Introduction

Border control is a crucial component of national security (U.S. General Accounting Office, 1999; Weber and Bowling, 2004). For a country the size of the United States, however, the 8,000-mile length of its land borders and coastlines poses a significant security challenge. Illegal migrant workers, drug couriers, foreign terrorists, smugglers, fugitives, and other criminals benefit from the geographic range and porous nature of both southern and northern American borders (Turner, 2004).

This paper explores the geographic patterns associated with illegal border crossings in order to increase the effectiveness of border security operations. The U.S. Border Patrol is responsible for securing the nation's land and water borders between official ports of entry (U.S. Department of Homeland Security, 2004, 2005). Their mission is to prevent illegal entry and to interdict and apprehend illegal entrants. The Border Patrol has employed personnel, technology, and intelligence in the effort to control illegal entry of people and contraband (Vila, 2003; Maril, 2004). Methods used at legitimate border crossing points include canine units, electronic security access gates, mobile X-ray scans of commercial vehicles, and Vehicle and Contraband Inspection Systems (VACIS). At active illegal border crossing locations, strategies involve vehicle, boat, and helicopter patrols, ground sensors, bollard fencing, vehicle barriers, and unmanned aerial vehicles (UAV).

Illegal border crossers, however, quickly adapt to such efforts, displacing to areas less well surveilled and patrolled (Kossoudji, 1992; Wilson, 2002). U.S. Border Patrol initiatives along the San Diego–Tijuana perimeter, for instance, helped curtail illegal migration into California, only to have it increase in Arizona (Anderson, 2003; Clarke and Guerette, 2005). It is difficult to effectively protect a long border, even if resources are unlimited. Despite a 12-foot high wall, 302 watchtowers, and a shoot-to-kill policy, someone escaped over the 27-mile long Berlin Wall every 2 days on average. A Rand study modeling drug interdiction efforts demonstrated that a high percentage of the border/coastline would have to be covered to have any significant impact on narcotic importation levels (Crawford et al., 1988).

The purpose of our research was to identify facilitating and inhibiting factors of illegal land border crossings in order to determine physical and human geographic features related to the probability of such movement. A geographic information system (GIS) can be used to analyze the relationship between these factors and the volume of illegal border crossings. Identifying the geographic patterns of illegal migration can help the Border Patrol optimize resource allocation and anticipate offender reactions (see also Ackleson, 2003).

Theory

An individual attempting an illegal land border crossing must access the crossing location, evade detection efforts, and escape. The criminal behavior of illegal border crossers, like that of other criminals, is not haphazard. Instead, it is subject to the rationality shaped by the individual's perceptions, beliefs, knowledge, and personal experiences. An individual's reactions to preventative measures and the availability of resources such as time and money also play important roles. A rational choice (cost/benefit or risk/reward) analysis assesses, summarizes, and compares the positive and negative impacts of competing alternative approaches in order to ensure effective decision-making (Cornish and Clarke, 1986; Clarke and Felson, 1993).

Traditional migration theory (Haynes and Fotheringham, 1984; Clark, 1986) proposes that conditions at the place of origin (e.g., poverty, unemployment, famine, war, etc.) repel or push people away from that location to other locations that present an attraction or pull (e.g., favorable immigration laws, better climate, high standard of living, job opportunities, etc.). Distance factors such as effort, time, and money (the expenditure of which increases with distance) influences migration decisions, as most migrants tend to travel as short a distance as possible in order to achieve their objective. The "beaten path effect" attempts to explain migration behavior with the concept that the well-trodden path is an obvious route for those following, and followers tend not to stray from that route. Migration theory provides insight to those factors most likely to influence high volumes of illegal border crossing activity.

Geographic profiling analyzes crime locations to determine the most probable area of offender residence (Rossmo, 2000). Criminals prefer to operate in familiar places, excepting areas too close to home (Brantingham and Brantingham, 1981, 1984). Geographic profiling, informed by the environmental criminology perspective, incorporates the principles of least effort, mental maps, routine activities, and rational choice. The locations of criminal activity are geocoded and manipulated to create a jeopardy surface, a three-dimensional probability surface resembling a topographical map with color ramped peaks and valleys indicating likely base locations for the offender. The principles underlying geographic profiling can also be used to help identify and analyze illegal immigration patterns.

Methodology

This research uses GIS to combine illegal entry spatial data with border region information layers in order to facilitate exploration of their relationship (Pick et al., 2001). A GIS is comprised of computer software, hardware, data, and personnel used to integrate, manipulate, analyze, and present a wide variety of information tied to specific spatial locations on the surface of the Earth (Fotheringham and Rogerson, 1994). Spatial data include geocoded (positioned on the surface of the Earth) vector and raster shapefile formats. Vector shapefiles use points, lines, and polygons to represent features, whereas raster shapefiles represent the surface of the Earth as a grid, with each grid cell having one and only one value. A variety of non-spatial types of information, such as images and databases, can also be integrated into a GIS. For analysis procedures, GIS data must be in the same file format (vector or raster).

Data

The dependent variable for this research is the location of illegal land border crossings across the Texas–Mexico border (excluding legitimate port of entry fraudulent immigration). The independent variables can be classified into physical and human geographic features. Physical geography considered in this analysis included hydrography, vegetation, natural region, soil, temperature, precipitation, and terrain. Human geography included major transportation networks (roads, railroads, and bridges), legitimate border crossing facilities, urban development, and population density.

Border Patrol data

The Border Patrol groups the 50 states and the island of Puerto Rico into 21 sectors, five of which cover Texas: El Paso,¹ Marfa, Del Rio, Laredo, and McAllen (renamed Rio Grande Valley since the start of this project). The research focused on the Del Rio Sector, which covers 205 miles of the Rio Grande along the border between Texas and Mexico. The DRS reaches 300 miles north of the border, encompassing 59,541 square miles of primarily farm and ranch land. While this includes 41 counties, only three – Val Verde, Kinney, and Maverick – are border counties. During the time period covered by our data, Del Rio had 8.3 per cent of the U.S., and 26.2 per cent of the Texas and New Mexico, illegal alien traffic.

For the DRS-dependent variable, a GIS point shapefile of 111 DRS Landmark Mile Markers (which are not placed at one-mile intervals, despite what the name suggests), following the Rio Grande along the Texas–Mexico border, was obtained from the U.S. Border Patrol, Department of Homeland Security (DHS). The Border Patrol Del Rio Sector enters their apprehension information into a data collection system called ENFORCE. From this system, we were provided a database, comprised of 52 separate variables, with records for 254,717 illegal entry apprehensions for the period October 1, 2000, to August 17, 2004 (almost four fiscal years²). A field in the database identifies the Landmark Mile Marker point closest to the location of the illegal entry apprehension.

Another field denotes the final disposition of the illegal border crosser. While most individuals who illegally cross the Texas–Mexico border are entering the United States for work purposes, a small number are engaged in criminal activity. A criminal disposition in the final disposition field indicates narcotics smuggling or other criminal activity. We analyzed all illegal entries and then criminal disposition entries specifically.

Unfortunately, the location identifiers are not georeferenced in ENFORCE. While the distance and direction from the apprehension event to the LMM and the name of the LMM are listed in the database, the exact location (e.g., latitude and longitude, UTM coordinates, etc.) of the LMM is not. Therefore, without additional information, it is not possible to place the event on a map.

DRS Border Patrol agents traveled the Rio Grande by boat with a global positioning system (GPS) and geocoded all their LMMs. We therefore were able to map and analyze all the data for the Del Rio Sector (a total of 254,717 events). While we had originally hoped to analyze the entire Texas–Mexico border, we were unable to obtain equivalent GIS point shapefiles for the LMMs in the other four Texas Border Patrol Sectors.

The Del Rio data was "cleaned" to improved geocoding accuracy. Initial geocoding had a high success rate and manual rematching of valid border entry points further increased accuracy (e.g., from 94.1 to 96.3 per cent, for 2001). This was accomplished through expanding the DHS geocoding reference by adding extra mile markers from the Border Apprehension Dataset. Some entry locations were eliminated as they represented cases in which an illegal border crosser was apprehended in a sector different from the one they initially entered.

The multiple years of data permitted us to analyze changes over time. The number of apprehensions (we worked with the entire official data population, not a sample), and the volume and range of geographic data for even one sector, provided us with the ability to do certain analyses and explore relationships of interest. Finally, the number of independent variables and the numerous necessary spatial data joins made working within even a single sector fairly demanding. The database of 254,717 entry points, with 52 fields, and associated independent variable values, was approximately 570,000 kB in size; every additional spatial join to new data took progressively, and significantly, longer (a spatial join attaches the attributes of one geographic layer to another so that the relationships between features in these layers can be determined).

Geospatial data

A variety of United States independent physical geography variable data is available for the Del Rio Sector. A classed Rivers and Streams GIS line shapefile was obtained from the Texas Department of Transportation (TxDOT), and a Lake and Reservoir GIS polygon shapefile from the United States Census Bureau (USCB) Topologically Integrated Geographic Encoding and Referencing System (TIGER). The International Boundary Water Commission (IBWC) has established a Rio Grande Gauging Station database, which includes the monthly minimum, mean, and maximum flow averages, and the latitude and longitude for individual gauging stations. The Texas Parks and Wildlife Department (TPWD) provided Vegetation and Natural Region GIS Polygon shapefiles. A soil polygon shapefile was available from the United States Department of Agriculture (USDA).

An assortment of United States independent human geography variable data exist for the DRS. For major transportation features or networks, TxDOT provided GIS line shapefile (mostly classed) data for Roads, Railroads, and Border Bridges, whereas the USCB provided (unclassed but named feature) GIS line shapefiles for Roads and Railroads. The Binational Border Transportation (BNBT) provided a Highway and Rail Texas and Mexico Border Crossing Facilities GIS point shapefile. The USCB provided a 2000 Census Block Demographic database and a Census Block GIS polygon shapefile, as well as an Urban (Development) GIS polygon shapefile.

Finally, the Texas Environmental Quality Commission (TEQC) provided Digital Orthophoto Quarter Quads (DOQQ) raster file data. Aerial photographs taken by the National Aerial Photography Program (NAPP) provided the base data for the creation of DOQQs. The DOQQs have a cell resolution of one meter and there is a slight overlap of coverage area into Mexico. The DOQQs provide a readily accessible look at specific areas of interest.

Data for the Mexico side of the DRS border were more difficult to obtain, and what we did find is of lower resolution and quality than United States data. GeoCommunity (GEOCOMM) provided a Water Bodies (rivers, streams, lakes, and reservoirs) GIS line shapefile; however, the features are sparse in comparison to the Texas water features. The Instituto Nacional De Estadistica Geografia E Informatica (INEGI) provided a Rivers (non-spatial) BMP Image, which is much more detailed, and displays feature classes. In addition, INEGI also provided Principales Tipos De Vegetatcion (vegetation), Fisiographia (natural region), and Geologia (type of rock – igneous, sedimentary, etc.) non-spatial PDF images, and an Edafologia (soils) non-spatial BMP image.

GEOCOMM provided Road and Railroad GIS line shapefiles which, while showing relatively decent major feature networks in the DRS (in comparison to other Mexico data sets), have much less detail than available United States data. The Center for International Earth Science Information Network (CIESIN) provided Municipio (county) and State GIS polygon shapefiles with 1990 Mexico Census demographic data originally compiled by INEGI. In addition, CIESIN also provided a Population Density (by square kilometer) raster derived from the 1990 Mexico Census Localities (cities and towns) point and State polygon demographic data originally compiled by INEGI. Finally, GEOCOMM provided a Populated Places polygon shapefile. The data dictionaries for these information sets sometimes contain important feature definition omissions. We used a Spanish-speaking researcher to contact and liaise with the Help Department of each individual INEGI regional office in order to obtain the relevant GIS layers.

Some of these variables are useful for predicting geographic variations in illegal immigration (e.g., proximity to urban areas, roads, bridges, etc.), whereas others are useful for predicting temporal variations in illegal immigration (e.g., temperature, Rio Grande water flow volumes, etc.). This article only reports on the findings of the geographic component of the predictive analysis.

Analysis

Data preparation

The preparation of the United States data for use in our analysis involved a variety of GIS procedures. Geocoding is the process of converting non-spatial data into spatial data within a GIS. The Illegal Entry Apprehensions (DRSBP) database was geocoded using the DRS Landmark Mile Marker (DHS) GIS point shapefile as a reference. The Rio Grande Gauging Station (IBWC), Texas Weather Station Historic Temperature, and the Texas Weather Station Historic Precipitation (NCDC) databases were geocoded into GIS point shapefiles using the latitude and longitude coordinates provided.

As vegetation features tend to be uniform along the Rio Grande (Ceniza Black Brush and Creosote Bush Brush or Crops), it was decided that identification of vegetation at increasing distances from the border would be more useful than merely identifying vegetation at the Landmark Mile Marker. A three-ring (5 km per ring) mask was placed around the Rio Grande, and each of the rings was used in turn to clip the Vegetation (TPWD) GIS polygon shapefile into three vegetation GIS polygon shapefiles for use in later analyses. A major highway runs parallel to, and a short distance from, the border. As it is well patrolled, and thus usually avoided, it was clipped from the roads network. The remaining roads were then used in the analysis.

Exploratory spatial analysis

Exploratory analysis of the data set created from the spatial joins involved creating summaries of all illegal entries and criminal disposition entries for various variables. Additional exploratory spatial analysis (ESA) was based on the results of a variety of statistical tests (Giovando and Zhang, 2005). Two correlation matrixes were constructed, one with all illegal entries grouped by landmark mile markers, and the other with criminal disposition entries specifically, grouped by landmark mile markers as the dependent variables.

Results and discussion

Individual characteristics

Several individual characteristics of apprehended illegal border crossers are captured by the U.S. Border Patrol in their ENFORCE database. The results of relevant offender attributes are discussed in this section.³

Almost all individuals apprehended by the U.S. Border Patrol in the Del Rio Sector were classified as participants (96 per cent of all illegal entrants and 96 per cent of criminal entrants). Participants were those individuals who were involved in the event (an illegal border crossing), but were not a smuggling organizer or perpetrator of a more serious crime. Principals were those individuals charged with a crime (e.g., smuggling contraband). Associates were those individuals involved in the crime but at a level lower than the principal (e.g., as a facilitator or organizer). Suspects were those individuals believed to be involved in criminal activity, but were not necessarily held in custody. Other is a catch-all category for anything not included in the other four event roles. Table 1 presents the breakdowns by event role for all illegal entries and criminal disposition entries. Almost all illegal border crossers gave travel or seeking employment as their reason for entry. Table 2 presents the breakdowns by declared entry reason for all illegal entries and criminal disposition entries.

Table 1 - Event role (all illegal entries and criminal disposition entries) – Del Rio Sector, Texas.

Full table

Table 2 - Illegal border crossers stated reason for entry (all illegal entries and criminal disposition entries) – Del Rio Sector, Texas.

Full table

Most illegal border crossers in the Del Rio Sector are male (90.9 per cent), adult (89.7 per cent), and single (64.6 per cent). Table 3 presents the breakdowns by personal characteristic for all illegal entries and criminal disposition entries. Figure 1 shows the age distributions for all illegal entries and criminal disposition entries. The criminal disposition entrants appear slightly older. The distributions are relatively similar, except for a peak around the 30-year age for criminal disposition entrants. The mean, median, and modal ages for all illegal entrants are 26.9, 25, and 18 years, respectively. For criminal disposition entries, the ages are 28.5, 27, and 30 years, respectively.

Figure 1.

Age at apprehension

Full figure and legend (78K)

Table 3 - Personal characteristics (all illegal entries and criminal disposition entries) – Del Rio Sector, Texas.

Full table

The vast majority of apprehended illegal border crossers in the Del Rio Sector possess Mexican citizenship (92.8 per cent). Most of the remainder is from other Central American or South American countries. The citizenship breakdown for criminal disposition entries is similar, with Mexico (85.3 per cent) again leading the list. Second place, however, goes to illegal border crossers claiming U.S. citizenship (8.5 per cent). The countries of citizenship most often claimed for both all illegal entrants and criminal disposition entrants are presented in Table 4. The U.S. Border Patrol also records information on country of birth and country of residence, but these data are very similar to the citizenship figures.

Table 4 - Citizenship (all illegal entries and criminal disposition entries) – Del Rio Sector, Texas.

Full table

Temporal patterns

The Border Patrol ENFORCE data were examined for temporal characteristics and patterns. The busiest months for all illegal entries are January to May (see Figure 2). March is the most active month, and December the least active. The monthly breakdown for criminal disposition entries is more varied, and appears to follow a rhythmic cycle (See Figure 3). January, May, and September are the busiest months, whereas October to December is the least busy.

Figure 2.

Entries by month – all illegal entries.

Full figure and legend (68K)

Figure 3.

Entries by month – criminal disposition entries.

Full figure and legend (71K)

There is little variation by day of month for all illegal entries. There is a slight drop off at the end of the month, but this is partly the result of the fact that the 12 months vary slightly in length. Again, the criminal disposition entries are more varied, with highs and lows appearing to follow a cycle of some sort.

Most of the illegal border crossing activity occurs from Thursday to Sunday (see Figure 4). Saturday is the busiest day of the week, whereas Monday is the least busy. The pattern is more varied and cyclical for criminal disposition entries (See Figure 5). Wednesday is the busiest day, whereas Sunday is the least busy.

Figure 4.

Entries by day of week – all illegal entries.

Full figure and legend (98K)

Figure 5.

Entries by day of week – criminal disposition entries.

Full figure and legend (93K)

There are two peak activity times of day for all illegal entries, the first from 10:00 am to 2:00 pm, and the second, from 8:00 pm to 10:00 pm (see Figure 6). There is less activity during the late afternoon to early evening period, from 3:00 pm to 7:00 pm, and little activity occurs in the early morning hours, from 2:00 am to 6:00 am. Criminal disposition entries again have a more varied pattern. The busiest time period is around 1:00 pm, and the quietest around 3:00 pm (see Figure 7). In contrast to all illegal entries, there is comparable less activity in the evening, and more in the early morning.

Figure 6.

Time of entry – all illegal entries

Full figure and legend (141K)

Figure 7.

Time of entry – criminal disposition entries.

Full figure and legend (165K)

Change over time in illegal border crossing activity was measured in two ways: (1) variability and (2) growth. Variability was calculated by first determining the mean number of illegal entries over the four fiscal years (2001–2004) for each landmark mile marker. The difference between the number of entries in a given year and the four-year mean was divided by the mean. The standard deviation was then calculated for the four years of data. Several LMMs with at least 100 entries⁴ had a standard deviation equal to or greater than 0.50. Substantial variability occurred in the urban areas of Del Rio and Eagle Pass, and in the rural areas to the northwest of both communities.

Growth was calculated by first normalizing the data by dividing the annual number of entries for each landmark mile marker by the total number of illegal crossings in the Del Rio Sector for that year, and multiplying by 100,000. The proportional change from FY 2001 to FY 2004 was then calculated. Several LMMs with at least 100 entries showed significant change in activity (defined as an increase or decrease of 50 per cent or greater). The LMMs on the northwestern and southeastern edges of the Del Rio Sector experienced significant increases in illegal border crossing activity, as did the area immediately outside the city of Del Rio. The LMMS with significant decreases in activity were primarily in the Del Rio and Eagle Pass urban areas.

Criminal disposition entries were not separately analyzed as there were too few cases for meaningful results (the mean annual number of criminal disposition entries per LMM was only 1.32).

Spatial patterns

The Border Patrol ENFORCE data were examined for various spatial characteristics and patterns. Figures 8 and 9 show, respectively, the number of all illegal entries and the number of criminal disposition entries by landmark mile marker, arranged in north to south order. Some areas are clearly much more active than others. Figure 10 presents the same information (all illegal entries), but in order of activity level.

Figure 8.

All illegal entries by landmark mile marker (October 1, 2000–August 17, 2004).

Full figure and legend (82K)

Figure 9.

Criminal disposition entries by landmark mile marker (October 1, 2000–August 17, 2004).

Full figure and legend (71K)

Figure 10.

All illegal entries by landmark mile marker in order of activity.

Full figure and legend (41K)

The degree of spatial clustering in illegal border crossing activity can be determined by comparing the distribution of all illegal entries by landmark mile marker to that expected by chance. Plotting actual entries against a uniform distribution (i.e., what is expected by chance) produces a Lorenz curve (see Figure 11), to which an index of dissimilarity or concentration can be applied. One such measure is the Gini coefficient (Taylor, 1977). In this case, it is equal to:

where G is the Gini coefficient; N the total number of observations (111); x_n the nth member of the uniform percentage frequency; and y_n the nth member of the illegal entry percentage frequency.

Figure 11.

All illegal entries border crossings Lorenz curve.

Full figure and legend (116K)

The Gini coefficient ranges from 0 to 1, with 0 indicating exact correspondence between the two sets of percentage frequencies and 1 indicating a complete lack of correspondence. The closer the Gini coefficient is to 1, the higher the degree of clustering. The Gini coefficient for the all illegal entries data was 0.64. An analogous measure, the Index of Dissimilarity, has a value of 0.49. Similar results are found when the data are analyzed annually – the Gini coefficient ranges from 0.64 (2002) to 0.68 (2004), and the Index of Dissimilarity from 0.48 (2002) to 0.52 (2004). Criminal disposition entries in the Del Rio Sector were even more clustered, with a Gini coefficient of 0.74 and an Index of Dissimilarity of 0.58.

Figure 12 shows a thematic map, using graduated colored circles to represent all illegal entry activity level by landmark mile marker for both the entire Del Rio Sector, and for the more active southern region (inset). Figure 13 shows the comparable activity level for criminal disposition entries only. For both all illegal entries and criminal disposition entries, the Del Rio-Acuña and Eagle Pass-Piedras Negras urban areas, southern Maverick County, and the border between Kinney and Maverick counties, were the busiest areas. Figures 14 and 15, respectively, show all illegal entries and criminal disposition entries as a bar graph, merged with the topographic landscape and political boundaries of the Del Rio Sector.

Figure 12.

All illegal entries.

Full figure and legend (194K)

Figure 13.

Criminal disposition entries.

Full figure and legend (190K)

Figure 14.

Illegal border crossings and physical landscape – all illegal entries.

Full figure and legend (126K)

Figure 15.

Illegal border crossings and physical landscape – criminal disposition entries.

Full figure and legend (117K)

Point data information, such as displayed in Figures 12, 13, 14, and 15, can often be confusing. Non-parametric statistical techniques have been developed to help make underlying spatial patterns easier to discern. One such tool is kernel density estimation, a robust spatial smoothing method. Continuous density surfaces, interpolated from point data (in our case, illegal border crossings), are typically displayed using a range of colors. Kernel density estimation filters out variability and is a valuable exploratory technique for identifying hot spots (see Bailey, 1994; McLafferty et al., 2000; Williamson et al., 2001).

Figure 16 shows kernel densities for all illegal entries by fiscal year. Figure 17 displays the kernel density for all illegal entries (all years combined), whereas Figure 18 displays the kernel density for criminal disposition entries (all years combined). These maps provide a better perspective on the most active illegal border crossing areas.

Figure 16.

All illegal entries kernel density, FY 2001–FY 2004

Full figure and legend (238K)

Figure 17.

Del Rio sector kernel density – all illegal entries.

Full figure and legend (232K)

Figure 18.

Del Rio sector kernel density – criminal disposition entries.

Full figure and legend (212K)

The number of all illegal entries, and the number of criminal disposition entries, per landmark mile marker are dependent variables we wish to predict from various physical geography, human geography, and environmental independent variables. Examples of two of the important GIS data layers used in this project include major transportation routes (Figure 19) and population density (Figure 20).

Figure 19.

Major transportation.

Full figure and legend (249K)

Figure 20.

Population density.

Full figure and legend (227K)

To help explore possible relationships, Pearson correlation coefficients (r) were calculated between the number of all illegal entries and various geographic and environmental variables (see Table 5). Similar correlations were calculated for the number of criminal disposition entries (see Table 6). Unless specifically stated as representing Mexico, the variables refer to U.S. geographic data. A negative correlation with a distance-to-feature variable indicates a positive correlation with proximity to that feature.

Table 5 - Correlations for all illegal entries – Del Rio Sector, Texas.

Full table

Table 6 - Correlations for criminal disposition entries – Del Rio Sector, Texas.

Full table

The correlated variables help paint a picture of the geographic preferences of illegal land border crossers. Their travel can involve several stops and stages (Spong, 2006):

• origin (in Mexico);
• waiting location (in Mexico, close to the border);
• staging area (on the south side of the Rio Grande);
• crossing (of the Rio Grande river);
• landing point (on the north side of the Rio Grande);
• intermittent destination (in Texas); and
• final destination (somewhere in the U.S.).

Each of these stops, and the travel routes between them, need to possess different geographic requirements for selection by an illegal immigrant. Proximity to closest Mexico urban area was significant for both all entries and for criminal disposition entries. The cities, towns, and villages on the south side of the Rio Grande provide waiting locations for those who want to illegally enter the United States. In some cases, these individuals will travel in a group led by a "coyote", and therefore must wait for the entry to be organized.

Crossing the Rio Grande can be dangerous, especially during times of high or fast water flow volumes.⁵ Proximity to natural bridges (river islands or sand bars, typically caused by deposition of granular material sediment) along the Rio Grande was significant, though proximity to constructed bridges, which are well guarded, was not. Also significantly correlated for both all entries and criminal disposition entries were distance from closest large flowing river, distance from flowing streams and small rivers (excluding the Rio Grande), distance from medium flowing river, and proximity to closest intermittent stream. This may be a function of the risk associated with crossing the Rio Grande near the turbulence created by two rivers joining. An intermittent stream, however, does not produce much of a disturbance, but does provide a walking route.

Along with intermittent streams, there are significant correlations for all entries with proximity to other features that provide walking paths, including railroad spur lines and rural highways. Criminal entries are more cautious, however. Proximity to railroad spur lines is still significant, but so is distance from railroad main lines and distance from paved county roads.

All entries were significantly correlated with proximity to various human settlement features, including urban area and city streets (most of the urban feature variables are highly intercorrelated). Interestingly, while the number of criminal disposition entries was significantly correlated with proximity to Mexican urban areas, it was not with proximity to U.S. urban areas. While a U.S. border town is invariably paralleled by a Mexican settlement, the converse is not always so. These findings outline the importance of Mexican settlements as set-up locations for both all illegal entry and criminal disposition entry crossings. The former are also influenced by the location of U.S. border settlements.

According to U.S. Border Patrol agents, vegetation may play a role in where illegal migrants choose to cross the border. The river cane (carrizo) which grows along the banks of the Rio Grande provides very effective cover for surreptitious entry into the United States. Unfortunately, the categorical nature of vegetation data and the relatively large size of the LMM sectors, precluded a suitable analysis, and we were not able to assess the influence of vegetation on the location of illegal immigration.

Illegal border crossers often set up staging areas on the south bank of the Rio Grande. U.S. Border Patrol agents advised us these areas were usually selected on the basis of certain geographic features, such as the presence of arroyos (wadis) to hide in, and existing paths along the north bank to facilitate ascent from the river. A positive differential in terrain elevation between the south and north banks was also preferred (i.e., locations where the south bank was higher than the north bank). This allowed migrants an effective means of scanning for Border Patrol vehicles on the U.S. side of the Rio Grande before crossing the river.

We therefore conducted a viewshed analysis in the Del Rio Sector. A high viewshed value indicates that the location has a good view of the nearest LLM area. This is partly a function of relative elevation. The viewshed analysis for all illegal entries at the half-mile range from a LLM is shown in Figure 21. Table 7 shows the correlations between both all entries and criminal disposition entries and viewshed values, for various distances (0.5 to 5 miles) and in both directions from the border (Mexico and U.S.). Out of the 12 possibilities, seven of the distance ranges showed a significant correlation between all entries and viewshed value, and eight of the distance ranges showed a significant correlation between criminal disposition entries and viewshed value. Viewshed may help explain the location of some of the criminal disposition entries.

Figure 21.

Viewshed analysis.

Full figure and legend (246K)

Table 7 - Correlations for viewshed analysis – Del Rio Sector, Texas.

Full table

Conclusion

The purpose of this research project was to identify physical and human geographic features related to the probability of illegal border crossing. A better understanding of the spatial behavior of illegal immigrants can help the U.S. Border Patrol anticipate and respond to such activity. We observed both spatial and temporal clustering, evidence of preferences for where and when the southern border is illegally crossed. The desirability of certain locations and times reflect rationale choices by illegal border crossers to the opportunities and risks presented by the physical and human environments.

Border control is an important and challenging element of national security. Recently, it has also been a highly controversial political issue (see Massey et al. (2002), for another perspective on the illegal immigration problem). The Secure Fence Act of 2006 was passed by the U.S. Senate in September 2006 and signed by President Bush a month later. The Act requires the construction of at least 700 miles of reinforced fencing along the U.S.–Mexico border, and outlines specific construction locations which reflect current high activity areas.

This initiative (assuming it is ever implemented) would still leave 1,300 miles of unfenced southern border. The Federal government's plan is to protect these areas with virtual fencing involving cameras, ground sensors, UAVs, and other forms of high technology surveillance. This net will be bolstered by increasing the number of Border Patrol agents. Despite these efforts, the potential for a partial border fence to produce spatial displacement is extremely high. Knowledge of the movement patterns of illegal border crossers and the environmental characteristics associated with desirable crossing points can assist the Border Patrol in anticipating local impacts of a border fence.

Beyond optimizing resource allocation, such knowledge could inform intelligence analysis and inter-jurisdictional information sharing. Agencies that may benefit, in addition to the Border Patrol, include the Department of Homeland Security, the Drug Enforcement Agency (DEA), the Bureau of Citizenship and Immigration Services (BCIS), county sheriff and small local police departments close to the border, and large metropolitan police departments in border regions.

This study was only a first step and consideration should be given to expanding the analysis to the entire U.S.–Mexico border. Examination of the relationship between environmental factors and the volume of illegal border crossings was accomplished through the use of a geographic information system. Geocoding all Landmark Mile Markers would facilitate further research and tactical analysis. Furthermore, if Border Patrol agents were issued global positioning system (GPS) units, they could accurately geocode apprehension sites and other locations of interest. Information has always been a precursor to effective security.

Notes

¹ The El Paso Sector includes greater El Paso, Texas, and all of New Mexico.

² The fiscal year for the U.S. Border Patrol runs from October 1 to September 30. We were given the 2004 database in mid-August, so we are missing approximately 6 weeks of data for that particular fiscal year. Initially, we were provided the complete national database for fiscal years 1998–2001. However, not all Border Patrol sectors entered their information into the new apprehension data collection system, ENFORCE, until the start of fiscal year 2001 (October 1, 2000). ENFORCE has an event location field. We subsequently received data for fiscal years 2002 to 2004, which also contained event location information.

³ The totals may not equal 254,717 as not all records had information for a given personal characteristic. In such instances, valid percentages are given.

⁴ LMMs with low activity levels can easily show large, but insignificant, change, as illegal immigrants are often arrested in groups. They were therefore eliminated from this analysis.

⁵ According to Border Patrol agents, illegal entries are more likely to be associated with time periods of moderate precipitation than with periods of high or low precipitation. The Amistad Dam and Reservoir is located in Val Verde County, not far from the city of Del Rio. The Dam was built for flood control, conservation, irrigation, power, and recreation. Operated by the International Boundary and Water Commission, the Amistad Reservoir regulates the flow of the Rio Grande for downstream users. During periods of low rainfall, more water is released for agricultural irrigation. Therefore, the water level of the Rio Grande can actually rise during extended periods of low rainfall, making illegal crossings more difficult.

References

1. Ackleson, J. (2003) Directions in Border Security Research. The Social Science Journal. Vol. 40, No. 4, pp 573–581. | Article |
2. Anderson, J.B. (2003) The U.S.–Mexico Border: A Half Century of Change. The Social Science Journal. Vol. 40, No. 4, pp 535–554. | Article |
3. Bailey, T.C. (1994) A Review of Statistical Spatial Analysis in Geographical Information Systems In Fotheringham, A.S. and Rogerson, P. (eds) Spatial Analysis and GIS. London: Taylor & Francis, pp 13–44.
4. Brantingham, P.J. and Brantingham, P.L. (eds) (1981) Environmental Criminology. Beverly Hills: Sage.
5. Brantingham, P.J. and Brantingham, P.L. (1984) Patterns in Crime. New York: Macmillan.
6. Clark, W.A.V. (1986) Human Migration. Sage University Paper Series on Scientific Geography, 7.Beverly Hills: Sage.
7. Clarke, R.V. and Felson, M. (eds) (1993) Routine Activity and Rational Choice. New Brunswick, NJ: Transaction.
8. Clarke, R.V. and Guerette, R.T. (2005) The Border Safety Initiative: Evaluation, Assessment and Recommendations for Strategic Action: Final Report. Final Report Submitted to the United States Border Patrol.Washington, DC: Department of Homeland Security.
9. Cornish, D.B. and Clarke, R.V. (eds) (1986) The Reasoning Criminal: Rational Choice Perspectives on Offending. New York: Springer-Verlag.
10. Crawford, G., Reuter, P.H., Isaacson, K.E. and Murphy, P. (1988) Simulation of Adaptive Response: A Model of Drug Interdiction (N-2680-USDP).Santa Monica, CA: RAND.
11. Fotheringham, A.S. and Rogerson, P. (eds) (1994)Spatial Analysis and GIS. London: Taylor & Francis.
12. Giovando, C. and Zhang, T. (2005 June) Spatial Knowledge Discovery Through an Integration of Visual Data Exploration with Data Mining. Paper presented at the UCGIS Summer Assembly, Jackson, WY.
13. Haynes, K.E. and Fotheringham, A.S. (1984) Gravity and Spatial Interaction Models. Sage University Paper Series on Scientific Geography, 2.Beverly Hills: Sage.
14. Kossoudji, S.A. (1992) Playing cat and mouse at the U.S.-Mexican border. Demography. Vol. 29, No. 2, pp 159–180. | Article | PubMed | ChemPort |
15. Maril, R.L. (2004) Patrolling Chaos: The U.S. Border Patrol in Deep South Texas. Lubbock, TX: Texas Tech University Press.
16. Massey, D.S., Durand, J. and Malone, N.J. (2002) Beyond Smoke and Mirrors: Mexican Immigration in an Era of Economic Integration. New York: Russell Sage Foundation.
17. McLafferty, S., Williamson, D. and McGuire, P.G. (2000) Identifying Crime Hot Spots Using Kernel Smoothing. In Goldsmith, V., McGuire, P.G., Mollenkopf, J.H. and Ross, T.A. (eds) Analyzing Crime Patterns: Frontiers of Practice. Thousand Oaks, CA: Sage, pp 77–85.
18. Pick, J.B., Viswanathan, N. and Hettrick, J. (2001) The U.S.–Mexican Borderlands Region: A Binational Spatial Analysis. The Social Science Journal. Vol. 38, No. 4, pp 567–595. | Article |
19. Rossmo, D.K. (2000) Geographic Profiling. Boca Raton, FL: CRC Press.
20. Spong, J. (2006 July) My Life as an Illegal. Texas Monthly. Vol. 34, No. 7, pp 102–105, 176–177.
21. Taylor, P.J. (1977) Quantitative Methods in Geography. Prospect Heights, IL: Waveland Press.
22. Turner, J. (2004) Transforming the Southern Border: Providing Security & Prosperity in the Post-9/11 World. Washington, DC: U.S. House of Representatives, Select Committee on Homeland Security.
23. U.S. Department of Homeland Security (2004) Yearbook of Immigration Statistics, 2003. Washington, DC: U.S. Government Printing Office.
24. U.S. Department of Homeland Security (2005) Yearbook of Immigration Statistics, 2004. Washington, DC: U.S. .Government Printing Office.
25. U.S. General Accounting Office (1999) U.S.–Mexico Border: Issues and Challenges Confronting the United States and Mexico (GAO/NSIAS-99-190).Washington, DC: US Government Printing Office.
26. Vila, P. (2003) Processes of Identification on the U.S.–Mexico Border. The Social Science Journal. Vol. 40, No. 4, pp 607–625. | Article |
27. Weber, L. and Bowling, B. (2004) Policing Migration: A Framework for Investigating the Regulation of Global Mobility. Policing and Society. Vol. 14, No. 3, pp 195–212. | Article |
28. Williamson, D., McLafferty, S., McGuire, P., Ross, T., Mollenkopf, J., Goldsmith, V. and Quinn, S. (2001) Tools in the Spatial Analysis of Crime In Hirschfield, A. and Bowers, K. (eds) Mapping and Analysing Crime Data: Lessons From Research and Practice. London: Taylor & Francis, pp 187–202.
29. Wilson, T.D. (2002) Border Games: Policing the U.S.–Mexico Divide [Review of Border Games: Policing the U.S.–Mexico Divide]. Review of Radical Political Economics. Vol. 34, No. 3, pp 343–384.

Acknowledgements

We would like to acknowledge the invaluable support of Kenneth Rohde, Ryan Cast, and Sean Thrash, of the U.S. Border Patrol. We also wish to thank Carl Schmiedeskamp, Michele Quinones, and Dr. Yongmei Lu, from Texas State University.