Extended Literature Review

General

In this section a more in-depth literature review on the specific subject of reforestation after forest fires is done to assess whether or not a robot is a desirable artifact to be created for such a purpose. Several aspects are investigated including but not limited to biodiversity, need for controlled seeding, the effectiveness of current methods and the costs of current methods. The general literature review concerning itself with the possibilities of robotics technology and the contemporary issues involving reforestation can be found in General Literature Review. General information about the project can be found at PRE2017 4 Groep6.

Biodiversity & Need for Control in National Parks

National Parks are located in most countries spread all over the world. In the United States only, the 59 national parks which were acknowledged at the end of 2017 span well over 400.000 km² (Sawe, B.E. 2017) . Even though National Parks are most commonly known as a touristic attraction, the reason they have originally been established is ‘’to conserve the scenery, natural and cultural resources, and other values of the park in a way that will leave them unimpaired for the enjoyment of future generations.’’ (The NPS Organic Act, 1916) . Meaning that, only taking the natural factors into account, the National Parks have to keep the wildlife as intact as possible. Thereby protecting it not only from human influences, but also from certain natural ones. The latter include, but are not limited to, typhoons, droughts, floods, earthquakes, and fires. Even though these phenomena are considered events which happen every once in a while and which are part of nature, their effect on a National Park is usually catastrophic. Usually, after one of these events, a substantial portion of the park is completely devastated, meaning that the wildlife needs to recover.

This paper will be limited to the phenomena of a forest fire originated from natural causes and the recovery of the National Park afterward. In order to recover a National Parks ‘natural scenery’, a clear definition of this must first be established. The natural scenery can be divided into three categories, being animals, plants, and non-living elements. This latter category spans the general topography of the Park, e.g. rivers, lakes, and mountains. Together these three categories form what is commonly known as an ecosystem, which is a term describing the relation between organisms and the physical environment they live in. Therefore, the goal of conserving the natural scenery can also be reformulated into conserving the current ecosystem. As the non-living elements are usually not as heavily influenced as the other two categories, they are from now on ignored in this analysis of the recovery of the National Park. Another concept linked to the conservation of nature is biodiversity. Biodiversity is the variability among living organisms from all sources, within and between species. The degree of biodiversity shapes the ecosystem, if there are many different species living in the park, they interact in another way with each other and their physical environment when compared with a situation wherein there are only a handful of species living in a similar environment. This shows that the biodiversity is of vital importance for the ecosystem and that a change in the park's biodiversity will result in a transformation of its ecosystem. (Greenfacts, 2018) The substantial dependence of natural scenery on the ecosystem and of the ecosystem on the biodiversity states that for a National Park to conserve its natural scenery, at least the biodiversity cannot change by any significant factor. It is, however, not claimed that this is enough for a park to conserve the natural scenery. It could be argued that other factors, like noise or horizon pollution, mean that the natural scenery is not conserved, but this is not discussed in this paper. Only the contribution of the biodiversity is taken into account.

Coming back to the recovery after a natural catastrophe, to say that a forest fire has a significant effect on the park's biodiversity is an understatement. Depending on the fire’s size, temperature, and the propagation speed it will wreak havoc over a large area of the national park which has disastrous consequences for all wildlife within the area. In order to get the park back to its original level of biodiversity, a certain degree of control is needed, as the original ratios of species should be established once again. Regaining this biodiversity is divided in both the animal and plant life, and can be done in multiple ways. This paper will be limited to the methods of regrowing the plant life, to be referred to as reforestation from now on. In order for reforestation to be effective, it needs to give all species a fair chance to return to their original population. This also requires to take into account the species-specific growth preferences, meaning factors such as temperature, nutrition, degrees of freedom, and exposure to sunlight. All these factors are heavily dependent on the plant's location, and thus on the location where the initial seed starts to sprout after the fire. Hence, the location of the seeds is of vital importance for reforestation. The three currently most used methods for reforestation are aerial, manual and natural reforestation, which will be further elaborated on below.

Current methods of reforestation

Natural Reforestation

One method of reforestation is natural reforestation. Natural reforestation relies on nature to return an area to forestland after the area is deforested. This returning of vegetation can happen through seeds that are carried by the wind, transported or buried by animals or that are dropped by mature trees (North Carolina Forestry Association, 2017) . This already leads to the first constraint of natural reforestation; there must be adequate living trees and animals around to enable natural reforestation, i.e. a nearby ecosystem has to be present. If there are no trees in the entire environment, there is only a negligible probability that seeds will be transported to the area. However, this study is concerned about returning a forest after a wildfire in a National park, in most cases, the fire is eliminated after a while due to human interference and this results in sufficient living trees left in the neighborhood to provide seeds.

In contrast to artificial deforestation, natural reforestation happens without the help of humans or machines. Because natural reforestation happens without the interference of humans, there is absolutely no control over natural reforestation. This means that the most dominant species, or the species that have not been destroyed by the fire, will regrow on the devastated area whilst other species which were originally located in this area are either all destroyed by the fire or take much longer to regrow, endangering them to vanishing from the National park since they cannot spread their seeds anymore. As is stated in the introduction from this wiki page, in order to rehabilitate the Natural park the biodiversity must stay in its original state to the best possible extent. This may happen with natural reforestation when the National park only consisted of one or a few species. However, when the National park consisted of many different species, which is often the case to create a nice scenery, natural reforestation provides no control and thus no guarantee that the old ratio of species will regrow. It can be said that the natural reforestation will always be good since nature can do what she wants, this is however not the case for a National park. Some species will always be dominant over other species, e.g. weeds in your own backyard. These dominant species are however not the (only) species you want to have in a National park. In order to recreate the original ratios of species in the National park, some level of control is needed which cannot be created with natural reforestation.

The benefit of natural reforestation is that the costs are extremely low. Because natural reforestation happens without the interference of humans, technology or other materials, there is also no money invested in the reforestation.

In the introduction, it is also stated that in order to achieve the preferred ratios of species, the different seeds need to be planted at different depths in the soil. It is obvious that this situation cannot be achieved by means of natural reforestation. Because most natural reforestation happens with seeds that are dropped by mature trees, these seeds will all end up on the ground and none will be buried, unless they are taken away by local wildlife. The level of control that is thus needed to recreate a National park cannot be obtained with natural reforestation.

Another source confirms this conclusion and states that only 7.9% of reforestation is done with natural reforestation (Lukaszewicz, 2002). This number is so low because where natural methods of natural regeneration fail or are unrealistic, artificial planting ensures the attainment of the main goal - sustainability of forest ecosystems.

Artificial reforestation has certain important benefits as to why it is often preferred over natural reforestation. It provides better control over tree spacing, more control over the species present in the new forest, the opportunity to plant genetically improved seeds or seedlings, and a higher rate of tree survival (North Carolina Forestry Association, 2017) . It can be summarized that when using artificial reforestation, the reforestation can be better managed than with natural reforestation, which is exactly what is necessary to recreate the National park as is stated in the introduction (United States Department of Agriculture, 2014) . Two main methods of artificial reforestation are manual reforestation and aerial reforestation. These two methods will now be discussed.

Manual reforestation

Manual reforestation is the oldest and most well-known method of replanting areas devastated by wildfires and the overall replanting of trees in general. However, manual reforestation is also a very slow method. It requires a large workforce and costs can quickly rise beyond proportions. It is estimated that manual reforestation can cost up to 62.000 USD for the initial replanting of a mere 1 by 1 km lot (Engel and Parrotta, 2001) . Maintenance is also required in the following years to make sure the forest grows as it is expected to be. This can cost up to 120.000 USD in the first 2 years only (Engel and Parrotta, 2001). This is a significant investment for national parks or other involved parties and other methods might be more cost efficient. More than 182.000 USD has to be spent on workforce and material. This is by far the most expensive and most labor intensive method and therefore not the most wanted solution. Manual reforestation is, however, preferable over methods for small areas.

This method of reforestation also poses significant health risks to the workforce (Finlay et al., 2012) . Since our focus lies on wildfires in national parks, this implies high pollution rates for soil, air, and water in the target areas. Ashes and toxic residues are often found in and near the devastated areas. There is an increase in mortality rate and symptomes related to the lungs occur at a significantly higher rate (Finlay et al., 2012) . This is perhaps the most important factor when deciding which method is to be used to replant an area after a wildfire.

Manual reforestation also has some benefits over the other primary replanting methods. This method is more precise compared to the other methods. Seeds can be planted at a more optimal depth and invasive or other unwanted species can be easily removed by the workforce on-site (Waldrop, 1998) . This means that the destroyed forest can be changed to a more healthy ecosystem by having the workforce suppress and promote certain species. The germination rate is therefore also relatively high compared to the other primary replanting methods (Waldrop, 1998) . However, this rate also depends on other factors like seed quality and differs per species. Seeds are also less prone to animals that use them as food as they are planted directly into the ground and are not easily accessible by animals. Seedlings and young trees can also be planted instead of seeds themselves to kickstart the growth of the forest. This method bypasses the inherent danger of the seed just lying on the ground. However, seedlings and saplings are more expensive and are fairly more difficult to move on-site when compared with the relatively small seeds. Recent advancements in seed quality have also made seed-only methods more beneficial, both in terms of costs and survival rate.

Manual reforestation is also already augmented by utilizing machines, by preparing the ground beforehand by means of subsoiling, machines can increase the growth rate and survival rate of the seeds (Waldrop, 1998) . Subsoiling also provides the option for placing the seeds even deeper into the ground, which can be beneficial for certain species. However, planting is still a labor-intensive practice and these large machines for subsoiling, often used in agriculture, cannot be used in all areas in national parks. Without machines, manual reforestation can still reach hill and mountain like areas that are normally not easily accessible for large machines.

A small robot could significantly decrease overall costs while still being able to a reach difficult areas. There are also no extra health risks for the human workforce since they can perform their function off-site. If high precision can be achieved without a significant loss of speed and a high germination rate, then small robots might provide an elegant, cheap and fast solution to the given problem.

Aerial reforestation

Aerial seeding is perhaps the most novel method of reforestation among the other options, which have generally existed for many centuries. Its main premise is a reduction in labor, as seeds can be sown at a much higher rate than manual seeding could ever produce and time-effectiveness, as an airplane can easily cover an area of several hectares at a much quicker rate than manual seeding using volunteers. However, the question remains if this method is truly beneficial in terms of actual saplings it produces and the costs it inherently carries, considering an aircraft of several metric tons needs to be lifted in the air burning kerosine and enormous amounts of seeds are spread.
Contrary to intuitive belief aerial seeding is in most cases not a standalone method, in order to be effective more often than not some preliminary groundwork is required to prepare the area to be seeded (in terms of boosting the receptiveness of the ground to the dropped seeds) (Régnière, 1982) . A very crude probabilistic model, taking into account two classes of possible areas (highly receptive due to site preparation or natural levels) and a constant occupation of highly receptive area per unit of area exists (Régnière, 1982) , which reveals that higher seeding rates do in general lead to more saplings, however the relationship is only linear in the case of pure natural occupation. If the site is prepared and occupation rates become higher, the relation between the number of saplings per unit area approaches a more or less square root relationship. This model further reveals that the variance of the pattern in saplings per unit area severely depends on the width between airplane runs, with lowest variance (corresponding to a uniformly seeded area, or in the case of multiple sown species a random pattern) only occurring at 1 meter distances. However, for each and every combination of width between airplane runs and seeding density a minimum invariance exists to create an optimal balance between the two. Using a purely random seed distribution (obtained by a uniform seeding density created by narrow spacing between airplane runs) as a measure for maximum obtainable sapling rate, it was found that the efficiency of a real process with a limited spacing decreases rapidly as the spacing becomes larger, although there is a compensating effect for higher seeding rates, albeit the amplitude of this compensation is much smaller than the amplitude of the decrease in efficiency at wider spacing. Very counterintuitive, for equal spacing between the airplane runs, the efficiency of the process first drops to a minimum of 88% after which it slowly increases for larger seeding drop densities, meaning that higher seeding rates do not necessarily mean better unless ridiculous amounts of seeds are used.

Furthermore, a second study in China comparing growth rates and carbon stock levels (equivalent for biomass) of both aerial seeded forests and naturally regenerated forests showed differences effectiveness between the two methods, with aerially seeded forests needing longer time to completely develop and hence always following behind naturally regenerated forests (Xiao et al., 2015) . This study was performed using the same tree species in previously highly naturally degraded areas. In the early years (10-20 years) of the new forests, the naturally regenerated forests seem to do better in terms of overall carbon stock, with an eventual conversion of the carbon stock values for the aerially seeded forest to carbon stock values of the naturally regrown forests for elder forests (50+ years). This faster growth rate of naturally regenerated forests is most likely caused by the self-capacity of natural forests to sustain themselves, which can only kick in at a later stage for an artificially planted forest by means of aerial seeding and by a higher occupation of carbon stock of the forest floor litter layer in the natural forests which provide many nutrients for trees. Thus overall, aerially seeded forests will eventually over time reach a state where the difference in their carbon stock is no longer statistically significant, however this approach in carbon stock is always from below and will never exactly converge.

Hence taking into account the above findings (Régnière, 1982) (Xiao et al., 2015) it is safe to conclude that aerial seeding can possibly be an effective method for reforestation, however it is largely context dependent on the problem. Considering one needs a pre-prepared area to increase the survivability of seeds and the airplane needs to make very tightly spaced runs to create a somewhat universally spread forest, the process becomes very time intensive in execution, whereas the main appeal of aerial seeding would be the simplicity in saving time as compared to manual seeding. In a situation where natural deposits of seeds are scarce, aerial seeding would be an optimal solution, since it simply would take nature too long to naturally recover the forest, even though the growth rate of natural forests are higher. However, considering we are dealing with forests fires in national parks this last situation is not very likely, as most often due to human intervention national parks do not completely vanish due to wildfires, rendering plenty of natural seed deposit left. In terms of control, aerial seeding gives little opportunities unless one chooses to specifically seed only one species of plant, as for any other desired degree of control a mixture of seeds has to be spread, for which different survival rates exist. Albeit, if one desires empty patches of land in the forest this could be done by severely increasing the spacing of airplane runs.
In terms of biodiversity, a similar conclusion can be drawn. Some level of biodiversity can be reached by mixing seeds from different trees and plants together for the airplane to drop, however this increases the difficulty of pre-preparing the site for different types of trees/plants to coexist together as each seed will have a different optimal depth and nutritional needs, whereas such a situation would also eventually be reached by nature. In terms of costs however, a study of the Society of American Foresters conducted in 1948 revealed that the costs of aerial seeding would be $7.25 per hectare (Westveld, 1949) , so taking into account inflation this would yield a cost of $75.39 (Bureau of Labor Statistics, 2018) . More recent sources report a cost of $60 per hectare for aerial seeding process only, with a whopping $630 per hectare if site preparations are taken into account as well (Canadian Silviculture, 2005) . The discrepancy between the values obtained by means of inflation and the more recent value is most likely caused by technological improvements making the process cheaper and more efficient, thus countering the inflation.
A popular novel alternative method for aerial seeding is by the employment of drones. Although this technique has a lot of potential, in its current state it is rather limited; communication technologies only allow a maximum distance of a couple of hundred meters for commands to be received by the drone (Fortes, 2017) , and even if the drones are build to be autonomous to overcome this distance limitation due to communication problems, they are still severely limited by their battery capacity which usually allows operation for somewhere between 30-90 minutes (Anderson, 2016) (Bustamante, 2015) .

All in all it can thus be concluded that in the case of a forest fire a robot can certainly be a proper solution as it gives a decent level of control if the resolution of the actuators is big enough and will certainly be an incentive for National Parks to switch to this novel technology if it can operate at a cost of less than $630 per hectare.

Conclusion

In the conclusion, a decision will be made on which current method of reforestation is most effective. This decision will be based on several factors which the research group considers important. The factors are:

• Effectivity with respect to time
• Costs
• Labour intensity
• Ability to restore biodiversity
• Effectivity with respect to resource
• Level of control
• Imposed health risk

The three methods will each get one of the four rankings per factor (--, -, +, ++). The ranking they get will be based on the literature review that is done per reforestation method. The reforestation method with the best overall score is considered to be the best reforestation method in the case of a forest fire in a National park. However, as the primary factors determining the effectiveness of reforestation are the degree of biodiversity and level of control, these criteria are assigned a weight of 2, whereas all other criteria have a regular weight of 1.

Natural reforestation and Aerial reforestation both receive one + with respect to biodiversity, this is because they have the ability to regrow several different species but the more dominant species will take over and the natural scenery will not recover. Manual reforestation receives two + because, with manual reforestation, complete recovery of the natural scenery is possible. Manual reforestation receives one + with respect to the time effectivity, this is because of a forest recovers faster when seeds are planted. Aerial and natural reforestation get a - and a -- respectively because aerial reforestation is only beneficial with respect to time in a non-fertile area and natural reforestation is very slow. Natural reforestation receives two + with respect to effectivity with resources because natural reforestation uses no additional resources and thus also has no waste. Manual reforestation receives one + because different seeds can be planted in the correct environment and therefore only the minimal number of seeds are used. Aerial reforestation receives two - because the seed-tree ratio is extremely high. Because of this high ratio and much fuel costs aerial reforestation receives two - with respect to costs. Manual reforestation also gets one - for costs because manual labor is always very expensive. Natural reforestation is free and therefore gets two +. Because manual labor is so intensive, manual reforestation receives two - when looking at the factor of labor intensity. With natural reforestation no labor is needed, thus this method receives two +. Aerial reforestation involves labor, this is however less intensive than with manual reforestation, the area that can be covered in a small amount of time is very big and this method therefore receives one +. Nobody manages the reforestation when it is done with natural reforestation, this method scores two - when looking at the level of control that can be fulfilled. Because with manual reforestation the location of each species can be determined you can exert a good level of control and this method thus receives two +. Aerial reforestation receives one + because you can determine the area at which you want to plant trees but you cannot control it per centimeter. Manual reforestation receives two - with respect to imposed health risk since it involves sending people in polluted and hence mildly toxic areas. Natural reforestation is ranked with two + on this aspect since the only involved health risk is minor injuries which might occur during the upkeep of the forest in the years to come, but this is something all methods of reforestation have. Aerial seeding receives one + for imposed health risk. This is because pilots will be exposed to low level of kerosine vapor when refueling their planes, which is a toxic substance. However, these vapor concentrations are low as they are quickly diffused into the surrounding area. Furthermore, flying at an alleviation exposes the pilots to higher levels of radiation, however seeding heights are rarely above the cloud level and are still in the lower level of the atmosphere where most of the radiation has already been filtered out, thus not posing a significant extra dose when compared to groundwork.

Table 3: Evaluation of the reforestation methods

	Natural reforestation	Manual reforestation	Aerial reforestation
Ability to restore biodiversity (x2)	+	++	+
Effectivity with respect to time	--	+	-
Effectivity with respect to resources	++	+	--
Costs	++	-	--
Labour intensity	++	--	+
Level of control (x2)	--	++	+
Imposed health risk	++	--	+
Result	+4	+5	+1

The result shows that natural reforestation and manual reforestation have a tight score of +4 and +5 points respectively. If the grating scale had a higher resolution (more than 4 options and/or less discrete weights) better distinction between the two could have been made. For now, they can thus be considered about equally good for reforestation after a forest fire with respect to the factors the group considers important, with the preference for manual reforestation arising from the evaluation matrix from table 3. Aerial reforestation, despite being commonly a used solution to a reforestation problem in the scope of a national park, is actually least favored. This is because in both important factors (biodiversity and level of control) it did not emerge victoriously.

This conclusion is interesting when a prototype of a robot which combats deforestation as a result of forest fires in National parks, is designed. This robot must follow the main design of manual reforestation since this method is considered most efficient. However, as can be seen in the table, there are also some improvement points for this method. Manual reforestation is very labor intensive and it is expensive as well. When a prototype for a reforestation robot is made, the robot must improve the way manual reforestation is done now with respect to labor intensiveness and costs, but the robot must be able to restore biodiversity and have a level of control at least to the same extent as manual reforestation methods have nowadays. A robot would, therefore, be a desirable way to improve manual reforestation since a robot is a good alternative to decrease the labor intensiveness of a job.