.

Narváez Figueroa, Cáceres Castellano & González Sanabria / INGE CUC, vol. 18 no. 2, pp. 47–60. January - June, 2023

Machine learning model for the classification of municipalities by illicit crops in Colombia from 2010 to 2020

Modelo de machine learning para la clasificación de municipios por cultivos ilicitos en Colombia de 2010 a 2020

DOI: http://doi.org/10.17981/ingecuc.19.1.2023.05

Scientific Research Article. Fecha de Recepción: 05/12/2022. Fecha de Aceptación: 14/12/2022.

Andrés Eduardo Narváez Figueroa

Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia)

andres.narvaezfigueroa@un.org

Gustavo Cáceres Castellano orcid_24x24

Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia)

gustavo.caceres@uptc.edu.co

Juan Sebastián González Sanabria orcid_24x24

Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia)

juansebastian.gonzalez@uptc.edu.co

.

To cite this paper

Narváez Figueroa, G. Cáceres Castellano & J. González Sanabria, “Machine learning model for the classification of municipalities by illicit crops in colombia from 2010 to 2020”, INGE CUC, vol. 18, no. 2, pp. 47–60, 2023. DOI: http://doi.org/10.17981/ingecuc.19.1.2023.05

.

Abstract

Introduction— The United Nations Office on Drugs and Crime (UNODC) classifies Colombia as one of the countries where drug trafficking and crime threaten the security, peace and development opportunities of its citizens.

Objective— This article presents the application of the unsupervised K-means classification algorithm to categorize municipalities with coca cultivation presence in Colombia. Methodology- The CRISP-DM methodology was used for data mining, and the PCA (Principal Component Analysis) algorithm was used for the correlation of variables.

Results— Multiple sources of information were used, such as: the number of hectares of coca per municipality, seizures, laboratories destroyed, manual eradication and fumigation, monitored by national institutions, in order to make crosses with socioeconomic and performance variables of the municipalities with coca crops in the period from 2010 to 2020. Based on the classification, the scenarios of each category were analyzed to find scenarios that allow elucidating the dynamics of the territories suffering from this scourge.

Conclusions— It was found that the behavior of coca-producing municipalities responds mainly to 4 groups. It was also found that the municipality of Tumaco in Nariño does not fit into any category since it exceeds the production with respect to the other municipalities.

Keywords— Unsupervised classification; illicit crops; data mining; fight against drugs; cocaine; Colombia

Resumen

Introducción­— La Oficina de las Naciones Unidas contra la Droga y el Delito (UNODC) clasifica a Colombia como uno de los países donde el narcotráfico y el delito ponen en riesgo la seguridad, la paz y las oportunidades de desarrollo de los ciudadanos.

Objetivo— Este artículo presenta la aplicación del algoritmo de clasificación no supervisado K-means para categorizar los municipios que tienen presencia de cultivos de coca en Colombia. Metodología: Se hizo uso de la metodología CRISP-DM, propia de la minería de datos, y para la correlación de variables se utilizó el algoritmo PCA (Análisis de Componentes Principales).

Resultados— Se utilizaron múltiples fuentes de información como: el número de hectáreas de coca por municipio, incautaciones, laboratorios destruidos, erradicación manual y fumigación, monitoreadas por la institucionalidad nacional, con el fin de realizar cruces con variables socioeconómicas y de desempeño de los municipios que tienen cultivos de coca en el periodo de 2010 a 2020. Partiendo de la clasificación, se analizaron los escenarios de cada categoría para hallar escenarios que permitan dilucidar las dinámicas de los territorios que sufren este flagelo.

Conclusiones— Se encontró que el comportamiento de los municipios productores de coca responde principalmente a 4 grupos. También se encuentra que el municipio de Tumaco en Nariño no encaja en ninguna categoría ya que excede la producción respecto a los demás municipios.

Palabras clave— Clasificación no supervisada; cultivos ilícitos; minería de datos; lucha contra la droga, Cocaína, Colombia

I. Introduction

Colombia did not begin as a coca growing power. It was initially a marimba (Colombian slang word for marihuana) growing power. In the 1960s, taking advantage of the hippie boom in the United States, marijuana crops appeared, mainly in the Serranía del Perijá and the Sierra Nevada de Santa Marta [1]. This bonanza did not last long due to the appearance of Californian marijuana in the narcotics market.

Between the 1970s and 1980s, the so-called coca bonanza phenomenon appeared, during which coca pasta base, a cocaine derivative, was being exported from Peru and Bolivia to Colombia. Then, it was processed and sent as cocaine to the United States. The high profits encouraged the growth of the business, allowing a rapid increase and expansion of production in far away regions such as those in the departments of Caquetá, Guaviare and Putumayo [2].

In the late 1980s and early 1990s drug trafficking cartels emerged as the main exponents in Cali, Medellín and Coastal areas of the country while implementing the entire coca production chain in Colombia, starting from the cultivation to the production of cocaine hydrochloride, including even money laundering.

The Colombian state response to the massive appearance of drug trafficking cartels [3], was the creation of the National Council of Narcotic Drugs (In Spanish, the CNE stands for the abbreviation of Consejo Nacional de Estupefacientes) through Law 30 of 1986 [4]. The CNE, among other things, regulates the areas where plants are cultivated in order to be later processed into drugs. Since 1987, the CNE has issued punitive decrees to control the processing of coca leaves. Law 30, in its article 7°, allows indigenous people to grow it for their own consumption, according to their cultural patterns, and establishes that the national government must promote crop substitution programs in areas where indigenous people and settlers have started to grow coca leaves for commercial purposes, before the Law was enacted [4].

At that time, the guerrillas used to exhibit an increasingly energetic presence in areas with oil, mining, illicit crops, border areas and important agricultural and livestock activity [5]. In many of these regions, large economic interests, whether or not linked to the world market, used to finance the proliferation of illegal security army groups and paramilitaries as a way to put an end to the guerrillas. The war increased the displacement of the rural civilian population, which has been historically affected by the territorial dispute between the different armed actors, as well as by the agrarian crisis. However, in Colombia, the drug problem and, within this, the problem of illicit crops, is relevant in the conflict discussion because it was once a financing source for both the insurgency and the development of paramilitary groups. Consequently, forced eradication actions have become part of the national security policy [5].

After more than two decades of implementing forced eradication strategies, the net area of coca cultivation has decreased from its peak by more than 50% [6]; but the area affected by coca cultivation has decreased by only 17% during the period from 2001 through 2010 [7]. Despite advances in surveillance, monitoring and even the implementation of manual eradication in many regions, as of 2010, 23 out of 32 departments reported coca plantations, reaching a total of 62,000 cultivated hectares of farmland [8]. Two decades of illicit crops attest that eradication as an isolated practice does not consolidate areas free of illicit crops [9].

Regarding the coca problem in Colombia, studies conducted by UCB (USA) correlate the presence of coca crops through a linear regression with the variables inherent to this problem, such as manual eradication and aerial spraying, to mention a few [6]. In addition, it also takes into account economic and social variables such as the number of human rights violations by illegal armed actors and the economic capacity of the vulnerable communities that suffer from this scourge in their territories. This research concludes that eradication was parallel to and even surpassed coca cultivation and, nevertheless, eradication has not managed to change Colombia’s status as a producer of almost half of the world’s coca leaf production [10], not to forget the investment made by the United States government in aerial spraying, and where the findings tell us that social investment, in addition to generating social welfare, emerges as a complementary response for controlling illicit crops.

In the study conducted by Uniandes (CO) propose a theoretical model to better understand the relationship between coca and the armed conflict in Colombia, based on the Ramsey dynamic optimization model [3]. The model depends on variables that are not easy to establish with accuracy, such as: the footprint of the illegal armed groups, territorial control of the groups, the salary earned by the portion of the population that works in coca production, the salary and cost of the military equipment used by the active guerrillas and the cost of maintaining control over the territory, all of which could possibly make the model difficult to implement.

They also look for a relationship between coca hectares and the behavior of these in neighboring towns with a coefficient of spatial correlation. To determine the presence of armed groups in the territories, they use a Boolean variable, where 1 shows if there are armed groups in a town and 0 if there aren’t any. This would show the action of the groups, but not their intensity [10].

Using GIS tools, UnManizales made a comparison of how coca cultivation coverage has changed in the municipalities of Cáceres and Tarazá in the department of Antioquia (Colombia), and how the landscape has been affected by coca cultivation [11].

Ujaveriana researchs compare the changes in forest coverage in the town of Tibú [12], as a result of illicit crops, in the period between 2000 and 2014 taken from the Colombian Environmental Information System (In Spanish, the SIAC stands for the abbreviation of Sistema de Información Ambiental de Colombia) and the extension of coca crops from the Integrated Illicit Crops Monitoring System (In Spanish, the SIMCI stands for the abbreviation of Sistema Integrado de Monitoreo de Cultivos Ilícitos), between 2005 and 2014 in order to build a timeline and find the relationship between these two problems and thus determine the environmental impact caused by these illegal crops.

In the two previous investigations [11], [12], it is found that GIS comparisons are made in different time periods, taking into account changes in the territories, but it is also found that other phenomena related to the dynamics of illicit crops are not included in the studies.

The studies mentioned above are mostly related to coca crops and the map analysis of illegal crops in Colombia using methods and tools from the field of systems engineering. Of course, there are many more studies referring to the problem of illicit crops and especially coca crops, both in Colombia and in Latin America, but mainly from an economic and social point of view.

UNED shows a repertoire of data mining techniques to be used in conjunction with GIS in order to deepen spatial analysis [13]. Within the repertoire, it is found that descriptive data mining techniques are the most appropriate for the proposed analysis, especially clustering, which, together with geovisualization, would be a powerful alternative to understand the dynamics of the illicit coca cultivation business.

Correspondingly, using data mining techniques, MARA carries out a project which is applied to agriculture in Argentina [14]. Using a combination of hierarchical clustering and the K-Means algorithm to obtain homogeneous groups of climate and soil, showing as a result that generalizations by region are not adequate, but by homogeneous environmental conditions. This research shows the possibility of performing data mining processes, and specifically the use of descriptive techniques such as clustering, specifically the K-Means algorithm in agriculture, showing the possibility of following this path in the proposed research.

The driving force of this research is to serve as a tool for decision makers to confront the problems generated by the drug market, starting with an initial descriptive analysis to look at the behavior of the variables and their distribution and a categorization of the towns where there is a presence of coca cultivation, based on the results of the exploratory analysis.

The objectives in terms of data mining are:

II. Methodology

This work was performed using the K-means algorithm, which refers to a clustering algorithm, where k refers to the number of centroids from which the grouping is performed. Also, the algorithm places each of the measurements within each of the k groups, taking into account the distance of each measurement to the centroid assigned to its group. Depending on the number of iterations that the algorithm repeats, the data subgroups are debugged (Fig. 1).

Fig. 1. Progress of K-means algorithm after several iterations.
Source: [15].

In order to observe the relationship between the variables, the PCA algorithm in R Studio PCA was used during the development. “PCA is one of the unsupervised learning techniques, which are usually applied as part of the exploratory analysis of the data” [16, par. 1]. In order to observe the correlation between variables, (the PCA) the Principal Component Analysis algorithm is used. The PCA is part of a multivariate analysis, allowing the analysis of high-dimensional data sets [16].

The CRISP-DM methodology was used specifically for data mining. It is divided into phases. In the first phase, understanding the problem, the objectives of the problem are determined, the situation is assessed including the resources, the objectives of the data mining are determined and the project plan is made. The objective of the applied data mining was to make an unsupervised classification of the towns in Colombia where coca crops are historically found. Based on this classification, scenarios were determined to evaluate which of the selected categories a town could belong to.

The databases of the coca leaf crop control monitoring system were considered for this research. The data on the monitoring of the presence of coca crops at a town level show the destruction of laboratories (primary production infrastructure and cocaine hydrochloride), manual eradication, seizures (coca leaf, coca paste and cocaine hydrochloride). Also, information regarding the market of this illegal economy, such as prices of coca leaf, coca paste and cocaine hydrochloride in pesos and dollars, as well as the average price of gold and the dollar as a source to keep track in case that the fluctuation of these two economic indicators involves any relation with the demand, the price and the number of cultivated coca leaf hectares in Colombia.

The second phase was called data comprehension. In this phase data were collected and the sources were explained focusing on the way they were coded while their quality was being verified. In order to carry out this task, a track of the official Colombian data infrastructure in the area of illicit crop was kept.

In terms of monitoring coca cultivation, the official data infrastructure in Colombia has two organizations that monitor the government’s fight against this illegal market. The first is the United Nations Office on Drugs and Crime (UNODC), which through the International Illicit Crop Monitoring System (SIMCI) tracks the census of coca cultivation in Colombia, and where coca prices throughout the production chain (coca leaf, coca paste and cocaine hydrochloride) can also be found. The second organization is the Colombian Drug Observatory (ODC), which has the Colombian Drug Information System (SIDCO) with the following databases of interest for the problem to be characterized: manual eradication, cocaine hydrochloride seizures, coca leaf seizures, cocaine base paste seizures, destruction of cocaine hydrochloride laboratories and destruction of primary production infrastructure laboratories.

It was also found that in order to categorize the towns, it is important to consider the variables that show the socioeconomic status of the towns, such as the Unsatisfied Basic Needs Index (UBN), the integral performance index and the category of the town.

The selected data include information related to coca cultivation crops as such, data on the institutional fight against the coca chain, economic indicators such as the price of the dollar and gold, socioeconomic information as well as a characterization of the towns. All information was disaggregated to a town level for the period between 2010 and 2020.

The third phase was called data preparation. In this phase, the data were processed and prepared for the application of data mining techniques. This phase was intended to be segmented in order to obtain the subsets, standardize and clean the data, calculate new data from existing data and format the data looking forward to facilitating their processing.

III. Results and Discussion

To start with the unsupervised classification process, the first thing to check is that the dataset is consistent in terms of the relationship between the variables that make it up.

This type of graphs are interpreted as follows “it indicates the % of variance explained by the first (Dim1) and second component (Dim2), positively correlated variables are grouped together or close together, while negatively correlated variables are plotted on opposite sides of the origin or opposite quadrants” [13, p. 1].

According to this, the first thing that can be observed is that the sum of the two axes Dim1 and Dim2 explain 50.13% of the total variance of the dataset, secondly, the behavior of the vectors corresponding to the variables that form the dataset shows some sort of stacking in their direction.

The stacking highlighted in red implies a relationship between the variables that describe the behavior of the crops and the dynamics concerning the illegal business such as hectares of coca, seizures, destruction of laboratories. The group categories highlighted in blue show a correlation between the variables that describe the behavior of the town and the variables corresponding to the integral performance index (Fig. 2). They are grouped in opposite parts to those that describe the BNI, which implies that the higher the performance index of the town, the lower the BNI.

Fig. 2. Correlation of variables PCA algorithm, first iteration.
Source: Own preparation.

There are 28 out of 92 variables that would explain 60% of the data and approximately 70 variables would be necessary to explain 99% of the data (Fig. 3). As can be seen in Fig. 5, the variables are related in general terms and the ones that have the greatest value in the axes in the PCA algorithm are those that refer to the hectares of coca per town, which indicates that the central variables in the dataset are the hectares of coca. This assures the consistency and relevance of the data for analyzing the problem of illicit coca cultivation in Colombia.

Fig. 3. Value of the variables according to PCA
Source: Own preparation.

It was decided to use 4 centroids to do the classification which will result in the classification of the data into four categories.

Having implemented the K-Means algorithm, the towns were grouped into four categories (Fig. 4). The algorithm classifies as follows: 15 towns are grouped in category 1; 229 are grouped in category 2; 24 are grouped in category 3; and in category 4 is only grouped 1 town out of the 269 towns with coca cultivation within the time period of this study.

Fig. 4. Group classification of towns using the K-Means algorithm.
Source: Own preparation.

Since the purpose of this type of algorithm is to group based on similar behavior among the variables that make up a group of data, finding a category composed of only one element is not ideal for classification projects.

Tumaco’s city performance mainly exceeds the performance of other cities or towns considering the number of hectares manually eradicated, seizures of cocaine hydrochloride and coca leaf, and a high number of cultivated hectares. As the city with the largest number of cultivated hectares, it was also the focus of efforts in manual eradication programs.

In order to obtain a more precise group classification, it was decided to iterate the implementation again, while temporarily excluding Tumaco. This was intended to observe the behavior of the rest of the towns and then add it to a category which best fits or consider it as a special case. In the second iteration, the PCA algorithm is re-evaluated to observe the correlation of variables in the dataset (Fig. 5).

Fig. 5. Correlation of variables PCA algorithm, second iteration.
Source: Own preparation.

The percentage of variance compared to that of the first iteration is lower, described by dimension one and two, while in the first iteration the sum of the two dimensions is close to 50%; in the second iteration it is close to 40%. In the case of the correlation of variables, they are observed to be closer, which implies a greater relationship between the variables, both those with a direct relationship and those with an opposite relationship.

There are 23 variables out of 92 that would explain 60% of the data and approximately 65 variables would be necessary to explain 99% of the data. A decrease in the number of variables explaining the behavior of the data can be observed, from 28 to 23 to explain 60% of the data, and from 70 to 65 to explain 99% of the data.

As shown in Fig. 6, the variables are related in general terms and the ones that have the greatest value in the axes in the PCA algorithm are those that refer to the hectares of coca per town, which indicates that the central variables in the dataset are the hectares of coca. Other variables that are important in the data are the destruction of primary production and cocaine hydrochloride laboratories. Having observed the correlation of variables, we proceeded to look for the appropriate number of centroids to develop the classification and then defined four centroids as the adequate number.

Fig. 6. Value of variables according to PCA second iteration.
Source: Own preparation.

The algorithm classifies this scenario as follows: 15 towns are grouped in category 1; 225 towns are grouped in category 2; 26 towns are grouped in category 3; and category 4 is made up of 2 towns (Fig. 7).

Fig. 7. Group classification of towns using the K-Means algorithm, second iteration.
Source: Own preparation.

As no categories formed by a single town were found, the results of the second iteration were taken as the final result of the classification and it was decided to classify the city of Tumaco for its particular behavior, as a special case, not belonging to any other category.

To analyze the results of the classification, we first analyzed the variables that define each category looking forward to understanding the dynamics that mark each group category (Table 1; Table 2; Table 3; Table 4).

Table 1.
Variables defining category 1 in the second iteration.

Variable

Test Value

Average Variable

Frequency

Global Average

Unsatisfied Basic Needs (UBN) Total

6.097

69.05

15

36.544

Unsatisfied Basic Needs (UBN) Rural

5.045

69.154

15

42.632

Category

–0.274

5.733

15

5.787

Cocaine hydrochloride seizures in 2015

–0.56

0

15

320.674

Coca leaf seizures in 2011

–0.571

14.8

15

3615.762

Cocaine hydrochloride seizures in 2010

–0.606

60.4

15

328.288

Cocaine hydrochloride seizures in 2018

–0.747

26.6

15

563.715

Cocaine hydrochloride seizures in 2016

–0.763

0

15

372.704

Cocaine hydrochloride seizures in 2014

–0.784

0

15

198.281

Cocaine hydrochloride seizures in 2011

–0.8

0

15

285.696

Coca Base Paste Seizures in 2014

–0.812

1.645

15

130.586

Cocaine hydrochloride seizures in 2012

–0.818

0

15

276.711

Cocaine hydrochloride laboratories in 2014

–0.847

0

15

0.433

Cocaine hydrochloride seizures in 2020

–0.855

0.007

15

493.673

Cocaine hydrochloride seizures in 2017

–0.892

0

15

597.163

Cocaine hydrochloride seizures in 2019

–0.939

0

15

384.31

Cocaine hydrochloride seizures in 2013

–1.004

0

15

270.053

Coca leaf seizures in 2010

–1.015

30.133

15

2953.333

Cocaine hydrochloride laboratories in 2010

–1.015

0

15

0.403

Cocaine hydrochloride laboratories in 2019

–1.018

0

15

0.94

Manual eradication in 2013

–1.063

3.953

15

68.147

Manual eradication in 2010

–1.073

7.867

15

102.971

Cocaine hydrochloride laboratories in 2012

–1.095

0

15

0.679
Source: Own preparation.
Table 2.
Variables defining category 2 in the second iteration.

Variable

Test Value

Average Variable

Frequency

Global Average

Integrated performance index in 2019

8.512

51.692

225

48.281

Integrated performance index in 2020

8.093

47.868

225

44.8

Integrated performance index in 2018

8.09

46.7

225

43.772

Integrated performance index in 2016

7.972

44.12

225

41.345

Integrated performance index in 2017

7.773

44.502

225

41.785

Integrated performance index in 2011

7.566

57.471

225

53.632

Integrated performance index in 2014

7.394

67.777

225

63.955

Integrated performance index in 2013

7.36

64.901

225

61.072

Integrated performance index in 2012

7.35

60.201

225

56.493

Integrated performance index in 2015

6.904

68.47

225

64.814

Integrated performance index in 2010

6.779

58.84

225

55.279

NBI Municipal capital

1.877

25.962

225

24.858

Cocaine hydrochloride laboratories in 2011

0.451

0.342

225

0.328

Category

0.393

5.796

225

5.787

Cocaine hydrochloride laboratories in 2010

–0.281

0.391

225

0.403

Cocaine hydrochloride seizures in 2013

–0.556

254.122

225

270.053

Cocaine hydrochloride seizures in 2011

–0.764

256.632

225

285.696

Coca Base Paste Seizures in 2014

–0.766

117.631

225

130.586

Coca leaf seizures in 2011

–1.003

2942.002

225

3615.762

Cocaine hydrochloride seizures in 2010

–1.066

278.15

225

328.288

Cocaine hydrochloride seizures in 2012

–1.164

234.773

225

276.711

Cocaine hydrochloride seizures in 2016

–1.219

309.288

225

372.704

Cocaine hydrochloride seizures in 2014

–1.672

153.258

225

198.281
Source: Own preparation.
Table 3.
Variables that define category 3 in the second iteration

Variable

Test value

Average Variable

Frequency

Global Average

Coca 2012

10.545

769.5

26

159.414

Coca 2013

10.327

803.692

26

155.138

Coca 2011

10.138

1100.692

26

217.06

Coca 2015

9.588

1565.104

26

295.24

Laboratories for primary production in 2015

9.576

59.5

26

12.31

Coca 2014

9.442

1137.692

26

224.511

Coca 2010

9.042

956.808

26

210.25

Coca leaf seizures in 2015

8.8

13827.39

26

2653.722

Coca 2017

8.513

2759.607

26

567.082

Coca leaf seizures in 2019

8.473

7208.682

26

1285.028

Coca 2018

8.389

2316.045

26

458.926

Coca 2016

8.219

2809.635

26

570.789

Coca leaf seizures in 2014

8.137

10284.76

26

1837.224

Laboratories for primary production in 2012

8.074

29.192

26

7.112

Laboratories for primary production in 2019

7.99

84.308

26

16.03

Coca leaf seizures in 2017

7.611

9815.185

26

1991.945

Coca leaf seizures in 2016

7.523

17731.408

26

3328.278

Coca leaf seizures in 2020

7.455

8817.058

26

1718.989

Manual Eradication in 2017

7.349

755.973

26

133.626

Laboratories for primary production in 2010

7.218

38.385

26

8.041

Coca Base Paste Seizures in 2018

7.186

617.083

26

123.93

Coca leaf seizures in 2012

6.964

6920.806

26

1478.528

Coca 2019

6.938

2611.581

26

532.259
Source: Own preparation.
Table 4.
Variables that define category 4 in the second iteration.

Variable

Test Value

Average Variable

Frequency

Global Average

Cocaine hydrochloride laboratories in 2019

13.132

35

2

0.94

Laboratories for primary production in 2017

12.417

335

2

13.194

Cocaine hydrochloride laboratories in 2017

12.045

26

2

0.896

Laboratories for primary production in 2018

11.886

356

2

14.679

Coca 2016

11.546

10120.41

2

458.926

Coca Base Paste Seizures in 2019

11.423

4565.017

2

170.705

Coca 2017

11.409

11675.145

2

567.082

Coca 2019

11.315

13351.225

2

532.259

Laboratories for primary production in 2020

11.301

379

2

16.672

Coca 2020

11.1

12517.715

2

499.817

Coca 2018

10.98

11877.31

2

570.789

Cocaine hydrochloride laboratories in 2012

10.237

18.5

2

0.679

Coca Base Paste Seizures in 2019

10.034

5164.034

2

219.591

Coca Base Paste Seizures in 2019

9.88

4000.757

2

162.148

Coca 2015

9.828

5215.555

2

295.24

Cocaine hydrochloride laboratories in 2018

9.79

21

2

0.963

Manual Eradication in 2020

9.737

8680.32

2

435.371

Coca leaf seizures 2018

9.707

35042.338

2

1569.84

Coca 2014

9.553

3717

2

224.511

Coca Base Paste Seizures in 2019

9.309

2538.96

2

123.93

Laboratories for primary production in 2014

9.183

118

2

7.299

Laboratories for primary production in 2016

9.158

242.5

2

15.381

Laboratories for primary production in 2019

8.59

293.5

2

16.03
Source: Own preparation.

The first category which is made up of 15 towns highlights most of the values for the total UBN and the rural UBN, which stands for unsatisfied basic needs. For DANE [17, par. 1] “the UBN seeks to determine, with the help of some simple indicators, whether the population’s basic needs are covered. Groups that do not reach a fixed minimum threshold are classified as poor”. This implies that they are economically and socially vulnerable towns with a small number of coca leaf, coca paste and cocaine hydrochloride seizures. In many situations, these seizures are found to be zero.

The second category which is made up of 225 towns, remarks heavily the values in the integral performance index, which, according to the Dictionary of Public Administration [18, par. 1] is “The Integral Performance Index (IDI), seeks to evaluate public management (in its programming, execution and follow-up stages) and decision-making in the use of municipal resources”. Another variable involved in the description of category 2 is the NBI for municipal capitals.

The third category, made up of 26 towns, shows that coca cultivation is the most important factor, however, the destruction of primary production laboratories and coca leaf seizures also define this category.

The fourth category, made up of two towns, refers to a heavy presence of cocaine hydrochloride laboratories and primary production. It is also important to take into account coca cultivation, which, unlike the third category, is much higher in this one (Fig. 8).

Fig. 8. Classification of coca-growing towns 2010-2020
Source: Own preparation

IV. Conclusions

References

[1] E. Garzón, A. Lopez, A. Reyes Posada, R. Rocha y S. Uribe, Drogas ilícitas en Colombia: su impacto económico, político y social. BO, CO: Ariel, 1997.

[2] E. Ciro, “Cultivando coca en el Caquetá: vidas y legitimidades en la actividad cocalera”, Tesis doctora­l, FCPyS, UNAM, CDMX, MX, 2016. Disponible en https://ru.dgb.unam.mx/handle/DGB_UNAM/TES01000751075

[3] A. Díaz y F. Sánchez, “Geografía de los cultivos ilícitos y conflicto armado en Colombia”, Trabajo grado, Fac Econ, UNIANDES, BO, CO, 2004. Disponible en http://hdl.handle.net/1992/7865

[4] M. Ramírez, Entre el Estado y la guerrilla: identidad y ciudadanía en el movimiento de los campesinos cocaleros del Putumayo. BO, CO: ICANH, 2001. Disponible en https://babel.banrepcultural.org/digital/collection/p17054coll10/id/2898/

[5] P. Sacipa, “Desplazamiento forzado y política de erradicación de cultivos ilícitos,” Scripta Nova, vol. 94, no. 39, Ago. 2001. Disponible en https://revistes.ub.edu/index.php/ScriptaNova/article/view/364

[6] E. Davalos, “New answers to an old problem: Social investment and coca crops in Colombia,” Int. J. Drug Policy, vol. 31, pp. 121–130. May 2016. https://doi.org/10.1016/j.drugpo.2016.02.002

[7] UNODC, “Informe Mundial sobre las Drogas 2014”, UN, NYC, NY, USA, V.14-04627 (S), 2014. Available: https://www.unodc.org/wdr2014/

[8] UNODC, Informe Mundial sobre las Drogas 2011. NYC, NY, USA: UN, 2011. Available: https://www.unodc.org/unodc/en/data-and-analysis/WDR-2011.html

[9] G. Buzai, C. Baxendale, N. Principi, M. Cruz, G. Cacace, N. Caloni, L. Humucata y J. Mora, Sistemas de Información Geográfica (SIG): teoría y aplicación. LUJ, AR: UNLU, 2013.

[10] UNODC, Informe Mundial sobre las Drogas 2013. NYC, NY, USA: UN, 2013. Available: https://www.unodc.org/lpo-brazil/en/drogas/relatorio-mundial-sobre-drogas.html

[11] D. Ortiz, “Influencia de los cultivos ilícitos como dinámica territorial de los municipios de Cáceres y Tazará (Antioquia) en los períodos 2007 – 2010, mediante herramientas de sistemas de información geográfica – SIG,” Trabajo Final, Fac Cienc Ing, UManizales, MZL, CO, 2018. Disponible en https://ridum.umanizales.edu.co/xmlui/handle/20.500.12746/4174

[12] L. Camargo, “Lineamientos para la gestión de la deforestación, generada por los cultivos ilícitos asociados al conflicto armado, en el Municipio de Tibú, en el contexto del posconflicto”, Trabajo de Grado, FEAR, UJaveriana, BO, CO, 2022. https://doi.org/10.11144/Javeriana.10554.38071

[13] J. Gil, Minería de texto con R. Aplicaciones y técnicas estadísticas de apoyo. MA, ES: UNED, 2021.

[14] MARA. “Las tres ideas de Sarmiento sobre la minería”. Proyecto Mara. Disponible en https://www.infomara.com.ar/ (consultado 2010 Sep. 28).

[15] C. Gil. “Análisis de datos scRNA-Seq con bioconductor”. Rpubs, 28 Feb. 2020. Disponible en https://rpubs.com/Cristina_Gil/scRNS-Seq_Bioconductor

[16] J. Hair, R. Anderson, R. Tatham & W. Black, Multivariate Data Analysis, 6th ed. TTN, NJ, USA: Prentice Hall, 2006.

[17] DANE, “Información Censo nacional de población y vivienda 2018”. Gov.co, 2018. Disponible en https://www.dane.gov.co/index.php/estadisticas-por-tema/pobreza-y-condiciones-de-vida/necesidades-basicas-insatisfechas-nbi#:~:text=La%20metodolog%C3%ADa%20de%20NBI%20busca,fijado%2C%20son%20clasificados%20como%20pobres

[18] Gov.co, “Índice de Desempeño Integral (IDI)”. Servicio al ciudadano-Glosario, 2022. Disponible en https://www.funcionpublica.gov.co/glosario/-/wiki/Glosario+2/%C3%8Dndice+de+Desempe%C3%B1o+Integral+%3COPEN_PARENTHESIS%3EIDI%3CCLOSE_PARENTHESIS%3E

Andrés Eduardo Narváez Figueroa. Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia).

Gustavo Cáceres Castellano. Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia). https://orcid.org/0000-0001-9621-3585

Juan Sebastián González Sanabria. Universidad Pedagógica y Tecnológica de Colombia. Tunja (Colombia). https://orcid.org/0000-0002-1024-6077

.
© The author; licensee Universidad de la Costa - CUC.
INGE CUC vol. 19 no. 1, pp. 47-60. January - June, 2023
Barranquilla. ISSN 0122-6517 Impreso, ISSN 2382-4700 Online
.