WO2009015069A1 - Methods and systems of evaluating forest management and harvesting schemes - Google Patents
Methods and systems of evaluating forest management and harvesting schemes Download PDFInfo
- Publication number
- WO2009015069A1 WO2009015069A1 PCT/US2008/070601 US2008070601W WO2009015069A1 WO 2009015069 A1 WO2009015069 A1 WO 2009015069A1 US 2008070601 W US2008070601 W US 2008070601W WO 2009015069 A1 WO2009015069 A1 WO 2009015069A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- forest
- harvesting
- management
- sample
- schemes
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Methods and systems for evaluating forest management and harvesting schemes are disclosed. In some embodiments, the methods and systems include the following: using a data-based, spatially-explicit forest simulator, evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a plurality of data fields; evaluating each of the plurality of data fields to determine whether it meets predetermined criteria; assigning a first value to those of the plurality of data fields that meets predetermined criteria and a second value to those of the plurality of data fields that does not meet predetermined criteria; training and cross-validating an untrained support vector machine using the first set of management and harvesting schemes and the corresponding first and second values thereby developing a trained support vector machine; and using the trained support vector machine, screening a second set of forest management and harvesting schemes.
Description
METHODS AND SYSTEMS OF EVALUATING FOREST MANAGEMENT AND
HARVESTING SCHEMES
CROSS REFERENCE TO RELATED APPLICATION(S) [0001] This application claims the benefit of U.S. Provisional Application No.
60/951,075, filed July 20, 2007, which is incorporated by reference as if disclosed herein in its entirety.
BACKGROUND
[0002] Sustainable efforts in the management of forest biodiversity and ecosystem services attempt to align conservation concerns with economic incentives. In North America, forest managers must often strive for sustainable forestry rather than the narrower objective of sustained timber yield. However, they are often asked to achieve what are still perceived as conflicting objectives: removal of forest products while maintaining the structural complexity, biodiversity, and ecosystem services of natural forest stands. Traditional silvicultural practices do not reflect these broadened objectives for forest sustainability.
[0003] Devising sustainable harvesting recommendations that take into account multiple management criteria such as timber yields, structural complexity, biodiversity, or other desirable forest attributes is challenging. This is because the spatial pattern of a partial harvest will have important implications for understory light levels for regeneration. Given the relatively limited dispersal distances of most North American tree species, the spatial distribution of seed trees will also have strong effects on the distribution and abundance of regenerating trees. Perhaps most important from an economic standpoint, the spatial pattern of a harvest determines the degree of release from competition for unharvested trees, with potentially dramatic effects on growth and survival of these individual trees. The long-term consequences of a given partial harvest for forest structure and value will be a function of both the immediate configuration of the residual stand and the subsequent growth of residual trees. Understanding the effects of partial harvests on interactions between individual trees is important to the development of sustainable management of forest ecosystems, particularly in uneven-aged stands of mixed species.
I of l9
[0004] Spatially-explicit, data-based forest models are often used to account for spatial complexity when developing management regimes/schemes for mixed, uneven aged forest stands. This class of ecological models can represent fine-scale spatial interactions between individual trees and has played an important role in advancing our understanding of the effects of processes such as natural disturbance, dispersal, and pathogens on forest stand dynamics. Yet, they have rarely been used in forest management because of a perceived lack of predictive power. In contrast, empirically rich growth and yield models have provided such predictive power, but only in a very limited set of environmental and structural conditions, primarily because these models cannot capture fine-scale spatial interactions between individuals in mixed stands.
[0005] Assessing the effects of partial harvesting on stand development in the context of multiple management goals also requires evaluation of a very large number of alternative interventions, e.g., partial harvests, with the goal of identifying those that optimize desired forest attributes. Traditionally, there have been two general approaches to this challenge. Growth and yield models can optimize multiple forest goals but require that the problem, that is, the objective function to be optimized, has a closed form solution. However, complex spatially-explicit interactions between individual trees for space cannot be expressed in closed form and as a result, this approach is not suitable to evaluate management alternatives that aim to manipulate structural complexity in mixed stands through partial harvests. The second approach relies on individual-based forest simulators to capture the effects of a specific management harvest intervention, e.g., harvest percentage of species X in a given spatial pattern, on mixed stand development at the expense of generality. It is often difficult to make recommendations using this approach because of the large number and range of input variables to be evaluated and the highly non-linear interactions between individual trees and demographic processes.
SUMMARY
[0006] Methods for evaluating forest management and harvesting schemes are disclosed. In some embodiments, the methods include the following: using a data-based, spatially-explicit forest simulator, evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, the first output data including a plurality of data fields; evaluating each of the plurality of data
fields to determine whether it meets predetermined criteria; assigning a first value to those of the plurality of data fields that meets predetermined criteria and a second value to those of the plurality of data fields that does not meet predetermined criteria; for the plurality of data fields, generating a corresponding set of the first and second values; training and cross- validating an untrained support vector machine using the first set of management and harvesting schemes and the corresponding first and second values thereby developing a trained support vector machine; and using the trained support vector machine, screening a second set of forest management and harvesting schemes to determine which of the second set of forest management and harvesting schemes as applied to the sample forest stand would generate an output data including a predetermined percentage of the plurality of data fields that meets the predetermined criteria if evaluated using the data-based forest simulator.
[0007] Systems for evaluating forest management and harvesting schemes are disclosed. In some embodiments, the systems include the following: a simulator module including a data-based, spatially-explicit forest simulator for evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, the first output data including a plurality of data fields; a scheme evaluation module for assigning a first value to those of the plurality of data fields that meets predetermined criteria and a second value to those of the plurality of data fields that does not meet predetermined criteria; a training module for training and cross-validating an untrained support vector machine using the first set of management and harvesting schemes and the corresponding first and second values thereby developing a trained support vector machine; and a screening module for using the trained support vector machine to screen a second set of forest management and harvesting schemes to determine which of the second set of forest management and harvesting schemes as applied to the sample forest stand would generate an output data including a predetermined percentage of the plurality of data fields that meets the predetermined criteria if evaluated using the data-based forest simulator. [0008] computer-readable mediums having computer-executable instructions for evaluating forest management and harvesting schemes are disclosed. In some embodiments, the computer-readable mediums have computer-executable instructions including the following: using a data-based, spatially-explicit forest simulator, evaluating
a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, the first output data including a plurality of data fields; evaluating each of the plurality of data fields to determine whether it meets predetermined criteria; assigning a first value to those of the plurality of data fields that meets predetermined criteria and a second value to those of the plurality of data fields that does not meet predetermined criteria; training and cross-validating an untrained support vector machine using the first set of management and harvesting schemes and the corresponding first and second values thereby developing a trained support vector machine; and using the trained support vector machine, screening a second set of forest management and harvesting schemes to determine which of the second set of forest management and harvesting schemes as applied to the sample forest stand would generate an output data including a predetermined percentage of the plurality of data fields that meets the predetermined criteria if evaluated using the data-based forest simulator.
BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The drawings show embodiments of the disclosed subject matter for the purpose of illustrating the invention. However, it should be understood that the present application is not limited to the precise arrangements and instrumentalities shown in the drawings, wherein:
[0010] FIG. 1 is a diagram of a method according to some embodiments of the disclosed subject matter; and
[0011] FIG. 2 is a schematic diagram of a system according to some embodiments of the disclosed subject matter.
DETAILED DESCRIPTION
[0012] Generally, the disclosed subject matter relates to methods and systems for evaluating forest management and harvesting schemes, which include the partial harvesting of forest stands. As shown in FIGS. 1 and 2, methods and systems according to the disclosed subject matter include the use of a traditional data-based, spatially-explicit forest simulator to train a support vector machine (SVM) learning algorithm. The combination of a traditional data-based, spatially-explicit forest simulator with a SVM
allows for the consideration of fine-scale spatial interactions between individual trees in mixed forest stands while evaluating a large number of forest management and harvesting schemes over a short period of time. In some embodiments, methods and systems according to the disclosed subject matter are embodied as computer-executable instructions included on a computer-readable medium.
[0013] Machine learning methods are powerful tools for the analyses of complex data and highly non-linear relationships such as those that typically occur in mixed forest stands. Among them, SVMs are considered a new generation of learning algorithms. SVMs have provided excellent results across many disciplines, require a minimal amount of fine-tuning, and theoretically guarantee performance, that is, they are capable of finding the absolute error minimum of any classification problem. Thus, they are generally more robust than other types of machine learning algorithms such as artificial neural networks (ANN). Training an ANN requires initialization with random numbers so finding the absolute error minimum cannot be guaranteed even if the adjustable parameters are set correctly, i.e., learning rate, momentum constant, number of hidden layers, etc.
[0014] Referring now to FIG. 1, some embodiments include a method 100 of evaluating forest management and harvesting schemes. At 102, using a data-based, spatially-explicit forest simulator, the application of each of a first set of forest management and harvesting schemes to a sample forest stand is evaluated, e.g., the simulator predicts the result of applying each scheme to a sample forest stand having particular known characteristics.
[0015] In one embodiment, each scheme evaluated had a particular number of input parameters. For example, in some embodiments, the following twenty-one input parameters were input into a data-based, spatially-explicit forest simulator known as SORTIE, a simulator developed by scientists at the Institute of Ecosystem Studies in
Millbrook, NY: (1) periodicity of harvest (in years); (2) radius of harvest area (in meters); (3) number of harvest sites within a stand (300 x 300 meters in the simulations); (4-21) the percent of trees to be harvested for each of nine species included in the model, including (4-12) the first orders trees in descending order by diameter at breast height (dbh), harvesting the largest trees first and (13-21) the second orders trees based on the degree of clumping (intra- and interspecific), with more crowded trees selected first. In this way,
forestry interventions that aim to harvest large, valuable trees and release subcanopy trees were mimicked. The range of input parameters is typically based on the plot size selected for the simulations and realistic parameter values.
[0016] Still at 102, for some embodiments of the disclosed subject matter, the results generated by the simulator include a first output data having a plurality of data fields such as timber yield, structural complexity, and biodiversity. Of course, in other embodiments, different data fields can be generated depending on the optimization goals, i.e., desired outputs, of the schemes being evaluated.
[0017] In some embodiments, for timber yield, the goal of the scheme was to increase economic yield from timber harvest for a 200 year period starting with a 100-year old stand. This time horizon was selected in an effort to capture the most productive and diverse age of the simulated stands. However, different time horizons can be used to suit other rotation schedules or management criteria. For timber yield, timber values were calculated from biomass estimates of individual trees and the amount of merchantable timber. Biomass was estimated using US Forest Service dimension analysis equations. Timber values were calculated using forest service data for board-feet per tree and prices from a 2005 survey.
[0018] Other data fields were selected to evaluate whether the scheme promotes the long-term structural integrity and biodiversity of the stand. The biodiversity and structural complexity of the managed stand after 200 years of harvesting interventions according to a particular scheme was compared with that of an unharvested stand of 100 years. To ensure the dynamic integrity of the stand, these metrics were compared over 100 years for both stands, that is, biodiversity and structural complexity for a 100 year period for an unharvested stand (100 to 200 yrs) was compared with that of a harvested stand after harvesting was completed (300 to 400 yrs). In essence, a 100 year old stand was managed according to a particular scheme for timber yield for 200 years with the goal of leaving it in a dynamic condition similar to the one in which it was found. This approach better reflects the manner in which forestry operations are scheduled, allows for harvest rotation between stands at the landscape level, and more importantly, ensures that forest are harvested to substantially increase or maximize timber yields while retaining the dynamic sustainability of the stand.
[0019] Still at 102, for some embodiments of the disclosed subject matter, evaluating the structural complexity includes comparing a probability density function of the sample forest stand after harvesting the sample forest stand according to each of the first set of forest management and harvesting schemes for a time period against a probability density function of the sample forest stand when it has not been harvested for the time period.
[0020] The rationale for structural complexity as a management goal is that stands that contain a variety of structural components, i.e., large trees and a good mix of species, are generally more likely to support a variety of resources and species. Most existing definitions assume that this is a static metric, e.g., frequency distribution of height or dbh. In some embodiments according to the disclosed subject matter, a structural complexity index was developed that explicitly incorporates the notion that stands are changing with time rather than assume that attributes such as height or dbh assume an optimal and static distribution. This index was calculated by comparing the probability density function (PDF) (p) for a stand that has been harvested to the PDF (q) for a stand that has not been harvested (and 200 years younger), over the period of time of interest to management (t) (100 years for this exercise) as follows:
[0021] Using the frequency distribution of diameter (dbh) as an example, q (x, f) is the probability of finding a tree with a dbh of x ± S (where δ is a small number) in an unharvested stand at time step t while p(x, f) is the probability of finding a tree of the same size at that same time step in a harvested stand. In this way we can calculate a single number that measures changes in the probability of finding trees for all possible diameters (dbh) at every time step as a result of forest harvest or changing forest population dynamics (i.e., succession). In effect, Eqn. 1 constitutes a measure of Kullblack-Leibler (KL) divergence, which measures the distance between the two dbh distributions. The smaller the KL divergence, the greater the similarity in dbh distribution between harvested and unharvested stands, therefore our goal was to minimize KL divergence between stands. In order to analyze the outcome of a harvest as a whole, we averaged the KL divergences for the nine species to create a stand-level measure of structural complexity, K.
[0022] Evaluating the biodiversity includes comparing an abundance of each species of the sample forest stand after harvesting the sample forest stand according to each of the first set of forest management and harvesting schemes for a time period against an abundance of each species of the sample forest stand when it has not been harvested for the time period. Harvesting can result in drastic changes in population numbers, which may not be reflected in dbh distributions. These changes in population abundance can in turn result in the loss of species. In fact, harvesting led to a significant drop in the numbers of individuals for most species regardless of how the stand was harvested. Therefore, typically, one goal of management schemes is to minimize the total percent change in population of the harvested stand in years 300-400 relative to an unharvested stand in years 100-200. Decreases in species abundance were considered but not increases because only the former represent a loss of biodiversity. However, if increases in the population of one species came at the expense of another species, the subsequent drop in the other species' abundance would have affected the measure of biodiversity. [0023] The metric of change in biodiversity was calculated as the average percent decrease in population number for those species that decreased in abundance across the 100 post-harvest years (300-400) for the harvested stand and years 100-200 for the unharvested stand. The abundance of species i was calculated in the unharvested stand as follows:
where TtJ is the number of trees of species i in year/. The abundance of the species i in the harvested stand was calculated as:
400
H1 = 2X [Eqn. 3] y=300
[0024] The percent change in abundance in years 300-400 compared to years 100-200 is:
Δ, is positive or negative depending on whether the abundance of species i increased of decreased as a result of harvesting at any time step. Since decreases in biodiversity were considered, overall change in biodiversity, B1 was calculated as follows:
B = sum(abs(min(Δ, 0)))/number of species [Eqn. 5] with a greater value of B indicating lower biodiversity.
[0025] At 104, each of the plurality of data fields is evaluated to determine whether it meets predetermined criteria. For example, if the goal is to identify schemes that promote biodiversity, only those schemes that have a biodiversity value above/below a certain predetermined threshold will be considered as desirable with respect to biodiversity, each output data field is evaluated against predetermined criteria to determine whether it is beneficial with respect to that particular parameter.
[0026] Then, at 106, each of the plurality of data fields that meets the predetermined criteria is assigned a first value and each of the plurality of data fields that does not meet the predetermined criteria is assigned a second value. In some embodiments, the first value is a positive one (+1) and the second value is a negative one (-1). Of course, other ways of rating the data fields can be used.
[0027] At 108, an untrained support vector machine is trained and cross- validated using the first set of management and harvesting schemes and the corresponding first and second values thereby developing a trained support vector machine. Pattern recognition algorithms such as SVMs rely on a set of training parameters to learn the connections between inputs, i.e., harvest scheme input parameters described above, and outputs, i.e., desired forest attributes such as timber yield, structural complexity, and biodiversity as indicated by the corresponding first and second values.
[0028] For example, in some embodiments, for each scheme simulated, the simulator is used to generate a plurality of data fields that include the following: (1) identification of
scheme simulated; and (2) a vector of length 19, which includes 9 values (+1 or -1) of KL divergence between harvested and unharvested stands, i.e., one for each species, 9 values (+1 or -1) of percent change in population abundance between harvested and unharvested stands, i.e., again, one for each species, and one value (+1 or -1) for timber yield. A portion of the output from the simulator is typically used to train the SVM. The remainder of the output from the simulator is used for cross-validation. Cross-validation aims to prevent over-fitting, which is when the SVM is biased to the training set and is unable to generalize when presented with data not included in the training set. Separation achieved with a training set of input parameters should be checked for generality through cross- validation with a set of patterns outside of the training set. The performance of the SVM classifiers in cross-validation can be measured against a random number generator that outputs +1 's and -1 's with the same frequency found in the training set. If the SVM outperformed the best in the random set, it is an indication that the trained SVM was able to generalize.
[0029] At 110, the trained support vector machine is used to screen a second set of forest management and harvesting schemes to determine which of the second set of forest management and harvesting schemes as applied to the sample forest stand would generate an output data including a predetermined percentage of the plurality of data fields that meets the predetermined criteria if evaluated using the data-based forest simulator. Because SVMs are typically only capable of classification into defined categories, rather than training the SVM to output a real number for any given quality, e.g., timber value, the output forest characteristics were binned as either "positive (+1)" or "negative (-1)". For example, each data field of each scheme screened will return either a positive (+1) or a negative (-1). The SVM makes this determination based on what it "learned" from the training set data. Various sorting or filtering can than be done to identify schemes that return desired outputs.
[0030] Referring now to FIG. 2, other embodiments of the disclosed subject matter include a system 200 of evaluating forest management and harvesting schemes. System 200 generally includes a plurality of interactive modules such as a simulator module 202, a scheme evaluation module 204, a training module 206, and a screening module 208.
[0031] Simulator module 202 includes a data-based, spatially-explicit forest simulator 210 for evaluating a first set of forest management and harvesting schemes 212 as applied to a sample forest stand, which are defined by stand input parameters 214, to generate a first output data 216. First output data 216 includes a plurality of data fields 218, e.g., which can include at least one of timber yield, structural complexity, and biodiversity.
[0032] In scheme evaluation module 204, each of plurality of data fields 218 is evaluated against a corresponding predetermined criteria 220 and assigned a first value 222 if meets the predetermined criteria and assigned a second value 224 if it does not meet the predetermined criteria. In some embodiments, first value 222 is a positive one (+1) and second value 224 is a negative one (-1).
[0033] In training module 206, an untrained support vector machine (not shown) is trained and cross- validated using first set of management and harvesting schemes 212 and corresponding first and second values 222, 224, thereby developing a trained support vector machine 226.
[0034] In screening module 208, trained support vector machine 226 is used to screen a second set of forest management and harvesting schemes 228 to determine which of the second set of forest management and harvesting schemes as applied to the sample forest stand would generate an output data 230 including a predetermined percentage of plurality of data fields 218 that meets the predetermined criteria if evaluated using data-based forest simulator 210. Output data 230 can be sorted to find those schemes that best achieve desired outputs. For example, second set of forest management and harvesting schemes 228 can be screened to return only a list of those schemes where at least seventy-five percent of the output data fields meet the predetermined criteria, i.e., were assigned a positive one (+1). Various screening strategies can be used, e.g., second set of forest management and harvesting schemes 228 can be screened to return only a list of those where the timber yield meets the predetermined criteria or by any other output data field.
[0035] Methods and systems and according to the disclosed subject matter provide advantages and benefits over known methods and systems. Accounting for spatial complexity is tremendously important for the development of sustainable management
practices in natural and managed ecosystems. In forestry, a better understanding of the effect of interactions between individual trees would allow us to develop sustainable and economically profitable forest management and harvesting schemes, particularly in mixed- species, uneven-aged stands. Such schemes should consider complex tradeoffs in the maintenance of a diverse stand composition and structure that meets ecological and aesthetic goals, while providing economically viable yields from harvests. Methods and systems according to the disclosed subject matter couple machine learning methods with a data-based, spatially-explicit forest simulator to identify timber harvesting regimes that can optimize and balance multiple management goals, specifically timber yields, with the maintenance of biodiversity and structural complexity, in mixed forest stands.
[0036] In tests, methods and systems according to the disclosed subject matter were able to screen a large number of harvesting/management schemes, e.g., approximately 100,000, and identify those that optimized desired outcomes. To quantify its performance, optimized and random parameter sets were sorted separately, analyzed, and compared the top 10% of input harvest parameters in the optimized and random set. Methods and systems according to the disclosed subject matter generated an improvement in all three desired forest attributes: 12.88% in change in biodiversity; 8.04% in structural complexity; and 57.83% in timber yield.
[0037] Developing sustainable harvesting regimes that optimize multiple forest attributes requires sophisticated computing approaches, such as machine learning methods. These methods can take advantage of existing data-based forest simulators to explore the effects of a very large number and range of management interventions on stand development. This approach is also more suitable to optimization of multiple management goals such as the maintenance of stand structural complexity and biodiversity, which may more relevant to the long-term sustainability of harvest opportunities. Methods and systems according to the disclosed subject matter combine machine learning methods with a data-based, forest simulator to screen a large of number of harvesting regimes and identify those that can optimize and balance multiple management goals, specifically timber value, biodiversity, and structural complexity, in mixed forest stands.
[0038] In tests, methods and systems according to the disclosed subject matter were able to find harvesting parameters in the screening set that outperformed, i.e., higher
timber value, lower KL-divergence, and lower change in abundance, any set of the 1,500 harvest schemes used in the training and cross-validation set. This means that the methods and systems according to the disclosed subject matter can be trained using a relatively small number of example schemes, generalize from what it learns, screen a large number of possible schemes by emulating the forest simulator, and identify those harvest regimes/schemes that meet the criteria of interest. It is estimated that the methods and systems according to the disclosed subject matter produces a significant time savings relative to using a simulator to individually evaluate each potential harvesting scheme.
[0039] Although the disclosed subject matter has been described and illustrated with respect to embodiments thereof, it should be understood by those skilled in the art that features of the disclosed embodiments can be combined, rearranged, etc., to produce additional embodiments within the scope of the invention, and that various other changes, omissions, and additions may be made therein and thereto, without parting from the spirit and scope of the present invention.
Claims
1. A method of evaluating forest management and harvesting schemes, said method comprising: using a data-based, spatially-explicit forest simulator, evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, said first output data including a plurality of data fields; evaluating each of said plurality of data fields to determine whether it meets predetermined criteria; assigning a first value to those of said plurality of data fields that meets predetermined criteria and a second value to those of said plurality of data fields that does not meet predetermined criteria; training and cross- validating an untrained support vector machine using said first set of management and harvesting schemes and said corresponding first and second values thereby developing a trained support vector machine; and using said trained support vector machine, screening a second set of forest management and harvesting schemes to determine which of said second set of forest management and harvesting schemes as applied to said sample forest stand would generate an output data including a predetermined percentage of said plurality of data fields that meets said predetermined criteria if evaluated using said data-based forest simulator.
2. The method according to claim 1, wherein said plurality of data fields include at least one of timber yield, structural complexity, and biodiversity.
3. The method according to claim 2, wherein evaluating said structural complexity includes comparing a probability density function of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against a probability density function of said sample forest stand when it has not been harvested for said time period.
4. The method according to claim 2, wherein evaluating said biodiversity includes comparing an abundance of each species of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against an abundance of each species of said sample forest stand when it has not been harvested for said time period.
5. The method according to claim 1, wherein said first value is a positive one (+1) and said second value is a negative one (-1).
6. The method according to claim 1, wherein said data-based, spatially-explicit forest simulator includes means for considering fine-scale spatial interactions between individual trees in mixed forest stands.
7. The method according to claim 1 , wherein said first set of forest management and harvesting schemes includes partial harvesting of said sample forest stand.
8. A system of evaluating forest management and harvesting schemes, said system comprising: a simulator module including a data-based, spatially-explicit forest simulator for evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, said first output data including a plurality of data fields; a scheme evaluation module for assigning a first value to those of said plurality of data fields that meets predetermined criteria and a second value to those of said plurality of data fields that does not meet predetermined criteria; a training module for training and cross-validating an untrained support vector machine using said first set of management and harvesting schemes and said corresponding first and second values thereby developing a trained support vector machine; and a screening module for using said trained support vector machine to screen a second set of forest management and harvesting schemes to determine which of said second set of forest management and harvesting schemes as applied to said sample forest stand would generate an output data including a predetermined percentage of said plurality of data fields that meets said predetermined criteria if evaluated using said data-based forest simulator.
9. The system according to claim 8, wherein said plurality of data fields include at least one of timber yield, structural complexity, and biodiversity.
10. The system according to claim 9, wherein evaluating said structural complexity includes comparing a probability density function of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against a probability density function of said sample forest stand when it has not been harvested for said time period.
11. The system according to claim 9, wherein evaluating said biodiversity includes comparing an abundance of each species of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against an abundance of each species of said sample forest stand when it has not been harvested for said time period.
12. The system according to claim 8, wherein said first value is a positive one (+1) and said second value is a negative one (-1).
13. The system according to claim 8, wherein said data-based, spatially-explicit forest simulator includes means for considering fine-scale spatial interactions between individual trees in mixed forest stands.
14. A computer-readable medium having computer-executable instructions for evaluating forest management and harvesting schemes, said instructions comprising: using a data-based, spatially-explicit forest simulator, evaluating a first set of forest management and harvesting schemes as applied to a sample forest stand to generate a first output data, said first output data including a plurality of data fields; evaluating each of said plurality of data fields to determine whether it meets predetermined criteria; assigning a first value to those of said plurality of data fields that meets predetermined criteria and a second value to those of said plurality of data fields that does not meet predetermined criteria; training and cross-validating an untrained support vector machine using said first set of management and harvesting schemes and said corresponding first and second values thereby developing a trained support vector machine; and using said trained support vector machine, screening a second set of forest management and harvesting schemes to determine which of said second set of forest management and harvesting schemes as applied to said sample forest stand would generate an output data including a predetermined percentage of said plurality of data fields that meets said predetermined criteria if evaluated using said data-based forest simulator.
15. The computer-readable medium according to claim 14, wherein said plurality of data fields include at least one of timber yield, structural complexity, and biodiversity.
16. The computer-readable medium according to claim 15, wherein evaluating said structural complexity includes comparing a probability density function of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against a probability density function of said sample forest stand when it has not been harvested for said time period.
17. The computer-readable medium according to claim 15, wherein evaluating said biodiversity includes comparing an abundance of each species of said sample forest stand after harvesting said sample forest stand according to each of said first set of forest management and harvesting schemes for a time period against an abundance of each species of said sample forest stand when it has not been harvested for said time period.
18. The computer-readable medium according to claim 14, wherein said first value is a positive one (+1) and said second value is a negative one (-1).
19. The computer-readable medium according to claim 14, wherein said data-based, spatially-explicit forest simulator includes means for considering fine-scale spatial interactions between individual trees in mixed forest stands.
0. The computer-readable medium according to claim 14, wherein said first set of forest management and harvesting schemes includes partial harvesting of said sample forest stand.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US95107507P | 2007-07-20 | 2007-07-20 | |
| US60/951,075 | 2007-07-20 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2009015069A1 true WO2009015069A1 (en) | 2009-01-29 |
Family
ID=40281741
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/US2008/070601 WO2009015069A1 (en) | 2007-07-20 | 2008-07-21 | Methods and systems of evaluating forest management and harvesting schemes |
Country Status (1)
| Country | Link |
|---|---|
| WO (1) | WO2009015069A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537432A (en) * | 2018-04-03 | 2018-09-14 | 北京林业大学 | A kind of the Multifunctional effect appraisal procedure and system of orest management planning |
| CN109002621A (en) * | 2018-07-25 | 2018-12-14 | 中国林业科学研究院资源信息研究所 | A kind of mean height and diameter of a cross-section of a tree trunk 1.3 meters above the ground calculation method for taking neighborhood and geographical difference into account |
| WO2022069802A1 (en) * | 2020-09-30 | 2022-04-07 | Aalto University Foundation Sr | Machine learning based forest management |
| CN117011085A (en) * | 2023-08-18 | 2023-11-07 | 中国林业科学研究院林业研究所 | Method and device for determining first time interval of forest stand |
| CN117575086A (en) * | 2023-11-24 | 2024-02-20 | 日照市岚山区高兴镇农业综合服务中心 | Forestry district ecological management method and system based on risk prediction |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6041173A (en) * | 1996-11-26 | 2000-03-21 | Assisi Software Corporation | Method and system for simulating forest management |
| US20030014342A1 (en) * | 2000-03-27 | 2003-01-16 | Vande Pol Mark E. | Free-market environmental management system having insured certification to a process standard |
| US20050087110A1 (en) * | 2002-11-18 | 2005-04-28 | Yoshikazu Kobayashi | Reforestation project system and reforestation project program |
-
2008
- 2008-07-21 WO PCT/US2008/070601 patent/WO2009015069A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6041173A (en) * | 1996-11-26 | 2000-03-21 | Assisi Software Corporation | Method and system for simulating forest management |
| US20030014342A1 (en) * | 2000-03-27 | 2003-01-16 | Vande Pol Mark E. | Free-market environmental management system having insured certification to a process standard |
| US20050087110A1 (en) * | 2002-11-18 | 2005-04-28 | Yoshikazu Kobayashi | Reforestation project system and reforestation project program |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108537432A (en) * | 2018-04-03 | 2018-09-14 | 北京林业大学 | A kind of the Multifunctional effect appraisal procedure and system of orest management planning |
| CN109002621A (en) * | 2018-07-25 | 2018-12-14 | 中国林业科学研究院资源信息研究所 | A kind of mean height and diameter of a cross-section of a tree trunk 1.3 meters above the ground calculation method for taking neighborhood and geographical difference into account |
| CN109002621B (en) * | 2018-07-25 | 2022-11-15 | 中国林业科学研究院资源信息研究所 | Forest stand average height and breast diameter calculation method considering neighborhood and geographic difference |
| WO2022069802A1 (en) * | 2020-09-30 | 2022-04-07 | Aalto University Foundation Sr | Machine learning based forest management |
| CN117011085A (en) * | 2023-08-18 | 2023-11-07 | 中国林业科学研究院林业研究所 | Method and device for determining first time interval of forest stand |
| CN117575086A (en) * | 2023-11-24 | 2024-02-20 | 日照市岚山区高兴镇农业综合服务中心 | Forestry district ecological management method and system based on risk prediction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107766929B (en) | Model analysis method and device | |
| CN106779087A (en) | A kind of general-purpose machinery learning data analysis platform | |
| CN108351985A (en) | Method and apparatus for large-scale machines study | |
| CN108446214B (en) | DBN-based test case evolution generation method | |
| CN106503723A (en) | A kind of video classification methods and device | |
| Phillips et al. | The elusive empirical shadow of growth convergence | |
| WO2009015069A1 (en) | Methods and systems of evaluating forest management and harvesting schemes | |
| CN113537807B (en) | Intelligent wind control method and equipment for enterprises | |
| Zou et al. | A novel network security algorithm based on improved support vector machine from smart city perspective | |
| CN111986027A (en) | Abnormal transaction processing method and device based on artificial intelligence | |
| CN115510042A (en) | Power system load data filling method and device based on generation countermeasure network | |
| Odegua | Applied Machine Learning for Supermarket Sales Prediction | |
| Dhandayudam et al. | An improved clustering algorithm for customer segmentation | |
| CN107273918A (en) | A kind of sample data classification determines method and apparatus | |
| CN106815600A (en) | For the depth co-ordinative construction and structural chemistry learning method of human behavior identification | |
| Drăgulescu et al. | Predicting assignment submissions in a multi-class classification problem | |
| CN111292062B (en) | Network embedding-based crowd-sourced garbage worker detection method, system and storage medium | |
| Maghsoudi et al. | Representing the New Model for Improving K-Means Clustering Algorithm based on Genetic Algorithm | |
| CN112819499A (en) | Information transmission method, information transmission device, server and storage medium | |
| Binoti et al. | Estimation of height of Eucalyptus trees with neuroevolution of Augmenting Topologies (NEAT) | |
| CN110363302A (en) | Training method, prediction technique and the device of disaggregated model | |
| CN115130619A (en) | Risk control method based on clustering selection integration | |
| DeNicola et al. | A stochastic model of epilithic algal succession and patch dynamics in streams | |
| CN114529063A (en) | Financial field data prediction method, device and medium based on machine learning | |
| Maleki et al. | A Novel Hybrid Model of Scatter Search and Genetic Algorithms for Software Cost Estimation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 08782128 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 08782128 Country of ref document: EP Kind code of ref document: A1 |