An astonishing collection of satellite imagery to show 21st century forest cover change - a focus on South America

Research by Hansen, Potapov and Moore et al (2013) was built on newly available high-resolution satellite data. Interestingly the U.S government held this data previously but did not allow it to be used in research - despite the the commonly recognised threats to the worlds forests and associated ecosystem services... (click for source)
What came to attention is the intense forest loss in South America: but not necessarily in the highly researched Brazilian Amazon! Evidence from Paraguay and Bolivia, for example, showed that the highest forest loss has been occurring in the subtropical forests of the world.

This GIF image shows the drastic change 2000-2012 in Paraguay, but please also pay attention to the zoom on the whole South American continent. Highlighted in red are areas that were converted by anthropogenic processes - it will shock you and justify the relevance of this blog further.

(click for source)

The scale of land cover change impact on the climate system: a complex matter

The paper by Hoffmann and Jackson (2000) explores how the conversion of tropical savannah to grassland influences climate and finds a similar pattern of climate evolution to what I have outlined in my previous post (concerning the conversion of tropical forest to grassland). I want to take this, slightly different example, to explore the complexity revolving around scale and location of impact.
Despite a common pattern of change (increased temperature, lower seasonal rainfall, lower ET), the study shows strong regional variations in the climatic effect of an assumed universal conversion of land cover.

Within each of the world regions examined, some areas existed where precipitation is expected to decline by considerably more than 10% (see the dark red shading in Figure 1). While the authors point to the stochastic nature of the model used as a potential reason for this prediction, it very likely represents the real variability to be expected. Notably, with uneven land cover change over a tropical biome (as realistically not ALL of the land will be converted) it is even less predictable where the effects will hit most strongly, posing challenges to risk management.

Figure 1: click for source

Another complicating factor to the location of climatic impact is that through deforestation, roughness length of the vegetation is decreased, having significant effects on circulation patterns and thus making the rainfall pattern in an area more uncertain.

What this paper touches upon too, but Pires and Costa (2013) examine in more detail, is the concurrent land cover changes happening on the South American continent, with accumulating and interacting effects on the climate and biomes. They show that inner Amazon forest regions keep their rainforest biome characteristics, but outer forest regions may cross a forest-savanna bioclimatic threshold even at low deforestation levels. Different regions vary in their resilience to be pushed to the next “climate-vegetation equilibrium state” (Oyama and Nobre, 2003). Due to the climate-vegetation feedbacks induced by the warming, drying climate response to deforestation locally and as spread effects of adjacent land degradation in the cerrado savannah, the bioclimatic boundary between rainforest and cerrado is predicted to move 500-1000km north into the Amazon basin. Hirota et al (2011) have a more methodological perspective on how to model the climate effects of land cover change – but highlight just as effectively that deforestation in the tropics in particular may have local, meso and even global scale effects, which interact with the climatic effects of other region’s land cover change. Thus, observed and also modelled climatic changes must be evaluated upon the imminent local drivers, as well as on the relative impacts of more far-reaching changes.

Brazil disappoints with increased deforestation rate just before Paris Climate Change Conference

In the news: theguardian
While deforestation figures had been reducing steadily in Brazil since the implementation of new forest conservation policies, just before the COP21 in Paris, official deforestation figures of the past year August 2014-15 were published that meant a major "setback" for Brazil. The rate of deforestation had increased by 16%. Brazilian environment minister Izabella Teixeira acknowledged the let-down and blamed the increased pressure on land-cover conversion to feed growing livestock-company demand.

The effects of large-scale tropical deforestation on the climate in the Amazon

I want to return to South America’s deforestation problem in this post, and in particular explore the significance of this particular land cover change in terms of its biogeophysical forcing of climate. The inclusion of land-use scenarios within SRES highlights the influence that large scale land cover change can have on future climates.

So how is the South American biosphere linked with atmospheric circulation? 

Plants can be considered the primary sites for the exchange of water, energy, carbon dioxide and other chemicals between the land and atmosphere and thus vegetation cover is an important control of the climate system (Bonan, 2008). Forests are seen to sustain the hydrological cycle through evapotranspiration (ET), cooling climate through cloud-cover and precipitation. The low albedo of tropical forests that would normally cause warming is offset by the cooling effect of ET. The high momentum of the hydrological cycle in the tropics results to about 40-60% of precipitation in Amazonia being recycled (Oyama and Nobre, 2013). (Tropical) forests are integral to the carbon cycle too – the Amazon rainforest is a net carbon sink if kept in healthy condition: and thus is expected to have a mitigating effect on anthropogenic global warming.

Due to the (sad) continuing pattern of land cover conversion from forest to grassland/pasture/crop-farms, research has explored the potential effect of this upon the climate. It is especially important to integrate the “climate services” of forests into climate change predictions of the future – but the complexity of interaction and impact-scale provide hurdles. 

Here, I want to highlight the regional/local impact of the conversion of the Amazon tropical forest to pasture upon different climate variables.

A total precipitation decrease and an evapotranspiration decrease is consistent with almost all studies – highlighting the weakening of the hydrological cycle through deforestation (Hirota et al, 2011). The change induced to surface albedo through the land cover change is the important parameter affecting precipitation: following a proposition of Charney (1975), we can explain that increased surface albedo reduces local convection by reducing heatflux into the lower atmosphere, thus inhibiting the primary precipitation-generating mechanism in the Tropics (Hoffmann and Jackson, 2000).

One might argue that the reduced evapotranspiration (ET) stems from decreased moisture surplus (/reduced precipitation). However, a modelling study by Hoffmann and Jackson (2000), looking at the conversion of tropical savannah to pasture, highlights that reduced rooting depth, energy and conductance are responsible for reduced ET, as it is also reduced in regions with no absolute precipitation decrease. I found this very interesting, as it gives an indication of how biological and physical processes are interlinked on the ground, producing a common effect at the scale of the local climate.

Overall, precipitation seems to decrease to a larger amount than ET (McGuffie et al, 1995), indicating a reduced moisture convergence in the Amazon region under deforestation, which can translate to reduced runoff and renewable freshwater supply.

Pasture has higher maximum day temperatures than forest due to less energy used in evapotranspiration (Gash and Nobre, 1997), with the difference being more distinct in the dry season. Some studies (e.g. the much-cited Nobre et al, 1991) suggest that the increased temperatures together with the reduction in precipitation lengthen the dry season (which is almost negligible in natural tropical rainforest biomes). This works as positive feedback: the potential for forest burning is enhanced, as well as recovery of deforested areas prevented. An example of this is given in the paper by Pires and Costa (2013), who highlight North-Eastern Brazil and the Bolivian Amazon as approaching their “bioclimatic savannization” tipping point due to this positive feedback.

The 4 graphs below depict the results of only one modelling study, however show clearly the direction of the above explained expected changes to the Amazonian climate, would deforestation continue and replace all tropical forest with pasture/grassland. 

Source: Nobre et al (1991)

                                               

How surface water in the Amazon is affected by land cover change 2-fold

(source: click here)

The way surface waters in the Amazon are interlinked with its land cover is summarised well in Coe et al (2009)'s paper. Through land cover changes, humans affect the quantity of surface waters by "changing how incoming precipitation and radiation are partitioned among sensible and latent heat fluxes, runoff, and river discharge and altering regional and continental scale precipitation patterns". 

While it is generally observed elsewhere on the planet that deforestation induced increased runoff, micro-scale comparisons of forest and deforested (pasture) land-cover responses to precipitation confirm this. Moraes et al (2006) looked at two areas in eastern Amazonia, and found that runoff increased from 3.2% to 17% of annual through fall when the land-cover was altered from forest to pasture.

On a much larger scale (a whole river catchment), the discharge of the Tocantins River (south-eastern Amazonia) was analysed together with precipitation records over two time periods that different in both their proportion of catchment used for agriculture vs natural forest, and the rate of deforestation (time-period 1 and 2). Here, we can draw a similar conclusion to that found on a micro scale - the river discharge was greater with greater pasture land-cover. However what I found interesting to evaluate upon is that this increase in discharge was not at all caused by an alteration to the precipitation pattern (the precipitation of the two time-periods were not significantly different!). The discharge in time-period 2 was 24% greater than in time-period 1, due to land-cover change altering the aggregate hydrological response of the catchment. 

(What is good to note here, and what I will touch on in the next blogs, is that this higher proportion of rainfall ending up in rivers and streams over pasture land is also due to less water being lost to the atmosphere via evapotraspiration compared to natural forest cover).

I want to give a quick preview to my next blog topics, in order to evaluate the points above within a larger climate-change context. It is predicted through many modelling studies that large-scale deforestation in the Amazon will reduce precipitation and moisture convergence over the land and also increase temperatures (e.g. Costa and Foley, 2000).

Thus, we have two opposing forcings on surface waters in the Amazon, both induced by deforestation (land-cover change): 1) the alterations made to the hydrological response to precipitation results in higher runoff proportions and thus increases river discharge. 2) the greater climatic response to a large-scale shift to pasture land-cover will reduce moisture convergence, reduce annual precipitation over the area and thus decrease river discharge.

This evaluation is obviously just plausible when looking at the aggregation of surface waters in the region. Local and regional changes to river discharge will more likely be determined by one effect over another other. Also, the large-scale changes in precipitation patterns over the Amazon have not yet been uniformly observed over the basin and are often also just regional responses to intense land-cover change rates. 

Burning biodiversity: degradation by fuel wood harvesting

I want to dedicate this post to a discussion of a specific paper - which deals with an aspect of South American land cover change I am very interested in, and that has not been extensively studied. 
Specht, Pinto, Albuquerque, Tabarelli and Melo (2015)(here) provide us with an approximation of the potential impact of fuel wood harvesting in the biodiversity hotspot of the northern Brazilian Atlantic Forest. 

(source) Transportation of fire wood in Brazil, that is to be turned into charcoal and sold in local settlements.

Despite the high urbanization rate in Brazil, rural communities are growing with rising population patterns. Globally, a high percentage of the households use harvested fuel wood as their main domestic energy supply (Africa: 58%, Latin America 15%) and its usage is negatively correlated to income level. This study estimated local settlement fuel wood consumption ( in seven localities in the northern Brazilian Atlantic Forest) using a series of questionnaires to link consumption to other socio-economic variables - upon which a larger regional approximation was made. 
Generally, 76% of the households use fuel wood regularly and consume on average 686 kg/person/year of tree biomass. However, poorer people use 961 kg/person/year due to the lack of access to alternatives. The burning mechanisms is highly inefficient- explaining the high average per capita consumption.

Using this insight, we can evaluate the impact these patterns have on forest integrity. 
In the Atlantic forest region the rural settlers collect the fuel wood predominantly from remaining native forests - disrupting them by selective logging and pathways. Extrapolating these case study findings to the whole continent, where the natural forest is intercepted with small settlements, we can expect permanently human-modified landscapes to become the main habitat configuration across most tropical countries in coming decades (e.g. opinion here).
These chronic sources of forest integrity disturbance have been widely underestimated by management strategies. The cumulation of all small-scale harvesting practises across the remaining forest cover, however, have (other than many other land cover changing practises) no market forces as drivers. Fuel wood harvesting is primarily a subsistence practise and driven by poverty and access to other alternative energy sources. This issue is a perfect example of how international biodiversity conservation cannot be fully successful without accounting for the full complexity of drivers. In an "overpopulated" world, these indirect socio-economic  impacts will only intensify and should not pushed away as an unrelated problem by forest management. 

Some methodological issues with impact-studies of land-cover change on biodiversity

5 out of the 25 global terrestrial biodiversity hotspots, identified by Myers et al (2000), are located on the South American continent.

Humid tropical rainforests include some 60% of all known plant and animal biodiversity – and the Amazon is the largest of such biomes on Earth.

The land use changes in South America (not only in the Amazon!) have induced the conversion of a high percentage of its land cover from forest (rainforest and savannah biomes) to agricultural pasture or monoculture plantations. A reduction in biodiversity is obvious – but there are problems with evaluating this change.

As most impact-studies are on a too short time-scale, biodiversity change post deforestation is often estimated by comparing modified land cover with "forest baseline". A point of discussion in these studies is whether the 'primary forests' indeed capture the realistic biodiversity change, as even categorized 'untouched' forest areas are likely to have been impacted by human uses, edge effects or selective harvesting (reference here is made to the impact of the impact of the more subtle process of degradation - see previous post for more info please!).

While further investigating the methods used to evaluate land cover changes, I became aware of the oversimplification of some land transformations (in particular through reading Lambi et al. (2003)). A certain land cover category consists of specified biophysical variables and other attributes of the earth's surface (e.g. soil, biota, topography, water resources etc.). Modelling studies use data sets to represent the land cover by a set of spatial units associated with the attributes included in the specified land cover 'category'. This way of grouping attributes into categories leads easily to a discrete representation of land cover. Using this approach applied to the real transformations of the South American continent, it oversimplifies the subtleties and lags of land cover conversions (e.g. deforestation to other specific agricultural use) and completely neglects land cover modifications (smaller changes that affect parts of the specified attributes of the land cover category without changing its actual classification completely) (Lambin and Geist, 2006)

I hope this raised your attention to the possible difficulties in examining the biodiversity changes of large spatial areas that have not been assessed in detail on the ground.

My final point integrates all these insecurities in assessing biodiversity changes: I want to alert to the fact that the importance of preserving the 'natural' forest cover stems from its unknown realistic value and extent of inherent biodiversity and rare species.