"The most interesting application of this methodology is in evaluating the impacts of new policies in the immediate years following enactment," said Nancy Harris of the World Resources Institute, US. "Until recently, forest change was analyzed on a decadal timespan with results coming out only years after the change occurred. We have the opportunity with this data to intervene or correct course if the desired results from new policies are not seen."

Spurred by initiatives to slow or reverse forest loss, many governments are developing forest monitoring systems that make use of satellite imagery. An example is the Global Forest Watch platform initiated by the World Resources Institute in 2013, which updates an interactive map with detailed information on forest loss all over the world.

But as Harris and colleagues point out, data sources are growing quickly and that puts a strain on data analysis. As spatial resolution increases, pixel counts rise exponentially. "It becomes harder to distinguish the signal you’re looking for – deforestation, for example – from the background ‘noise’ that may exist in the data," said Harris. Another problem is subjective: defining the scales for analysis that give the most meaningful results.

The researchers have tackled the problem with two big-data tools, Spark and Hadoop, to configure the high-resolution data into a structure suitable for hot-spot analysis. Harris explains that the process essentially stacks two-dimensional maps for different years, producing a giant, three-dimensional "cube" of data. In this cube, every location can be analyzed with its neighbours to determine whether a region contains statistically significant levels of deforestation.

The methodology has confirmed that policies attempting to limit deforestation in Brazil have shifted the practice to the unprotected Cerrado ecoregion. In Indonesia the technique showed that a forest moratorium has not been effective in slowing loss and may have intensified loss, as loggers with concessions to clear land are doing so quickly before any new limits come in place. Meanwhile, in the DRC there was clear widespread intensification of forest loss along road networks amid rapid population growth, which, said Harris, "is likely to increase in the absence of clear policies to address this [deforestation] trend".

The researchers plan to apply their methodology again within a year, once new data for 2015 and 2016 are made available. "These more recent years of data will allow us to see where new hotspots are appearing or intensifying," said Harris. "We also expect to run this analysis on a wider range of countries and regions. Incorporating these more recent years will increasingly provide an opportunity for policymakers, companies and other actors to react in a timely manner to unplanned, illegal or unexpected forest loss."

The same type of analysis could be applied to fire data too. "In Indonesia illegal land fires are an annual problem and in El Niño years have grown catastrophically," Harris said. "Understanding where across the landscape fire-prevention programmes are most needed is one potential solution to reduce the impacts of fires in the future."

The study is published in Environmental Research Letters (ERL).

Related links

Related stories