|Summary||On forest carbon cycle||
How to run
|General description||Examples for model output||
CASMOFOR produces lots of tables and graphs as output. All numbers must always be evaluated against the questions that were put when building the scenarios, and also against the values the user set before and during running of the model.
In order to demonstrate the functionalities of the model, how the program works,
and what types of outputs it generates, several examples are provided below by
the very questions that were defined in the
This question is answered by running one batch with a set area. In the example below, the species are the following: Hybrid poplar, Black locust and Pedunculate oak, their distribution by yield class is 20% in yield class 2, 40% in yield class 4 and 40% in yield class 3, respectively. Settings for all other scenario parameters must also be provided (e.g. the afforestation is made on cropland in 80% of the area etc.).
If three area scenarios are to be compared, then three batches of different areas must be run so that all other settings otherwise are the same. In case an afforestation project of a given area includes more than three tree species, then several batches must be run.
From the data that is produced by CASMOFOR after it has been run three exemplary area scenarios (I: 400 thousand ha; II: 200 thousand ha; III: 100 thousand ha) with the same predefined settings (species composition and distribution etc., see above), the following graph can be constructed:
The above graph demonstrates that the overall amount of carbon sequestered is related to the afforested area. Thus, the area needed to sequester a certain amount of carbon, assuming a given set of afforestation parameters, can simply be estimated by running the batch(es) of known parameters with a selected area, say 1000 ha, and then upscale or downscale the area using the ratio of the estimated amount of carbon to the amount to be sequestered.
Note that the amount sequestered on the afforested land changes over time, therefore, setting the amount of carbon to be sequestered is not enough, rather, it is also necessary to set the time frame.
It can be concluded from the data produced by CASMOFOR, or even using simple logic, that the overall amount of carbon sequestered is linearly related to the area afforested. In other words, for example, afforesting two thousand ha of land sequesters twice that much of carbon than one thousand ha does, provided that the distribution of the area by species and yield class is the same.
It is also worth analysing the results for one single area scenario, for which an example is shown on the graph below:
The graph demonstrates that
Note also that, whereas
the increase is slow in the biomass pool, the decrease in both the biomass pool,
as well as, although to a lesser extent, the overall system carbon is usually of
abrupt nature due to thinnings and especially final cutting. As a result of
these forestry operations, carbon is transferred from living compartments to non-living
compartments, from where
the carbon is gradually released back to the air. The increases in the amount of
carbon in the overall system are due to tree growth, as well as the increase of
the forest area (as long as the afforestation campaign lasts), and the
reductions are due to gradual carbon losses from dead roots, litter, deadwood
and soil, and abrupt losses due to harvests. Harvests usually take place in
single years, and the carbon in the firewood immediately leaves the system (that
is why total system carbon can abruptly decrease substantially). Even larger
abrupt reductions occur if carbon in the wood products is also considered
outside the system boundary (as in the accounting system under the Kyoto
Protocol and related provisions).
As afforestation campaigns are usually done in several years, the amount of sequestration depends on the temporal distribution of afforestation. In the examples below, afforestation was assumed to take place for 25 years, however, with an increasing rate in one scenario (from 10,000 to 15,000 ha a year), and with a decreasing rate (from 10,000 to 5,000 ha a year) in another. (Note that the input of the area that changes over time was entered using the user.xls file.) The differences, in this case, are easily observable, but mainly affect the absolute amount of carbon fixed. This is because the main factor in this case is the silvicultural regime, i.e. the distribution of thinnings and final cuttings over time: adding carbon to the system (e.g. by adding area) is always a slow process, while taking out carbon by harvesting trees much faster affects the amount of the carbon in the afforested area.
Different species grow at a different overall rate, and at different rate over time. For example, black locust and poplars are fast growing species, whereas oak and beech are slow growing ones. By selecting the species one can affect the growth rates for the afforested land. The silvicultural operations (i.e. the time and intensity of thinnings and final harvests) are also species-specific. It must, however, be noted that species selection is a rather difficult task, as the occurrence of species is site specific, so not all species can dwell on all sites. The example below shows but one such a case.
As mentioned above, tree growth rate is site-dependent. However, almost all processes in forests are site dependent as they usually depend on temperature, water supply, nutrient availability and their distribution over time and space. However, due to lack of data, site dependency in CASMOFOR is manly limited to site-dependent growth rates and silvicultural regimes. You find more details on site-dependent growth rates here, on silvicultural regimes here, and on site in general and here. The effect of site is simply modeled by specifying how much of the various species are afforested in the various yield classes.
The graph below demonstrates that areas of the same unit size with the same species sequesters different amounts of carbon if they are of different yield classes. The upper curve shows data for yield class 2, the middle curve for yield class 4, whereas the lower curve for yield class 6. The curves are different not only because of different growth rates due to different yield classes, but also because the timing and the intensity of silvicultural operations and final cuttings also depend on the site class.
Before planting trees, one usually has to do soil preparation. This is an operation during which soil is disturbed due to which soil may loose some carbon. There are very rare case studies that have assessed the amount of this lost carbon. In order to assess the effect of this loss, one may set the amount of carbon lost at the beginning of running the model. For more details, click here. The graph below demonstrates importance of the loss of carbon relative to the overall carbon content of the entire system. First, carbon is lost from the soil at the time of forestations (red coloured area) at a relatively rather large at the beginning. The amount of carbon lost reduces over time, then disappears (soil carbon stock changes become positive, see the brown coloured area at the top), but the amount of soil carbon loss reappears, at a small scale, at year 2049 when the stand is clearcut and regenerated, and soil is disturbed again. Note that, when simulating this scenario, a rather high rate of emission due to disturbance was selected that is unlikely to occur in real-word afforestations. Note also that a positive amount of soil carbon changes (i.e., soil carbon gain) appears as a brown patch at the top, which means that, between about 2030 and 2049, and then after about 2053, the net result of losses from, and gains in, the soils is a net gain.
The question is answered by running two batches, because the cropland/grassland ratio is assumed to be common for the maximum three species scenario of a batch. For the first batch, the share of cropland is set to 100%, whereas it is set to 70% for the second batch, which otherwise has the same settings. In the first case, net carbon stock cases are always positive. However, in the second case, CASMOFOR produces the graph below for an even-aged forest area:
The graph demonstrates that afforesting on grassland has a significant effect
with respect to both the soil carbon pool, as well as the total amount of carbon
that can be sequestered in the afforested area, at least for several years. If
the share of grassland is 50% or more, soil continuously loses carbon.
This question is answered in an example of simulating Black locust stands that have a rotation period of about 40 years. First, an even-aged stand is simulated, i.e. a stand where the afforestation was done for one year. For any distribution over site, i.e. yield classes, and by setting a long-enough projection period, the dynamics of harvests and reforestations become cyclical as the blow graph demonstrates, as it is assumed that all forests are regenerated after either the final cutting, or any theoretical natural disturbance. Thus, the total amount of carbon fixed in the system will take the same values cycle after cycle. Note that the above-ground biomass changes between the maximum value and zero, whereas many other pools do not drop to zero, depending on their specific life-cycle.
However, when simulating an age structure where an equal area of forest is found for each year, most pools are saturated (after about one rotation time, i.e. 40 years, after conducting afforestations for each year for a period of a rotation time, i.e. for 40 years) but do not lose carbon over time as the graph below suggests. However, this only holds true as long as there are no major disturbances in one or more age classes.
Note that it is not possible to simulate non-regular natural disturbances in CASMOFOR, however, the effect of final harvests, which are similar in nature, is included. Therefore, it can be concluded that the more age structure is diverse, and the larger forest area is, the less is the effect of natural disturbances and final harvests in terms of permanence; the only difference can be the timing of loss of carbon if the natural disturbance occurs earlier than the final harvest. However, also in this case, it is more about transferring carbon from one compartment to another than about losing carbon from the system. Consider also the example below.
Mortality is a natural phenomenon that occurs in each forest due to smaller or larger natural disturbances like wind, fire or snow or ice breaks. CASMOFOR includes a module that simulates such a mortality of trees by drawing a random event from a theoretical disturbance distribution. This also means that, for each separate run, the model produces different mortality events.
For the same afforestation program as above, CASMOFOR, produces the below graph when also simulating mortality:
Note that there are two major effects of mortality. One is that the carbon stock of both the above-ground biomass, as well as the belowground biomass compartment changes as mortality events occur. (For conifers, the amount of leaves would also change.) However, the dead root and dead wood compartments also change, but the dirction of change is quite the opposite. Because of this, the system does not lose carbon. On the other hand, the total amount of carbon in the system is by some 10% less than in the undisturbed system. This is mainly because, although the growth rate is not affected, the carbon of the disturbed biomass eventually remains in other pools for a shorter time than in the biomass itself, so it is emitted into the air rather than kept in the forest.
It must be underlined here that, due to the poor understanding of the complicated processes and to lack of proper data, the models in CASMOFOR were not designed to accurately simulate the effects of disturbance on the carbon content of the forest, therefore, the simulated results only partly demonstrate real-world events.
When taking decisions on what type of projects to implement, e.g. whether industrial or forestry projects, an important information is how much it costs to sequester a unit amount of CO2. This can be calculated by dividing the amount sequestered by the net costs (i.e., costs minus revenues) of establishing and maintaining a forest area. For one ha of a Black locust forests, these costs and revenues, which depend on the site, quality and age of the forest, can be seen on the graph below for two rotation periods:
If many stands are established, and all costs and revenues, as well as all carbon sequestered are accumulated, the net costs of sequestering a unit of carbon (HUF/t CO2 fixed) shows a tendency as demonstrated by the figure below:
The curves, either for the total amount of carbon fixed, or for the carbon sequestered by the (belowground + aboveground) biomass, show that net costs sharply decrease after the afforestation program has started, but are in the negative territory, which means that fixing carbon by forests costs some money. However, net costs approximates zero after as short as two decades. What is more, after some 50 years in this scenario, the costs become "negative", which means that afforestations sequester carbon while they provide us with net revenue. This demonstrates that afforestations are among the most cost effective, although not necessarily the quickest means of climate change mitigation.
Note that it is possible to adjust, in the "summary" sheet of the results file, the market price of a tonne of CO2 sequestered, as well as the HUF/EURO rate.
Note also that economic data are highly market dependent, that is, they may quickly change from time to time and from place to place.
The sensitivity of the model to parameters cannot be calculated by simple mathematical formulas due to the complexity of the system. Therefore, the sensitivity is studied by running a so called Monte Carlo analysis. This involves running the model many times and changing the value of selected factor(s) each time the same batch is simulated by a random value. When the same system and the same scenario is run many times with different parameters, an error distribution is produced which can be used to characterize the sensitivity.
There is a separate module where the sensitivity can be analysed. This module requests the user to specify (1) how many times the simulation should be run, and (2) what the standard deviation of the error of the variable(s) to be studies is, and (3) whether the error assumed is random or systematic. The first graph below shows (1) by a red arrow, (2) by a red circle, and (3) by a red rectangle.
After the Monte Carlo settings have been provided to CASMOFOR, the scenario to be studied must be also set. Then, CASMOFOR runs the number of times that has been specified above (in the example above, 100 times). NOTE that this may take many minutes, or even hours.
The results are found in the "results_XXX".xls file, where XXX is the name of the run you provide when running the program. The file contains statistics, and graphs based on these statistics. The below two graphs show selected results.
The figure above demonstrates, for a species in the scenario example, the total and mean, minimum, maximum, as well as standard deviation and number of runs for each calendar year of the simulation (note that only some of the first years of the simulation are shown here). These data characterize the sensitivity of the accuracy of the estimation of the system for the amount of carbon sequestered by this species.
In contrast, the table below, which again only shows a part of the entire table, is an example for the same statistics for all compartments of the forestry system separately.
The minimum, mean and maximum values, but also standard deviation and 95% confidence interval can be printed also graphically by species and pool, which are demonstrated in the below graphs (the first group of three graphs shows absolute values, whereas the second group shows relative values). By modifying the various model parameters, the graphs transparently show the sensitivity of the system to the simulated values over the calendar years. Species/pools, and the statistics to be shown on the graphs can be selected from the two pull-down menus on the top.
Graphs with absolute values:
Graphs with relative values:
Finally, to check what the actual standard deviation of the various parameters is, you can analyse two tables that can be found in the output file. One is a summary of the standard deviations by parameter and species (sheet "statistics", beginning from line 177, only a section is shown below):
The other is a detailed table of the generated parameter values for each run (sheet "statistics", beginning from line , only a section is shown below; note that Column 1 has different values for each run, other factors always have the same value):
This webpage was last modified by Zoltan Somogyi 29 June 2014.