a follow-up blog post, I will explain why my three original arguments stand, and respond briefly to a few of the points raised by Dr. MacDicken and colleagues.

Update: Please read my response in the comments section below.