## Friday, March 29, 2013

### Lesson 8 - REDD Case Study

This lesson shows us how to project deforestation rates based on socioeconomic variables. An econometric model specifies the statistical relationship between economic variables pertaining to a particular economic phenomena, in this case deforestation. It is useful to use an econometric model to simulate deforestation because it takes into consideration the many complex dynamic phenomena that produce deforestation as a result. These types of simulation models are important for evaluation of the potential for future deforestation under a business-as-usual scenario. In the case of this example, this is important in order for countries to receive incentives from reducing their carbon emissions from deforestation. REDD stands for “reducing emissions from deforestation and forest degradation” and is an idea that developing countries who do this over time, can receive financial compensation for doing so. This would require a country’s emission levels to be lower than historical levels and would require them to know historical emission levels and whether they were high or not. A model like this is important to predict future deforestation rates based on multiple economic variables to see if receiving compensation for REDD is a possibility for certain countries.

The model used for this lesson contains a spatial lag regression that is applied to compute the influence of five variables on deforestation: Percentage of crop areas, cattle herd density, percent of protected areas, proximity to paved roads, and migration rates. The five variables listed are the five static variables that are used to determine the probability of deforestation. The spatial lag regression makes it so that the deforestation does not happen right off in respect to the spatial variables and instead occurs over time and spreads over time. This models “spatial neighborhood matrix” allows neighboring cells, in this case municipalities, to have an influence on the cell in question. This means that each cell is not independent of the surrounding cells and the variables in the surrounding cells will effect it. In terms of deforestation, this means that if one cell becomes deforested, then it is likely that neighboring cells will become deforested as well over time. Because of this, we would expect to see deforestation occur in expanding clumps instead of in random patches.

After running the model with various cattle herd and crop rates, it was determined that the smaller the rates the better because this will produce a lower, steady rate of deforestation and CO2 emissions. A high rate of cattle herd and crops means a rapid increase of land use and therefore a rapid increase in deforestation and CO2 emissions and an overall unstable state.

The model is simulated for 20 years and a map output is produced for each year. Below are three output maps from the model starting with the first year, the 10th, and the last year:

At first, it looks like these images are all the same and there is no difference. But looking at the amount of yellow in the first image compared to the last one, you can see that the yellow areas are more filled in and have less green areas between them. This shows that the areas surrounded by deforestation (yellow) eventually become deforested as well and that deforestation is similar to the spreading of a disease. The most worrying part about this is that in the first year, there were very small areas of deforestation in the middle of green (forested) patches. Those were very small and may not have caused many problems in the first year, but because of the spreading of deforestation, these areas have become much larger by the last year. At this point they will cause problems such as habitat fragmentation for wildlife species. It also seems that the deforestation has spread faster between 2010 and 2020 as opposed to 2001 and 2010. This is worrisome because it means that the rate of deforestation is increasing as time goes on. I could not get any data to show up for cattle herds, but below is a graph for annual deforestation and a graph for crops for years 2001, 2010, and 2020:

As you can see, the crop rates are similar every year, but steadily increase every year. The seasonal changes are always the same, but the overall annual rate for crops increases. For deforestation, however, the seasonal changes are not always the same. 2001 and 2010 are similar in that their peak deforestation rates were at the same time, but in the year 2020, the peaks were at the times where deforestation was lowest in the past. Overall, deforestation rates for 2020 do not seem to be higher than they were in previous years. In 2001 and 2010, deforestation rates seem to be the opposite of crop rates but in 2020, deforestation rates become a bit more similar to the seasonal changes of the crop rate.

In the scenario parameters group, I changed the cattle herd expansion and crop expansion values from 0.05 to 0.5. This simulates a rapid increase in land use. Below are the output maps for years 2001, 2010, and 2020 again:

As you can see in these output maps, the spread of deforestation occurs at a much higher rate and by 2020, a larger amount of forested areas have become deforested because of the demand for crops and cattle herds. It is impossible to see these images as something that is not a problem. The amount of forested land left after just 20 years is alarming. Below are the graphs for the annual crops and deforestation rates:

As you can see, for the crop rates, they started out low in 2001, increased a large amount by 2010, and then dropped down to 0 by 2020. With deforestation, the highest rates were in 2001 and they continually dropped in 2010 and even more in 2020. Maybe the rates for both the crops and deforestation decrease by 2020 because there is drastically less land to use for crops and deforestation because most of it has already been converted.

Next is estimating the carbon losses from the two scenarios presented above. This will be done with the carbon bookkeeping model which is a very common approach to estimating carbon emissions. This model calculates annual carbon emissions by determining annual deforestation (from our previous model) and overlaying these deforested areas on a map of forest carbon biomass. This model assumes that carbon content is 50% of wood biomass and 85% of carbon stored in trees is released into the atmosphere during deforestation. This model was run for both scenario 1 and scenario 2. The graphs for the annual carbon emissions in tons for each scenario, CO2 emissions, and the graphs for annual deforestation in hectares for each scenario are below:

As you can see, these graphs all look the same. With the small (0.05) cattle herd and crop rates, the annual rates of deforestation and carbon emissions were fairly steady without much fluctuation. With the rates much higher (0.5), the rates of deforestation and carbon emissions instantly were higher and fluctuated much more. Overall, the second scenario emitted much more CO2 into the atmosphere and much more deforestation took place. Therefore, it is best if the cattle herd and crop rates remained relatively low so the CO2 emissions and deforestation rates would remain low and steady.