3.1. GIS Applications
GIS applications in the forest sector are vast and diverse [8
]. Typically, the GIS allows a range of application from essential functions for spatial analysis to the application and development of statistical and mathematical modelling [9
]. In terms of forest utilization, the most important functions, which have been widely and variously explored, are focused on forest road network planning and aimed to support the identification of the most suitable harvesting systems or the evaluation of forest accessibility [8
]. Slope, roughness, and several other morphological parameters, for example, may present significant spatial constraints for enterprises especially when high-quality stands of timber are located at considerable distances from existing logging roads. Such problems are aggravated during poor economic conditions, as forest companies may not be able to access the quality of timber required to maintain profits [10
]. In recent years, multiple studies have shown the advantage of precision forestry and the use of geographic information system (GIS) in forest road network analysis and planning [11
]. This new approach allowed to further stress the forest road multifunctional aspects; in this direction, a recent paper highlighted like the combination of GIS-multiple-criteria decision-making (MCDM) [12
] approaches can properly assist forest road planning for forestry and touristic purposes [13
]. One of the first GIS applications in forest road planning was the determination of mean extraction distance considering the actual forest road network. Thus, it was possible to assess and plan the most suitable extraction system focusing on the minimization of the total costs of timber extraction. Enache et al. [14
] and Duka et al. [15
] developed different GIS models to calculate extraction distance, including correction factors to consider road sinuosity and slope variation.
Another interesting GIS application is possible in the designing of a planned forest road network, considering various environmental and logistic parameters. In recent years, Enache et al. [16
] used GIS and multi criteria analysis (MCA) in Romania, to identify an optimal road network considering management, costs, environmental, and social factors. The developed model was then tested and validated. A reduction of mean skidding distance from 864 m to 255–268 m was reported, leading to an increase in productivity of timber extraction from 7.5 m3
/h to 11.7 m3
/h and to an increased contribution margin from 21.2 €/m3
to 25.1 €/m3
. Enhancement of forest infrastructure reduced CO2
emissions due to timber harvesting and transport from 8.52 kg/m3
to 7.3 kg/m3
. The integrated GIS and MCA is one of the preferred tools for the correct forest management for assessing the relative importance of the economic, environmental, and social criteria. In particular, the GIS approach is necessary in sites where it is fundamental to encourage the creation of a biomass supply chain network. In fact, a low economic value of biomass stock allows no room for incorrect choices concerning trails, location of energy plants, and logistical transportation. The importance of a correct optimization to reduce transportation costs is highlighted by several studies published over the past decade [17
Another GIS model to automatically design skid-trail networks in order to reduce skidding costs and soil disturbances was implemented by Contreras et al. [23
]. This model simulates tree-bunch locations, creates a feasible skid-trail network across the harvest unit, estimates skidding cost and soil recovery cost for each skid-trail segment, and finds the best network design that connects each tree-bunch to landings, reducing both skidding and soil recovery costs.
In forest road network planning, GIS could be even more efficient if integrated with other technological instruments such as linear programming. A new skid trails pattern developed in this way resulted in a 16.4% increase of work productivity, 44% reduction of skid trails length, and 44.29 ton/ha of prevented soil losses, compared to not-planned interventions [24
Finally, concerning forest road network planning, Parsakhoo et al. [25
] demonstrated the efficiency of GIS-planned skid trails networks also in close-to-nature forestry interventions, consisting of little interventions with low biomass removal, which are mostly applied in protected areas. In this study, a skid trail network was developed to extract marked trees from stand sites to landing sites using a GIS-based decision support system (DSS). The techniques were applied in a stand where single trees are felled in close-to-nature conditions. Results showed that on average the length of the route decreased by 6.65% to 19.22%.
An example of the efficiency of GIS in forest operations planning is given in Figure 2
, which is a real application by the authors of precision forest harvesting in a Central Italy turkey oak coppice forest yard, utilizing the methodology proposed by Picchio et al. [11
]. In the above-mentioned figure, it is possible to see the great difference in terms of skid trails number and length between a GIS-planned (Figure 2
b) and a not-planned (Figure 2
The subsequent step of GIS application in forest utilization is aimed to plan interventions considering both road network, environmental, and topographic conditions of the area. This implies the concept of “Accessibility” or “Openness”. That is, considering the peculiarity of a forest estate/parcel, (i) is it possible to access the forest in order to harvest wood material? (ii) Which machinery is more suitable for this?
The first scientific work with this focus was implemented last year by Synek and Klimanek [26
], who developed a GIS model that indicates the most environmentally friendly extraction system, considering topographic, climate, machine equipment, and stand characteristics.
In 2016, Laschi et al. [27
] devised a GIS approach to classify a forest ownership according to three accessibility classes: Accessible, barely accessible, and not accessible. In this study, the concept of accessibility was designed according to the time needed to reach a specific point in forest ownership starting from the actual forest network.
However, according to forest harvesting management, a concept of accessibility based on time criterion is less reliable than one based on distance from the existing road, even better if this distance is a “real distance”, calculated considering slope instead of a horizontal distance [28
One of the most recent papers that used GIS to classify a forest ownership in Turkey, according to the distance from existing roads, and so identifying the most suitable extraction system according to this distance, is Caliskan et al. [29
]. The authors developed a model for timber extraction systems analysis, considering terrain morphology and, secondary, forest road network. Chainsaw–small-size cable crane (36.76%) and chainsaw–medium-size cable crane (27.94%) were selected as the most suitable timber extraction systems for the steep terrain study area, according to the model. They were followed by chainsaw–forest tractor (23.52%), chainsaw–agriculture tractor (10.29%), and chainsaw–sledge yarder (1.49%) [29
]. A slightly more complex project was produced by Picchio et al. [28
]. In this study, a GIS model was developed to classify two forest estates in Central Italy as accessible or inaccessible areas for extraction by tractor with a winch and/or lightweight cable yarder, which are the most common extraction systems used in that area. Then, a “Least Cost Path” analysis was performed to design new hypothetical skid trails that could make all ownerships surface accessible. Finally, the authors made a survey analysis to validate the developed model and found a strong correlation between the model validation and the actual accessibility in the forest areas.
In the above-mentioned works, the study area included entire forest estates, usually hundreds or thousands of hectares. The first attempt to use GIS in a more accurate analysis and planning of a forest road network on a relatively small area (ca. 20–60 ha), dimensions comparable to a single forest yard intervention, was carried out in Picchio et al. [11
]. In this study, three different GIS models, developed in previous studies, for the identification of forest winch accessible areas, were applied and field-validated in two different study areas in Central Italy. All three models showed optimum results in the prevision of winching areas in both study areas.
The next step should be the development of correction factors to further increase model efficiency, and their integration with forest harvesting scheduling models, such as the one developed by Vopenka et al. [30
], inserting all these systems within the forest management plan.
3.2. GNSS Instruments
The global navigation satellite system (GNSS) has become one of the most popular techniques for fast and accurate positioning in open spaces. This method has been used in many areas of mapping because of the low cost and simplicity of use, compared to the standard surveying technique [31
] defined the development of this system 30 years ago as the most important recent innovation in the field of remote sensing, replacing traditional (manual, analogic, etc.) surveying methods with GNSS survey methods. The availability to collect information on machine performance and function allow the collection of information such as distances traveled, machine status, and productivity of the machine at each location.
Regarding the contribution of GNSS to the forest utilization sector, it is possible to identify two different groups of scientific contributions: The first group includes articles that show GNSS applications for the improvement of various forest utilization techniques; the second one is composed of articles that analyze and/or try to improve GNSS position accuracy under forest canopy cover.
Regarding the first group of articles in the analyzed period (2013–2019), most scientific papers focused on the possibility of using GNSS in order to define machine work productivity. Though this, at first glance, could seem interesting only from an academic point of view, it is actually an essential evaluation for technical-practical aims. In fact, all economic and logistic evaluation in forest yards, for example stumpage value evaluation, is strongly linked to work productivity.
GNSS devices allow a very interesting possibility to monitor operational time in forest logging operations with high level of mechanization (processor, cable crane, and wheeled skidder), with an error of working time evaluation ranging from 2.75% to 7% in comparison to the considerably more complex and costly chronometric field relieves [33
Moreover, GNSS devices also showed feasibility for helicopter logging productivity analysis [37
]. GNSS technology could be also used to evaluate forest utilization impacts, considering that reducing soil and topsoil impacts, linked with forest operations, is a central and essential aim of sustainable forest management research [38
]. In fact, Veal et al. [39
] studied the accuracy of this application to define areas where repeated traffic could lead to an excessive soil compaction or other undesirable impacts on the local environment.
Ellis et al. [40
] showed the importance of GNSS technology in the evaluation of soil impact, claiming that such devices are currently more precise than UAV-LiDAR in detecting skid trails network linked to forest utilization.
According to the above, equipping all forest machines, and not only the newest ones, with GNSS devices, and sharing the recorded position with the control Institutions personnel, could be a powerful instrument for forest operations supervision, as it allows remote control of machines [33
Together with GNSS, modern harvesters are equipped with computers able to collect and store a great deal of data on stem measurements, harvesting production, and machine parameters. These data are automatically collected by the measurement system unit at the harvesting head, linked to the OBC systems of the machine [41
]. The information is recorded using a de facto standard called StanForD (standard for forest data and communication), which is used by all major manufacturers of cut-to-length (CTL) machines across the world [42
]. There is a number of standard files produced when operating with StanForD, including: Apt (cross-cutting instructions), prd (production files), pri (production individual files), drf (operational monitoring data), and stm (individual stem data) [44
]. Apt files are produced by the user, whereas the others are produced by the machine computer. These files can be used by forestry companies and contractors to manage production aspects [45
]. Although StanForD files contain useful data, the process of extracting, storing, and analyzing them is complex. Software, for example SilviA, is used by both John Deere and Waratah to make StanForD files easier to be read, created, and edited. Advanced software, such as Timber Office from John Deere and Ponsse Opti from Ponsse, can be used to manage operations and for fleet control [46
]. Moreover, several harvester control systems have a navigation system capable of displaying a range of base layer maps, which can include raster and vector data, such as digital elevation models (raster feature), stand maps (polygon feature), and power lines (line feature). The operator can navigate with a map displaying stand boundaries as well as restricted or dangerous areas based on the outputs presented on the machine’s computer screen. Additional functions such as recording points (e.g., features of interest) and calculating areas are available in some systems. An example of GIS data development of StanForD data from a forwarder OBC system is presented in Figure 3
Integrating GNSS and StanForD data could lead to various important and interesting forestry applications such as: Developing forest yield maps, useful for harvesting and management planning or evaluating work productivity to other parameters, for example stem diameter at breast eight (DBH), species, shift (day/ night), slope, and operator [46
The last application of GNSS technology analyzed by scientific research in the last years, with regard to the above cited first group of articles, is GNSS application for workers’ safety.
This is a key factor in sustainable forest management, especially considering that logging consistently ranks among the most dangerous occupations [47
]. With a rate of 136 fatal injuries per 100,000 workers (91 fatalities total) in 2016, logging workers had by far the highest fatal injury rate in the United States [49
]. Location-sharing devices, like global navigation satellite system–radio frequency (GNSS-RF) technologies, which share geographic coordinates, and radio-frequency identification (RFID) transmitters, capable of local relative positioning of worker proximity to equipment, have potential to increase workers’ safety on logging operations [50
GNSS-RF devices that facilitate location sharing in off-the-grid areas without cellular service include receivers that could synchronize with mobile phones or tablets by using Bluetooth and transmit GNSS locations throughout local networks, as well as dedicated radios with GNSS capabilities. These latter devices were originally developed for military applications, but in recent years, they have been increasingly considered for worker safety uses in natural environments [50
Location sharing technology could increase team safety during logging operations and facilitate both injury prevention and response. For instance, location sharing devices with help alerts could allow isolated individuals to notify the coworkers or off-site response services of any emergency. In cases of incapacitation, automatic position updates may help coworkers if an individual requires aid. In both cases, geographic coordinates are shared to assist response efforts [51
Considering the consistent importance of this technology’s implementation in forest yards, many authors have addressed this issue.
Keefe et al. proposed the development of a system that allows operators to see the location of ground workers and other equipment on a digital display in real time, using location-sharing devices [52
This system may increase workers’ safety during logging, thereby reducing the incidence of fatal and near-fatal injuries [50
]. Another scientific work developed a system in which virtual geofences encompassing high-risk areas during logging operations can be monitored by a mechanism to detect and alert operators of the presence of ground workers in hazardous areas [53
]. Zimbelmann et al. expanded this concept to include the detection of workers and equipment in motion through real-time proximity analysis [55
However, to improve the potential of this technology and guarantee workers’ safety, it is fundamental to enhance positioning accuracy [56
From this point, it is possible to explore the second group of scientific works, i.e., papers that analyze and try to improve GNSS positioning accuracy under forest canopy cover; this is a topic in which scientific research has been highly active in recent years [3
First of all, it is important to provide a brief introduction on the problem.
GNSS technology works well in unobstructed open spaces and all GNSS manufacturers provide the accuracy of their receivers assuming them to work without any obstacles [63
The fact that forest canopy cover may suppress the satellite signal is not taken into account [68
]. Many factors linked with forest conditions can influence positioning accuracy. Forest canopy is a barrier for signal propagation, so the final radio wave is weak, and the reflection causes an elevated signal-to-noise ratio, which is called multipath effect [69
]. The base idea of multipath is strictly connected to signal reflections from objects located near the receiver, which finally causes an error in distance measurements. There are many software and hardware solutions to weaken this effect, however the strong forest influence is not completely solved yet [72
]. Moreover, the multipath effect is strengthened by high moisture conditions and by the presence of leaves [72
Considering the above, today the position of one object (a machine or an operator) under forest canopy cover can be detected by GNSS technologies with an accuracy of 2–7 m with an average of 3 m and a maximum of about 20 m [3
That being the case, enhancing GNSS positioning accuracy is a primary goal and it could be reached by integrating GNSS with other technologies like RTK, RBN DGPS [74
], IMU sensors [3
], or RF [75
]. Using RTK, it is possible to reach centimeter-level accuracy [76
]. Moreover, it is possible to have a positioning error below 1 m in forest conditions also by using RBN DGPS [74
Kaartinen et al. [3
] integrated a GNSS receiver with an IMU sensor to test the positioning accuracy compared to simple GNSS technology and it was found that GNSS-IMU reached an error of 0.7 m while the mean error of simple GNSS technology was between 4 and 9 m. An overall summary of various technologies of positioning correction is given in Table 1
To summarize, with current GPS-GNSS differential postprocessing technology, or by integrating GNSS with IMU sensors, it is possible to reach a positioning precision with an error below 1 m. Up to now, many precision forestry applications have been allowed, but still without the possibility of a complete machine automation.
3.3. Machine Sensors
Another important application in forest operation improvement is the integration, within forest machines like harvesters or forwarders, of different kinds of sensors able to detect particular parameters, which could be important to support forest utilization in various ways. As reported by Borz [77
] in 2016, equipping forestry machines with a sensor system is very important to improve forest operations, and this should be done not only for CTL machinery, but these sensors should be also implemented on winch-assisted machineries like winch skidders. Very few papers have, in fact, analyzed sensors’ usage in winch-assisted vehicles.
Analyzing recent scientific papers on the matter, the most investigated usage of sensors in the forestry sector has been dedicated to work productivity evaluation, which, as written in the above paragraphs of this review, is a very important parameter for multiple purposes.
The first kind of sensor used for this application has been the machinery vibration sensor, which has been demonstrated to be very useful in machine productivity analysis, with particular reference to delay time identification, quantification, and explanation [78
]. In particular [79
], a mean difference between the cycle times was obtained of an estimated <1 s.
Other studies have instead employed machine monitoring systems and OBC to evaluate work productivity. Manner et al. [80
] used the John Deere Timber Link to evaluate forwarding productivity and Brewer et al. [81
] adopted StanForD through Ponsse Opti2 information system. Both studies reached good performances in work productivity evaluation.
Again, with regards to productivity analysis, multi-camera security systems showed the possibility of analyzing the work cycles of a John Deere 540G cable skidder [82
Concerning the implementation of various kinds of sensors on forestry machines, an interesting publication is Ding et al. [83
], who described a novel stumpage detection method for forest harvesting, based on a 2D laser scanner and infrared thermal imager. According to this method, the stumpage information is captured by the two sensors and fused via image fusion and laser matching. Then, stumpage features can be extracted from the fused information. Next, an SVM (support vector machine) classifier model is constructed by sample training, according to the feature data. Finally, in contrast to SVM with default parameters, three different optimization algorithms were proposed to optimize SVM parameters. The results showed that this method could reach a detection rate of 96.7%. Ultrasonic sensors could instead be used for measurement of ruts depth in forwarding operation focusing on environmental impact characterization; according to the study results, ultrasonic sensors provide sufficient accuracy to characterize depth of ruts in 1.5 m long segments of strip-roads, including dynamic data on depth and length of ruts after each pass [84
]. Using several different tools mounted on forestry machines, Marinello et al. [85
] highlighted a clear relationship between roughness parameters and the vibration intensities in order to monitor and study the effects of different road surfaces on vehicle stability.
With regard to the economic function of forests, a very interesting work is Sandak et al. [86
]. In this study, a sensorized processor was developed featuring the following sensors: Near infrared (NIR) spectrometer and hyperspectral cameras to identify surface defects, stress wave and time of flight sensors to estimate timber density, hydraulic flow sensor to estimate cross-cutting resistance, and delimbing sensors to estimate branches number and approximate position. Moreover, the processor prototype also deployed an RFID UHF system, which allowed the identification of the incoming tree and marked each log individually, relating the quality parameters recorded to the physical item and tracing it along the supply chain [86
What this sensorized processor makes possible is an evaluation of timber quality and, linking with StanForD data and a machine monitoring system like Timber Link, a complete evaluation of stumpage value.
Another application of electronic sensors in forest operation is positioning sensors. Correct positioning of a harvester head—the part in which knowing the position is essential for cutting operation automation—is currently limited by two problems: The above-cited GNSS positioning error and the position of GNSS receiver, which is not located on the processor head but somewhere within the cabin. Starting from these claims, Lindroos et al. [87
] analyzed various positioning methods i.e., angle/range sensors, tilt sensors, joint sensors, and IMU sensors. Angle and range-based methods derive the position by estimating the angles and/or distances between a given number of sensors and the harvester head. Joint estimating methods calculate the position based on the geometry of the crane combined with direct measurements of joint angles and displacements. Tilt sensors estimate the static position by sensing the head’s orientation with respect to the earth’s gravitational field. Finally, IMU sensors use a combination of accelerometers, gyroscopes, and magnetometers [87
]. The authors’ analysis highlights the joint sensors and IMUs as the methods with the greatest potential for implementation thanks to their accuracy, cost-effectiveness, and solidity [87
Some kinds of sensors are, of course, also usable to pursue one of the most important aims of sustainable forest management, i.e., forest workers’ safety. One application is on-field continuous measuring of cable tensile force during winching operation, which can be performed by cable tensile force measurement device like Cable-Bull®
SR22/800 XR sensor (manufactured by the Honigmann Industrielle Elektronik GmbH) [88
]. A summarizing view of the contribution of various sensors and electronic devices to SFM is provided in Table 2
An Overview of Forest Machine Automation Purpose
In machinery, robotics, and engineering, the degree to which a specific task of a given machine is automated is known as the level of automation (LOA).
In the forest utilization world, the most advanced technology machines are harvesters and, secondly, forwarders. Nevertheless, these require almost complete operator input and so even mechanized harvesting or forwarding extraction method could be considered to have a LOA 0 [90
LOA 1 products, such as computer-assistance for motion control, entered the market only a few years ago [91
]. In particular, there are cranes equipped with motion sensors [92
], computer support to the boom-tip [93
], reduced crane vibration systems [93
], active suspension [94
], and hydraulic valves equipped with software control [95
Future challenges are reaching LOA 2 and LOA 3, which respectively consist of an operator who choose an operating action according to the machine suggestion and in tele-operated or unmanned forest vehicles [91
To summarize, forest engineers using GIS are currently able to analyze a forest parcel, assessing the best extraction systems and automatically tracing new skid trails network to optimize bunching-extraction with minimal costs, minimum environmental impacts, and maximizing workers’ safety. GIS-created files can be integrated in harvesters’ and forwarders’ OBCs in order to display on the screen an optimal skid trails network and the geofences of restricted-access areas. GNSS technology could allow one to position a forest operating machine with a precision of about 3 m that could decrease at values lower than 1 m by integrating GNSS with other technologies such as RF or IMUs or adopting differential correction. Harvesters and forwarders are equipped, or could be, with various sensors which allow one to identify the position of trees to be harvested; have a better knowledge of the processor head position; and perform an on-field evaluation of work productivity, utilization costs, timber quality analysis, and wood tracing.
Considering the actual technologies, the above is the best possible and obtainable result. To reach the subsequent step, consisting of tele-operated or unmanned forest vehicles, the integration on forest machines of simultaneous localization and mapping (SLAM) algorithms seems to be needed [100