3.1 The BCS Freeware Operation
The key steps of the BCS freeware operation for the simulation of a biogeochemical cycle is herein described. The first interaction between the user and the BCS freeware is a blank window (Figure 1). The starting window is divided into two major areas: Tables Area (TA) and Information Area (IA).
The following step consists in creating a biogeochemical cycle and including the respective boxes information, which are shown in the IA. BCS freeware allows for the inclusion of an unlimited number of boxes, each one of them with its own name and reservoir. Then, a spreadsheet appears in the TA, prompting for matter transfers between boxes throughout the biogeochemical cycle (K matrix). That matrix, described in Section S1 (Supplementary Information), is calculated by the ratio between the reservoir and a certain flux. Nevertheless, the BCS freeware also allows the user to include fluxes directly. Figure 2 shows an example of a window containing the K matrix (TA) and a list of the boxes (IA) that compose the biogeochemical cycle.
The final step is setting the time span of the simulation. The BCS freeware makes it possible to use discrete time values (days, months, years) or an implemented automatic time span selection. For the second alternative, a halting criterium must be provided (reservoir difference between iterations).
Finally, a results window is shown (Figure 3). That window shows all of the input information, as well as the data matrices used by the algorithm. It also displays interactive charts which present the reservoirs temporal profiles.
Single or multiple disturbances in a biogeochemical cycle can be easily implemented in the BCS freeware. Once a simulation ends, it is possible to generate a new K matrix and boxes list, containing the respective simulated end values. That second K matrix is fully editable, making it possible to simulate disturbances by changing the content of its cells. The previous process can be repeated indefinitely, making it possible to simulate multiple in-series disturbances. Figure 4 shows a flowchart of the BCS freeware operation.
3.2 BCS Freeware Accuracy
In order to check the accordance between the results obtained using the BCS freeware and the other two existing software, the sulfur biogeochemical cycle (Figure 5a) was taken as an example. The corresponding K matrix is given (CHAMEIDES; PERDUE, 1997) (Figure 5b). That cycle is at steady-state conditions, what can be in Figure 6, which shows the graphs for two compartments (Soils and Atmosphere), generated by the BCS freeware. Those two compartments are the ones that really change after perturbations, due to their respective small reservoirs (compared to the other three compartments).
One disturbance was tested and the respective results compared (the cycle behavior until another steady state is reached): 1st) an increase in the mass transfer rate between S sediments and Soil reservoirs, caused by mining and other land usages; and 2nd) an increase in the mass transfer rate between S reduced sediments and Atmosphere reservoirs, caused by the burning of fossil fuels. That perturbation was implemented by changing k12 from 2.5 × 10-9 to 2.0 × 10-9, k23 from 5.0 × 10-9 to 4.0 × 10-9 and k24 from 4.47 × 10-9 to 1.3 × 10-8, respectively (CHAMEIDES; PERDUE, 1997).
This scenario lasted for 130 years. Afterwards, the mass-transfer coefficients were returned to their original values (2.5 × 10-9, 5.0 × 10-9 and 4.47 × 10-9) for further 1000 years. Figure 7 summarizes the results showing the final responses for the Atmosphere reservoir, which is the more affected one by the proposed anthropogenic perturbation.
The pairwise cumulative standard errors (%) between the software outputs were calculated (Table 1). At steady-state conditions, the outputs were virtually the same. However, small differences were noted during the perturbations performed. It is noteworthy that, T-D and T-M themselves produce slightly different outputs. BCS outputs are closer to the T-D ones. All of the errors between BCS and T-D and T-M, but one (perturbation 2, compartment 3), are smaller than the ones between T-D and T-M. Nevertheless, those are all smaller than or equal to 2.7%, which are fairly acceptable. The raw outputs are shown in Section S2.
Table 1. Pairwise cumulative standard errors (%) between the software outputs.
Disturbance
#
|
Compartments
|
Comparisons
|
T-D × T-M
|
T-D × BCS
|
T-M × BCS
|
1
|
1
|
0.00033
|
0
|
0.00033
|
2
|
0.0078
|
0
|
0.0078
|
3
|
0.82
|
0
|
0.82
|
4
|
1.7
|
0.86
|
1.5
|
5
|
0.016
|
0
|
0.016
|
2
|
1
|
0.00066
|
0
|
0.00066
|
2
|
0.016
|
0
|
0.016
|
3
|
0.80
|
1.4
|
1.4
|
4
|
2.7
|
0.86
|
2.7
|
5
|
0.32
|
0
|
0.32
|
Note: At steady-state conditions, the outputs were identical.
3.3 BCS Freeware Ease of Use
Although statistically identical, the three software differ mainly on their ease of use. The BCS freeware was designed to be a standalone software, with an easy graphical user interface (GUI), which is familiar to undergraduate students. It is also capable of directly simulating high numbers of reservoirs, even with multiple-stage perturbations. T-D is limited by the technology available at that time and does not have several functionalities which are presently easily performed by the BCS freeware, not mentioning an outdated and unfamiliar GUI. Although T-M is able to perform all of the functionalities present in the BCS freeware, it requires that the user has previous knowledge of matrix calculation software, which may not be the case depending on the course taken and the user’s seniority. Table 2 presents the key features of the three software.
Table 2. Key features of the tree software.
FEATURE
|
BCS
freeware
|
T-D
|
T-M
|
Requirements
|
JRE 1.8 or higher
|
DOS emulation software
|
Matrix calculus software
|
User Interface
|
Java Swing, common GUI structures
|
Text-based structures
|
Mainly command lines
|
K Input
|
Editable spreadsheet
|
Text fields
|
Not possible
|
# of Reservoirs
|
Unlimited
|
Ten
|
Unlimited
|
Flux Input
|
Direct, editable spreadsheet
|
Not possible
|
Direct,
one-by-one input
|
Multi-Stage
|
Direct, stages retain previous data
|
One at time, stages do not retain previous data
|
One at time, stages do not retain previous data
|
Graph
|
Customizable, interactive
|
Fixed, few points
|
Fixed, only zoom functionality
|
The BCS freeware has a simpler configuration, even requiring the JRE. In fact, the JRE might be already present on the user’s computer, as it is also necessary to a large number of other software and functionalities. If not, the JRE is easily installed.
T-D has an outdated, less interactive user interface, as it runs on a non-windows environment, impairing parallel tasks. T-M relies mainly on command lines in the selected matrix calculation software. Those command lines usually have specific, although similar, syntaxes for each software. The BCS freeware, on the other hand, was created with Java Swing components, which include usual GUI structures, requiring less time for new users to master its use.
Two relevant advantages of the BCS freeware are: 1) the possibility of directly inputting mass fluxes instead of mass-transfer coefficients; and 2) multi-stage simulation. T-D requires fluxes to be transformed into mass-transfer coefficients, adding a somewhat tedious step to the simulation. T-M only accepts mass fluxes, but they must be inputted one by one, each one of them in a different prompt line. If the user mistypes any of the inputs, he/she is forced to restart the whole feeding routine. When performing a multi-stage simulation (multiple perturbations), T-D and T-M work with batch calculations, which the user setting up each perturbation manually. With the BCS freeware, the user can do that by simply pushing a button. It is worth mentioning that the BCS freeware retain all of the data of multi-stage simulations, making it possible to observe all the perturbations performed in a single graph.
Finally, the BCS freeware provides fully interactive and customizable graphs as part of the simulation results. T-D produces rudimentary graphs, with a low number of points. Alternatively, T-D provides the polynomial expression for each reservoir, in order the user to plot the data somewhere else. Nevertheless, that polynomial must be manually copied to the graphing software, what is tiresome and prone to add typing mistakes, especially when a large number of reservoirs are simulated. T-M provides graphs, whose only functionality is the zoom in and zoom out tool, besides a table with the x,y coordinates of the graph.
3.3 BCS Freeware as Teaching Tool
In the end of course, the students anonymously answered an assessment questionnaire. Statements were given and five answers were possible (totally agree, partially agree, neither agree nor disagree, partially disagree, totally disagree), besides open questions. When questioned about the use of case studies as evaluation tools instead of written exams, the students consistently preferred the former.
Other questions were: “Regarding the simulation tasks, what is your opinion about their use in the course? Were they helpful in the learning process? Do you believe that this tool could have been better explored? Were the deadlines appropriate?” The answers mainly revealed that the students accepted well the simulator, the case studies and that the proposed deadlines were enough. Here are some excerpts of the answers:
“I liked the use of simulations in the course, this [BCS freeware] software gives a very good sense of what happens with the cycles and, consequently, with the entire [planet] life when biogeochemical cycles are disturbed, what is very interesting for learning. I believe that this tool was well explored in the course with proper deadlines.”
“The simulation tasks were very useful to the learning process, as they make it possible to analyze and understand the biogeochemical cycles in a more practical way. The simulator was very well explored and the written reports were a proper form of evaluation. The deadlines were enough for running the simulations, but the assistance of the tutor was necessary.”
“I believe that the use of the simulation tool was important to the learning process in the course. I believe that it was well explored, and it would be even better, if the simulated problems were even closer to real-life problems. The deadlines were OK.”
Furthermore, during the elaboration of the reports, the students were also able to evaluate whether the computational output data had real physical meaning and related well with what they have learnt in classes, what is an important discussion as computational modeling becomes increasingly used in scientists daily lives (CHAMIZO, 2013).
It was interesting the fact that the students began to relate the simulation results with possible environmental problems, once the biogeochemical cycles were disturbed. It became clear their evolution throughout the course, as they started to realize how small disturbances would call for very high periods of time for reaching the new stationary state; moreover, they started to wonder whether that new state would be suitable for human beings to thrive. As they handed out the reports, they also attempted to identify short-, medium-, and long-term consequences of the proposed disturbances.