Geointelligence against Illegal Deforestation and Timber Laundering in the Brazilian Amazon

: Due to the characteristics of the Southern Amazonas Mesoregion (Mesorregi ã o Sul do Amazonas, MSA), conducting on-site surveys in all licensed forestry areas (Plano de Manejo Florestal, PMFS) is an impossible task. Therefore, the present investigation aimed to: (i) analyze the use of geointelligence (GEOINT) techniques to support the evaluation of PMFS; and (ii) verify if the PMFS located in the MSA are being executed in accordance with Brazilian legislation. A set of twenty-two evaluation criteria were established. These were initially applied to a “standard” PMFS and subsequently replicated to a larger area of 83 PMFS, located in the MSA. GEOINT allowed for a better understanding of each PMFS, identifying illegal forestry activities and evidence of timber laundering. Among these results, we highlight the following evidences: (i) inconsistencies related to total transport time and prices declared to the authorities (70% of PMFS); (ii) volumetric information incompatible with o ﬃ cial forest inventories and / or not conforming with Benford’s law (54% of PMFS); (iii) signs of exploitation outside the authorized polygon limits (51% of PMFS) and signs of clear-cutting (43% of PMFS); (iv) no signs of infrastructure compatible with licensed forestry (24% of PMFS); and (v) signs of exploitation prior to the licensing (19% of PMFS) and after the expiration of licensing (5%).


Introduction
According to official data, about 97% of the forests in the State of Amazonas were still preserved until 2012. However, in that same year, its annual deforestation rate reached 523 km 2 (an increase of 4% when compared to 2011), and more recently reached a total of 1421 km 2 during 2019, which represents about 14% of all that was deforested in the entire Amazon region in that period [1]. Such an increase in deforestation is due to multiple factors, among them the recent infrastructure works in the region, mainly roads. In addition to this, with the strengthening of preventive and repressive measures against illegal deforestation in the states of Mato Grosso, Rondônia, and Pará, historically the largest deforesters, the Amazonas state is today the last timber frontier in the Brazilian Amazon: illegal wood extracted from public areas there is of excellent quality and can be obtained at even lower cost than in other regions [2,3].
Forestry activities in the Brazilian Amazon presuppose the existence of authorization from a competent environmental administrative authority, which can be carried out in the form of a Sustainable Sustainable Forest Management Plan (Plano de Manejo Florestal Sustentável, MFS), according to the provisions of the Brazilian Forest Code [4].
In brief, for the approval of a PMFS, the landowner must initially carry out a forestry inventory of the respective area, which contains the exact location of each tree, its species, and volume, and also identify protected specimens and seedlings such as those located in non-exploitable zones (e.g., hillsides and riversides).
Based on these data, the PMFS is prepared, following a series of legal requirements such as the dimensions of roads, storage yards, and with the limitation on the maximum volume to be explored as 25 m³/ha [4].
These requirements aim to ensure that forestry activities will take place in a sustainable and planned way, respecting the mechanisms of sustaining the ecosystem. Indeed, one of the main elements to be considered by the environmental authorities in the licensing of a PMFS is the capacity of the natural regeneration of the managed species, and the landowners must commit themselves and carry out the forest exploitation in such a way as to promote the regeneration cycle of the area until new specimens replace those that have been cut [5].
Once the forest inventory and the PMFS is approved, designated trees may then be felled and sold. A PMFS account is created in the Forest Origin Document Management System (Sistema de Controle do Documento de Origem Florestal, SisDOF) of the Brazilian Institute of Environment and Renewable Natural Resources (Instituto Brasileiro de Meio Ambiente e dos Recursos Naturais Renováveis, IBAMA).
The SisDOF account works similarly to a bank account: forestry credits equivalent to the volume and species registered of the forest inventory are "deposited". These credits will be deducted with each sale and transferred from the PMFS to the account of the respective buyer at SisDOF by registering the respective transaction in the system and issuing a transport permit, called the Forest Origin Document (Documento de Origem Florestal, DOF), which follows the product to the destination and must be presented to the authorities in the case of inspections during the transportation [6]. Figure 1 represents the legal process of forest products from the PMFS to the timber industry. Unfortunately, the experience in combating illegal deforestation in the Brazilian Amazon shows that, in addition to the irregularities practiced inside the private rural properties authorized to carry out sustainable forest management (such as cutting intensity beyond the allowed limit, cutting of protected species or any species in Areas of Permanent Preservation (Áreas de Preservação Permanente, APP) such as hillsides and river banks), PMFS are also used by criminals for the legalization and trade of illegally obtained forest products from other private and public non-authorized areas such as indigenous lands and national parks [7][8][9][10][11][12][13][14]. Figure 2 represents the visual outline of illegal forestry in Brazil.
ISPRS Int. J. Geo-Inf. 2020, 9, x FOR PEER REVIEW 3 of 32 Unfortunately, the experience in combating illegal deforestation in the Brazilian Amazon shows that, in addition to the irregularities practiced inside the private rural properties authorized to carry out sustainable forest management (such as cutting intensity beyond the allowed limit, cutting of protected species or any species in Areas of Permanent Preservation (Áreas de Preservação Permanente, APP) such as hillsides and river banks), PMFS are also used by criminals for the legalization and trade of illegally obtained forest products from other private and public nonauthorized areas such as indigenous lands and national parks [7][8][9][10][11][12][13][14]. Figure 2 represents the visual outline of illegal forestry in Brazil.  [6,9].
Due to the characteristics of the Amazon region, conducting on-the-spot inspections with regularity and frequency in all areas of PMFS has proven to be unfeasible, notably due to the resources and time required, which greatly undermines state action in combating these crimes [14,15]. The present study seeks to demonstrate that such verification can be carried out successfully, based on geointelligence (GEOINT) techniques, especially from the multitemporal analysis of satellite sensor image analysis and its comparison with the official data contained in the SisDOF.
The rest of this article is organized as follows. In Section 2, a review of the literature is conducted regarding GEOINT and its application to the evaluation of PMFS in the Brazilian Amazon. Section 3 presents the objectives, materials, and methodology proposed. Section 4 presents the results, while the discussion and conclusions are presented in Sections 5 and 6, respectively.

Literature Review
Geointelligence (GEOINT) is defined by the National Geospatial-Intelligence Agency (NGA) as: "[...] exploitation and analysis of images and geospatial information to describe evaluate and visualize the physical features and georeferenced activities in the world. Geointelligence consists of: Images, Image Intelligence and Geospatial Information" [16]. The definition provided by the NGA proves to be quite satisfactory for the purposes of this work as it allows us to identify the three components of GEOINT, namely: (i) imagery; (ii) image intelligence (IMINT) and; (iii) geospatial  [6,9]).
Due to the characteristics of the Amazon region, conducting on-the-spot inspections with regularity and frequency in all areas of PMFS has proven to be unfeasible, notably due to the resources and time required, which greatly undermines state action in combating these crimes [14,15]. The present study seeks to demonstrate that such verification can be carried out successfully, based on geointelligence (GEOINT) techniques, especially from the multitemporal analysis of satellite sensor image analysis and its comparison with the official data contained in the SisDOF.
The rest of this article is organized as follows. In Section 2, a review of the literature is conducted regarding GEOINT and its application to the evaluation of PMFS in the Brazilian Amazon. Section 3 presents the objectives, materials, and methodology proposed. Section 4 presents the results, while the discussion and conclusions are presented in Sections 5 and 6, respectively.

Literature Review
Geointelligence (GEOINT) is defined by the National Geospatial-Intelligence Agency (NGA) as: "[...] exploitation and analysis of images and geospatial information to describe evaluate and visualize the physical features and georeferenced activities in the world. Geointelligence consists of: Images, Image Intelligence and Geospatial Information" [16]. The definition provided by the NGA proves to be quite satisfactory for the purposes of this work as it allows us to identify the three components of GEOINT, namely: (i) imagery; (ii) image intelligence (IMINT) and; (iii) geospatial information (Table 1).
However, one has to keep in mind that this division is merely conceptual, and, in practice, the three components merge, as shown in Figure 3. Table 1. Components of GEOINT (adapted from [16]).

Imagery
Visual record of any resource or environment (natural or man-made), its objects, and related activities that contain the respective geographic positioning data produced by satellites, aerial platforms, unmanned aerial vehicles (UAVs), or other similar means.

Image Intelligence
Technique of extracting useful information from the interpretation or analysis of images (for example: the use of digital image processing techniques, vegetation indexes, classifications, etc.) and collateral data (all that may contribute to the understanding and interpretation of images, including intelligence data and information from other sources).

Geospatial Information
Information related to the Earth's surface that identifies the location, geometry, and attributes of environments, constructions, objects, resources or phenomena that occupy it (maps, statistical data, tables, digital files, geoprocessing operations etc.).
ISPRS Int. J. Geo-Inf. 2020, 9, x FOR PEER REVIEW 4 of 32 information (Table 1). However, one has to keep in mind that this division is merely conceptual, and, in practice, the three components merge, as shown in Figure 3. Table 1. Components of GEOINT (adapted from [16]).

Imagery
Visual record of any resource or environment (natural or man-made), its objects, and related activities that contain the respective geographic positioning data produced by satellites, aerial platforms, unmanned aerial vehicles (UAVs), or other similar means.

Image Intelligence
Technique of extracting useful information from the interpretation or analysis of images (for example: the use of digital image processing techniques, vegetation indexes, classifications, etc.) and collateral data (all that may contribute to the understanding and interpretation of images, including intelligence data and information from other sources).

Geospatial Information
Information related to the Earth's surface that identifies the location, geometry, and attributes of environments, constructions, objects, resources or phenomena that occupy it (maps, statistical data, tables, digital files, geoprocessing operations etc.). Despite its great increase in recent years, the scientific literature on GEOINT and its application in the evaluation of PMFS in the Brazilian Amazon is still relatively scarce. Part of this is due to the fact that the exploitation of wood through forest management is selective: only two to five trees per hectare (<25 m³/ha), and satellite images available to the public are generally of medium or low resolution [14,15,17]. It was believed that it was not possible to evaluate PMFS on a large scale without resorting to high-resolution images, however, studies in the field of remote sensing have demonstrated for several years that it is possible to detect and evaluate the quality of logging with medium resolution satellite images and even estimate the impacts of exploitation on forest biomass [18][19][20][21][22][23][24][25][26]. In addition, a recent investigation demonstrated that Landsat images are useful for detecting logging activity in the same year of exploitation with the identification of storage yards larger than 290 m² [27]. Despite its great increase in recent years, the scientific literature on GEOINT and its application in the evaluation of PMFS in the Brazilian Amazon is still relatively scarce. Part of this is due to the fact that the exploitation of wood through forest management is selective: only two to five trees per hectare (<25 m 3 /ha), and satellite images available to the public are generally of medium or low resolution [14,15,17]. It was believed that it was not possible to evaluate PMFS on a large scale without resorting to high-resolution images, however, studies in the field of remote sensing have demonstrated for several years that it is possible to detect and evaluate the quality of logging with medium resolution satellite images and even estimate the impacts of exploitation on forest biomass [18][19][20][21][22][23][24][25][26]. In addition, a recent investigation demonstrated that Landsat images are useful for detecting logging activity in the same year of exploitation with the identification of storage yards larger than 290 m 2 [27].
It should also be noted that several Brazilian government agencies such as the Brazilian Federal Environmental Agency (Instituto Brasileiro do Meio Ambiente e Recursos Naturais Renováveis, IBAMA) and the Brazilian Federal Police (Polícia Federal, PF) have used these technologies with relative success to detect irregularities in PMFS. This is possible because, even if there are serious limitations to the detection of the felling and dragging of trees in low and medium spatial resolution images, the infrastructure of the PMFS (consisting of storage yards and primary and secondary roads) can be detected and evaluated using this type of image. This allows for the mapping of selective wood exploitation, but also, in certain cases, estimation of their respective impacts [25,28,29].

Objectives
The objectives of this research were:

•
Evaluate how the use of GEOINT based on medium resolution imagery can be used on understanding the real situation of forest exploitation in areas authorized by the government in the Brazilian Amazon; and • Verify if the PMFS of MSA are being carried out in compliance with the respective environmental standards, being truly sustainable, or whether laundering the timber illegally extracted from other areas.

Materials
The data for this research were obtained from public agencies, namely: • Identifying data for motor vehicles authorized to transport forest products in the Amazon as well as their carrying volume capacity from IBAMA [41]; and v.
SisDOF data referring to all commercial timber transactions carried out between 2013 and 2018 from IBAMA [42].

General Overview
The research consisted of two stages: (i) the definition, based on an exploratory analysis of available materials (satellite imagery, geospatial information and collateral materials) and the study of the behavior of a "standard" PMFS, of a set of standard indicators (or analysis criteria) that could be applied to evaluate the regularity of PMFS; and (ii) the application of these criteria in relation to a number of PMFS in the defined study area.
The GEOINT methodology applied during this research was strongly influenced by the Kahaner Intelligence Cycle [43] and more specifically, its adaptation by Mellión [44]. Figure 4 illustrates how the Meillón model was applied for a better specification and methodological division of the two stages above described by dividing the research into the following steps:

•
Planning phase (in red), composed in this case of the activities of definition of the standard PMFS and the study area as well as the data and resources necessary for the research; • Collect phase (in green), consisting of the obtaining and initial validation of the data and resources necessary for the evaluation of the "standard" PMFS and the study area; • Processing phase (in purple), consisting of the definition of tools and methods to be applied to each of the datasets (spatial and non-spatial) as well as the creation of the definitive dataset; • Analysis phase (in blue), consisting of the integrated analysis of the data obtained and prepared in the previous phases and its documentation for later evaluation and comparison of the results; • Diffusion phase (in orange), consisting of the presentation of the information obtained and the respective conclusions.  In the following sections, each of the four methodological phases are synthetically described.

Planning
In this phase, the "standard" PMFS and the area of study area were defined as well as the data and resources necessary for the research.
A "standard" PMFS was selected from an area that started its forestry activities in 1993 and has been certified by the Forest Stewardship Council (FSC) since 1997, the first to obtain this certification in Brazil [45]. The standard PMFS was distributed in nine different rural properties with a total area of approximately 270,000 ha, of which 139,000 ha were destinated for forestry management activities, subdivided into 47 production units, identified by the respective years of the cutting cycle ( Figure 5) In the following sections, each of the four methodological phases are synthetically described.

Planning
In this phase, the "standard" PMFS and the area of study area were defined as well as the data and resources necessary for the research. A "standard" PMFS was selected from an area that started its forestry activities in 1993 and has been certified by the Forest Stewardship Council (FSC) since 1997, the first to obtain this certification in Brazil [45]. The standard PMFS was distributed in nine different rural properties with a total area of approximately 270,000 ha, of which 139,000 ha were destinated for forestry management activities, subdivided into 47 production units, identified by the respective years of the cutting cycle ( Figure 5) [46]. The choice of the "standard" PMFS was based on research carried out in 2002 [47], which characterized them as follows: • The PMFS area was planned to be exclusively exploited under the selective logging regime at a proportion of 2.5% per year, between 2002 and 2022; • 100% of commercial trees in the area were inventoried and the PMFS, approved by environmental agencies, was duly designed to prevent overexploitation; • In addition to inventorying commercial species, the company carried out an inventory of 0.5% of the area (1/200), one year before and one year after exploitation in order to understand the impact of logging on forest regeneration and species diversity. • When settling in the region, the company came across about 50 families that already lived there, without, however, having the respective property titles. They were individuals who already lived there. The company chose to assign property rights to them instead of filing lawsuits to expel them. Most of these people were also employed at the company. Additionally, the choice of the Southern Amazonas Mesoregion (Mesorregião Sul do Amazonas, MSA) as the study area ( Figure 6) was justified by the fact that traded forestry products and illegal deforestation rates had been increasing yearly between 2014 and 2018 [1][2][3], and that according to SisDOF, 63% of all products traded in the Amazonas in the same period were from the MSA [42]. Likewise, most of the administrative fines for environmental offenses that occurred in Amazonas between 1996 and 2017 (88%) also took place at the MSA [39], which makes this region the one with the highest incidence of irregularities and administrative fines in the State of Amazonas. The choice of the "standard" PMFS was based on research carried out in 2002 [47], which characterized them as follows: • The PMFS area was planned to be exclusively exploited under the selective logging regime at a proportion of 2.5% per year, between 2002 and 2022; • 100% of commercial trees in the area were inventoried and the PMFS, approved by environmental agencies, was duly designed to prevent overexploitation; • In addition to inventorying commercial species, the company carried out an inventory of 0.5% of the area (1/200), one year before and one year after exploitation in order to understand the impact of logging on forest regeneration and species diversity. • When settling in the region, the company came across about 50 families that already lived there, without, however, having the respective property titles. They were individuals who already lived there. The company chose to assign property rights to them instead of filing lawsuits to expel them. Most of these people were also employed at the company.
Additionally, the choice of the Southern Amazonas Mesoregion (Mesorregião Sul do Amazonas, MSA) as the study area ( Figure 6) was justified by the fact that traded forestry products and illegal deforestation rates had been increasing yearly between 2014 and 2018 [1][2][3], and that according to SisDOF, 63% of all products traded in the Amazonas in the same period were from the MSA [42]. Likewise, most of the administrative fines for environmental offenses that occurred in Amazonas between 1996 and 2017 (88%) also took place at the MSA [39], which makes this region the one with the highest incidence of irregularities and administrative fines in the State of Amazonas.

Data Collection
The data collection phase consisted of obtaining and initially validating the data and resources mentioned in Section 3.2. Most data and resources are publicly accessible and were directly obtained through the websites of the above-mentioned bodies and entities. Additionally, the following were The study covered a period ranging from 2014 to 2018, corresponding to 83 licensed PMFS as shown in Figure 7.

Data Collection
The data collection phase consisted of obtaining and initially validating the data and resources mentioned in Section 3.2. Most data and resources are publicly accessible and were directly obtained through the websites of the above-mentioned bodies and entities. Additionally, the following were

Data Collection
The data collection phase consisted of obtaining and initially validating the data and resources mentioned in Section 3.2. Most data and resources are publicly accessible and were directly obtained through the websites of the above-mentioned bodies and entities. Additionally, the following were collected: (i) data of the polygons of the PMFS located in the MSA and all transactions carried out through SisDOF in the state of Amazonas, between 01/01/2014 and 08/21/2018, were obtained from the respective environmental agencies through the Federal Police's Environmental Crimes Enforcement Division [31,42]); (ii) data from the standard PMFS (polygons, inventories, etc.) were kindly provided by the technical responsible person [46]; and (iii) inspection and forensic reports from IBAMA and PF were obtained from the respective bodies, under commitment of their use for statistical purposes [36,37].

Processing
The processing phase consisted of the definition of the tools and methods to be applied to each of the datasets (spatial and non-spatial) as well as the creation of the definitive dataset. Initially, an exploratory evaluation of all reports produced by the PF and IBAMA as well as all other spatial and non-spatial data obtained, was carried out in order to identify the most common irregularities found on those PMFS. This resulted in the establishment of 22 evaluation criteria for PMFS ( Table 2).

Criteria For Non-Spatial Data Justification
2.1 Product received after valid dates Illegal activity.

DOF Canceled
Suspicious. Possible fraud for the use of a single DOF and its forest credits to transport more than one cargo.

DOF issued during rainy season (December to March [48])
Suspicious. In the rainy season, logging and transport is not viable in many regions, once transport routes are largely unpaved and in poor traffic conditions.

Irregularities related to the forestry inventory
It is possible to use the data from the RADAM project to assess whether the data on species and volumes presented in forestry inventories are consistent with those contained in the official survey [12]. In addition, the analysis carried out in this research indicates that the mathematical model of Benford [51], can also be very useful in the evaluation of forest inventories.

Total volume traded identical to the authorized volume
Suspicious. Losses occurs during the process of forest exploitation (logs crack during the felling of trees), as also is common the presence of tree hollows not measured in the original forestry inventory.
2.14 Fines for labor law violations Demonstrates past/present irregularities.

Exploitation intensity over 25 m 3 /ha
Illegal. The legislation of the state of Amazonas establishes that the maximum exploitation intensity of PMFS is up to 25 m 3 /ha [17]. Criteria 1.1 to 1.7 (spatial data) and 2.10, 2.11, and 2.14 (administrative fines applied for environmental or labor irregularities) focused on assessing the regularity of forest management, with a positive response to any of these items meaning irregularity. Criteria 2.1 to 2.9, 2.11 to 2.13, and 2.15 (non-spatial data) were based on the information of the forestry inventory (2.12) and all data trade transactions in SisDOF (remaining items). These data, when analyzed together with the satellite imagery, indicate a greater or lesser degree of suspicion of the existence of frauds for the laundering and trade of forestry products illegally extracted from other areas.
Once the evaluation criteria were defined, the following methods and software were used to process and organize the image dataset:  This dashboard was also combined with the electronic spreadsheet developed by Steven Miller [56] to test the applicability of the statistical model known as Benford's law for evaluating the data of forest inventories and roundwood volumes traded through SisDOF. This statistical model is This dashboard was also combined with the electronic spreadsheet developed by Steven Miller [56] to test the applicability of the statistical model known as Benford's law for evaluating the data of forest inventories and roundwood volumes traded through SisDOF. This statistical model is applicable to specific datasets produced naturally (such as river areas, population, molar weights, atomic numbers, and mortality rates) and has been successfully used for decades to detect economic and financial fraud [57]. It considers the frequency probability of the first valid digit, that is, the leftmost digit, different from zero [56][57][58]. In contrast to a homogeneous distribution of the nine possible digits (1 to 9), which would correspond to the expected frequency of any digit being 11.11%, Benford's law notes that the frequency follows a logarithmic behavior, represented by the formula: P(digit =) = log 10 1 + 1 digit (1) which means that the probability of the first digit being 1 is something close to 30.1%, whereas digit 9 will appear in only about 4.6% of the observations [51,[56][57][58][59][60]. Figure 9 allows us to understand, graphically, what would be the valid first digit and how the distribution of that digit occurs according to Benford's law.
ISPRS Int. J. Geo-Inf. 2020, 9, x FOR PEER REVIEW 13 of 32 applicable to specific datasets produced naturally (such as river areas, population, molar weights, atomic numbers, and mortality rates) and has been successfully used for decades to detect economic and financial fraud [57]. It considers the frequency probability of the first valid digit, that is, the leftmost digit, different from zero [56][57][58]. In contrast to a homogeneous distribution of the nine possible digits (1 to 9), which would correspond to the expected frequency of any digit being 11.11%, Benford's law notes that the frequency follows a logarithmic behavior, represented by the formula: which means that the probability of the first digit being 1 is something close to 30.1%, whereas digit 9 will appear in only about 4.6% of the observations [51,[56][57][58][59][60]. Figure 9 allows us to understand, graphically, what would be the valid first digit and how the distribution of that digit occurs according to Benford's law.

Analysis
The above methodology was initially applied to the "standard" PMFS, which presented adequacy to most of the 22 evaluated criteria, except for the following findings: (i) criterion 1.6: exploitation outside the authorized polygon in 2009 (~1 ha) and 2016 (~3 ha); (ii) criterion 2.5: existence of a large number of transactions with the identity of Internet Protocol numbers (IPs) and; (iii) criterion 2.6: 31.1% of the total volume of roundwood traded during the rainy season. Furthermore, criteria 2.5 and 2.6 were not considered suspicious since the identity of IP numbers was not corroborated by other indications of fraud (criteria 2.3, 2.4, 2.6-2.9), and the standard PMFS had good transport infrastructure, being located near a paved highway.
This methodology was subsequently replicated to the 83 PMFS located in the MSA.

Preliminary Results Obtained during the Processing Phase
During the processing phase, the analysis and comparison of data on roundwood volumes traded through SisDOF in the State of Amazonas from the years 2014 to 2018 as well as the official forest inventories contained in RADAM where it was found that the respective volumetric data presented a logarithmic distribution conforming to the theoretical model known as Benford's law [51,60]. The evaluation of conformance with Benford's law was based on the Chi-square adherence test. Although the level of significance most widely used in the verification of Benford's law is 5%, it

Analysis
The above methodology was initially applied to the "standard" PMFS, which presented adequacy to most of the 22 evaluated criteria, except for the following findings: (i) criterion 1.6: exploitation outside the authorized polygon in 2009 (~1 ha) and 2016 (~3 ha); (ii) criterion 2.5: existence of a large number of transactions with the identity of Internet Protocol numbers (IPs) and; (iii) criterion 2.6: 31.1% of the total volume of roundwood traded during the rainy season. Furthermore, criteria 2.5 and 2.6 were not considered suspicious since the identity of IP numbers was not corroborated by other indications of fraud (criteria 2.3, 2.4, 2.6-2.9), and the standard PMFS had good transport infrastructure, being located near a paved highway.
This methodology was subsequently replicated to the 83 PMFS located in the MSA.

Preliminary Results Obtained during the Processing Phase
During the processing phase, the analysis and comparison of data on roundwood volumes traded through SisDOF in the State of Amazonas from the years 2014 to 2018 as well as the official forest inventories contained in RADAM where it was found that the respective volumetric data presented a logarithmic distribution conforming to the theoretical model known as Benford's law [51,60]. The evaluation of conformance with Benford's law was based on the Chi-square adherence test. Although the level of significance most widely used in the verification of Benford's law is 5%, it also uses a level of significance of 1% to minimize the probability of detecting the false evidence of fraud [58]. In both cases, eight degrees of freedom were used.
The following results were obtained: • The volumetric data of all transactions contained in SisDOF for the State of Amazonas for both the total volume by species as well as the total volume of species traded per hectare presented conformity with Benford's law; • The official volumetric data of the RADAM Project for both for total volume per species and total volume of species per hectare also showed conformity with the model; • The sum of the "Diameter at Breast Height" (DBH) values of all individuals of the same species, per hectare, or totals for the entire study area also proved to be in conformity (RADAM volumetric data); and • Changes in the data scale (daily, weekly, monthly, quarterly, half-yearly, or annual), did not significantly change the results obtained. Figure 10 shows that the values obtained through the SisDOF [42] and RADAM [38] found strong conformity with this mathematical model. also uses a level of significance of 1% to minimize the probability of detecting the false evidence of fraud [58]. In both cases, eight degrees of freedom were used.
The following results were obtained: • The volumetric data of all transactions contained in SisDOF for the State of Amazonas for both the total volume by species as well as the total volume of species traded per hectare presented conformity with Benford's law; • The official volumetric data of the RADAM Project for both for total volume per species and total volume of species per hectare also showed conformity with the model; • The sum of the "Diameter at Breast Height" (DBH) values of all individuals of the same species, per hectare, or totals for the entire study area also proved to be in conformity (RADAM volumetric data); and • Changes in the data scale (daily, weekly, monthly, quarterly, half-yearly, or annual), did not significantly change the results obtained. Figure 10 shows that the values obtained through the SisDOF [42] and RADAM [38] found strong conformity with this mathematical model. Similarly, it was found that the total volume traded in the State of Amazonas from 2014 to 2018 of Allatoma lineata (jequitibá) conformed with Benford's law ( Figure 11), but the same analysis for the traded volumes of Tabebuia serratifolia (ipê), a species that is known to be fraudulently overestimated in forestry inventories due to its high commercial value [12], did not conform to Benford's law ( Figure  12).  Similarly, it was found that the total volume traded in the State of Amazonas from 2014 to 2018 of Allatoma lineata (jequitibá) conformed with Benford's law ( Figure 11), but the same analysis for the traded volumes of Tabebuia serratifolia (ipê), a species that is known to be fraudulently overestimated in forestry inventories due to its high commercial value [12], did not conform to Benford's law ( Figure 12). Similarly, it was found that the total volume traded in the State of Amazonas from 2014 to 2018 of Allatoma lineata (jequitibá) conformed with Benford's law ( Figure 11), but the same analysis for the traded volumes of Tabebuia serratifolia (ipê), a species that is known to be fraudulently overestimated in forestry inventories due to its high commercial value [12], did not conform to Benford's law ( Figure  12).

Analysis Results
The results of the spatial analysis are presented for comparison purposes for both the standard PMFS and the 83 PMFS evaluated. Similarly, the results of non-spatial analysis are also presented for both the standard PMFS and the 83 PMFS evaluated as well as for all the PMFS in the State of

Analysis Results
The results of the spatial analysis are presented for comparison purposes for both the standard PMFS and the 83 PMFS evaluated. Similarly, the results of non-spatial analysis are also presented for both the standard PMFS and the 83 PMFS evaluated as well as for all the PMFS in the State of Amazonas and those located in the MSA, whenever the data were available.
Some figures are shown as examples of the irregularities found for a better understanding. PMFS were identified by its feature identification number (FID), with no mention of individualizing data such as names, geographic coordinates, etc.

Analysis Results
The results of the spatial analysis are presented for comparison purposes for both the standard PMFS and the 83 PMFS evaluated. Similarly, the results of non-spatial analysis are also presented for both the standard PMFS and the 83 PMFS evaluated as well as for all the PMFS in the State of Amazonas and those located in the MSA, whenever the data were available.
Some figures are shown as examples of the irregularities found for a better understanding. PMFS were identified by its feature identification number (FID), with no mention of individualizing data such as names, geographic coordinates, etc.    83 MSA PMFS: 20 PMFS (17% of the sample) had infrastructure incompatible with forest management practices, 10 referring to cases of no sign of yards and roads in the area and 10 referring to situations incompatible with PMFS and/or the volumes traded from these areas. Figure 16 shows complete absence of infrastructure and signs of selective logging inside PMFS. MSA PMFS: 20 PMFS (24% of the sample) had infrastructure incompatible with forest management practices, 10 referring to cases of no sign of yards and roads in the area and 10 referring to situations incompatible with PMFS and/or the volumes traded from these areas. Figure 16 shows complete absence of infrastructure and signs of selective logging inside PMFS.      Figure 19 shows presence of yards, roads and selective cutting signs inside three PMFS (4888, 2987 and 4191).  Figure 19 shows presence of yards, roads and selective cutting signs inside three PMFS (4888, 2987 and 4191).       STANDARD PMFS: 31.3%. It should be noted that the standard PMFS has a good access road structure and it is located close to a highway, which would, in theory, allow exploration throughout the year. Most PMFS located in Amazonas and specially in MSA, however, do not have such structure.

Spatial Data
MSA 83 PMFS: 24% of the areas had DOFs issued during the rainy season at a level higher than 15% of the total volume sold. Anyway, some of the properties had volumes extracted in the rainy season above the average such as FID 2262 (32.5%) and FID 3468 (44.7% and 37%). FIDs 4936 and 4499, despite being areas belonging to the same holder and being located very close, had their exploitation the in rainy season carried out quite differently, with 1% of exploitation during the rainy period in FID 4936, against 44.3% of exploitation at the same time for FID 4499 ( Figure 23). This finding, in addition to various other irregularities found for both FIDS, may show fraudulent transactions in SisDOF. However, this is an indicator that must be evaluated in each specific case, together with other factors such as the existence of roads and adequate infrastructure for the extraction and transportation of products, distance to the recipient, etc.  Figure 23). This finding, in addition to various other irregularities found for both FIDS, may show fraudulent transactions in SisDOF. However, this is an indicator that must be evaluated in each specific case, together with other factors such as the existence of roads and adequate infrastructure for the extraction and transportation of products, distance to the recipient, etc.   MSA 83 PMFS: 70% of the PMFS had DOFs issued whose total transport time and distance, between the sender and the receiver, would have resulted in an average speed greater than 40 km/h. FID 4034, also mentioned in the criterion 2.8, presented transport duration times that would result in speeds above 1000 km/h ( Figure 25). MSA 83 PMFS: 48% of the PMFS had DOFs issued whose total transport time and distance, between the sender and the receiver, would have resulted in an average speed greater than 40 km/h. FID 4034, also mentioned in the criterion 2.8, presented transport duration times that would result in speeds above 1000 km/h ( Figure 25).  The complete evaluation of each of the 83 PMFS would require further analysis by obtaining complete documentation and conducting on-site inspections, in order to confirm the irregularities and estimate the extension of the environmental damage. However, the results obtained provided an overview of the multiple possibilities and the importance of GEOINT for the assessment of PMFS. Table 3 shows the frequency and percentage of PMFS that presented each of the irregularities or signs of fraud evaluated, according to the criteria previously described in Table 2. Table 3. Frequency (f) and percentage (%) of irregularities/signs of fraud. Individual data and results for each of the 83 PMFS as images, maps, and graphs are available in the Supplementary Materials.

Discussion
The analysis of images and other spatial data allowed us to evaluate the current situation of forestry exploitation carried out in the 83 PMFS submitted to examination, identifying, among other points: (i) possible overlap with protected areas; (ii) where and when forest exploitation took place; (iii) what type and intensity of this exploitation (clear cut, selective cutting, opening of roads, yards etc.); (iv) its dimensions; and (v) the regularity of exploitation in relation to the authorized polygon and the areas of permanent preservation (APP).
In its turn, the analysis of non-spatial data (documents and databases obtained from public agencies) allowed for a better understanding of the entire context in which the obtained geographic information is inserted, notably: (i) if the referred exploitation took place in an authorized area and, if so, if it was carried out in accordance with the respective authorization, its conditions and limits; (ii) if the transport and sale of these forest products followed the respective legal procedure (there are no inconsistencies or evidence of fraud), with the wood being effectively delivered to the informed recipient; and (iii) or, if on the other hand such data, when analyzed together with the spatial information, point to the existence of evidence of fraud in the forest inventories (overestimation of volumes of higher value commercial species) as well as the existence of a transaction commercial and/or transport simulated in the respective systems, just to legalize forest products illegally extracted from other areas. Table 4 shows the frequency ranking by the criteria previously described in Table 2. With regard to the assessment of possible fraud in forest inventories, it should be noted that during the analysis work, it was possible to demonstrate that the volumetric data contained in the forest inventories (and, therefore the volume data of wood marketed through SisDOF) were expected to conform to the mathematical model known as Benford's law [51], which has long been used to detect economic and financial fraud.
The analysis, carried out using data from official forest inventories across the Southern Mesoregion of Amazonas, showed that this model can be applied to the following volumetric datasets: (i) total volume (by species or per hectare); (ii) number of specimens (total per species or per hectare); and (iii) sum of the diameter at breast height (DBH) values of all specimens, by species.
It was also found that the volumetric data traded in the State of Amazonas of Tabebuia serratifolia (ipê), a species that is known to have been fraudulently overestimated in forest inventories due to its high commercial value [12], were not in conformity with Benford's law. This observation reinforces the importance of the methodology used in this investigation and opens up a new range of possibilities for the realization of truly sustainable management in the Brazilian Amazon.
Nonetheless, we are aware that the analyses and results herein obtained have limitations, especially those related to the low spatial resolution of the satellite images used and the frequent presence of clouds in the region. Such limitations, however, refer exclusively to the option of using images available to the general public and not to specific limitations of the GEOINT methodology employed, which can be easily overcome with the acquisition of images with higher spatial resolution and those produced by radars.

Conclusions
Combined analysis of spatial and non-spatial data allowed for a better understanding of the context of each PMFS, especially as evidence of simulated transactions to legalize forest products extracted illegally from other areas. Low/medium resolution imagery can be used to identify selective logging infrastructure (access road and storage yards). Collateral material has paramount importance in understanding the imagery and geospatial information context as well as in reaching robust conclusions about the legality of forestry activities in a specific area. Business intelligence tools are valuable for analyzing and processing a large amount of collateral material, allowing for the results to be produced faster, as they discover hidden patterns in data. Benford's law is valuable for evaluating forestry inventory data and is useful for detecting possible fraud.