Spatial statistics used by geographers

Detail

Why geographers adjust statistics for location

• Ordinary statistics include the mean (the average) and standard deviation (how far values spread from the average). They describe numbers but ignore where those numbers occur.
• Geographers study places, so they adapt statistics to take location into account. This goes beyond NCERT.
• Spatial statistics are methods that use location (a district, a village point, the distance between places) as part of the calculation.
• There are two classic tools:
• the location quotient, which compares regions;

• nearest neighbour analysis, which looks at the pattern of points on a map.

• The input data comes from sources such as the Census of India (since 1872) and the NSS (National Sample Survey, since 1950). Both give figures district by district.

Location quotient (LQ)

• Location quotient (LQ): a ratio that compares two shares:
• a region's share of one activity (for example, the share of its workers who are in textiles);

• the nation's share of the same activity.

• Formula:

• LQ = (workers in activity i in the region ÷ all workers in the region) ÷ (workers in activity i in India ÷ all workers in India)

• Worked example (illustrative numbers):

• Suppose 20% of a district's workers are in textiles.
• Suppose 5% of India's workers are in textiles.

• Then LQ = 20 ÷ 5 = 4. The district has four times the national concentration of textile workers.

• How to read the value:

LQ value	Meaning	What it tells the planner
> 1	The region specialises in the activity	The activity is part of the region's economic base, which "exports" goods or services to other regions
= 1	The region's share is the same as the nation's	No special concentration
< 1	The region is under-represented	The region probably brings in this good or service from elsewhere

• Uses:
• It finds industrial clusters. A cluster is a group of firms of one industry in one area, such as textile districts in Gujarat, Tamil Nadu or Punjab (see manufacturing-industries-india).
• It finds a region's economic base: the activities that bring money into the region from outside.

• It compares regions fairly, because it uses shares and not raw totals. A big state and a small district can be put side by side.

• Limits:

• LQ shows concentration, not efficiency or output.
• It depends on how the regions are drawn. If district borders change, the result can change. This is the MAUP (modifiable areal unit problem: results change when boundaries change).
• High specialisation has a downside. If that one industry suffers a slump, the whole region is hit hard.

Nearest neighbour analysis (NNA)

• Nearest neighbour analysis: a method that measures whether points on a map (villages, wells, shops) are clustered, random or regular.
• It was developed by ecologists P.J. Clark and F.C. Evans (1954). Geographers later adopted it for settlements.

• Three steps: 1. Measure the distance from each settlement to its nearest neighbour. Take the average (observed mean distance, D̄obs). 2. Work out the distance you would expect if the same number of settlements were scattered at random over the same area: D̄exp = 0.5 ÷ √(n/A), where n = number of settlements and A = area. 3. Divide the first by the second to get the index: Rn = D̄obs ÷ D̄exp, which works out to Rn = 2D̄obs √(n/A).

• Reading Rn:

Rn value	Pattern	What it looks like on the ground
≈ 0	Clustered	All points crowd at one spot
≈ 1	Random	No clear order
up to ≈ 2.15 (2.1491)	Perfectly regular	Points evenly spaced in a hexagonal (six-sided) grid, like Christaller's central places

• In later studies, any index below 1 is read as clustered and any index above 1 as dispersed. Software also gives z-scores and p-values, which test whether the pattern could just be chance [5][6].
• Use in India: classifying rural settlements (see human-settlements). The NCERT types match the index like this:
• Clustered or compact villages give a low Rn. They are found in the fertile alluvial plains of the north and in the north-eastern states.
• Semi-clustered and hamleted villages give a middle Rn. They are found in the Gujarat plains, parts of Rajasthan, the middle and lower Ganga plain, Chhattisgarh and lower Himalayan valleys.

• Dispersed settlements give an Rn moving towards regular. They are found in the hills of Meghalaya, Uttarakhand and Himachal Pradesh, and in parts of Kerala.

• Limits:

• Rn depends on the boundary of the study area. A larger area makes points look more clustered.
• It treats each settlement as a point and ignores its size.
• It uses straight-line distance, not actual travel distance over rivers or hills.

Companion tools in modern settlement studies

• Recent studies use NNA together with two other tools [5]:
• Kernel density estimation: a smooth "heat map" of how dense the points are. It shows where the clusters are.

• Standard deviation ellipse: an oval drawn around the points. It shows which direction the cluster spreads or shifts.

• Each tool answers its own question [5]:

• NNA: "Is it clustered?"
• Kernel density: "Where is it clustered?"

• Standard deviation ellipse: "How does the cluster move?"

• A colocation quotient extends the LQ idea. It tests whether two kinds of features (for example, crime spots and bus stops) tend to occur near each other [7].

Spatial statistics inside GIS

• GIS (geographic information system): computer software that stores maps and data tables together, so you can calculate on maps (see remote-sensing-gis).
• Today, LQ and NNA run as built-in GIS tools, which joins statistical processing with spatial analysis:
• The software reads village points or district boundaries.
• It computes the index.

• It draws the result as a map.

• Indian platforms that supply the base data:

• Bhuvan, ISRO's geoportal built by NRSC (Hyderabad):
  • you can view remote-sensing data on it and download satellite data free of cost [2];
  • government departments can host their own mapping applications on it [2];
  • it works as a reference platform for spatial and non-spatial databases [2].

• PM GatiShakti National Master Plan, built by BISAG-N:

  • it is a GIS platform with data layers on infrastructure, geographic features and population [3];
  • about 1,700 data layers have been integrated: 969 from States/UTs and 731 from Central Ministries/Departments (2025) [3];
  • it maps economic, social and logistics nodes (hubs), so planners can spot missing connections [3].

• National Geospatial Policy 2022:

• it is aligned with PM GatiShakti, which links 16 key Ministries [4];
• geospatial data created with public money is to be openly accessible [4];
• it targets a high-resolution topographic survey and mapping system and a highly accurate Digital Elevation Model (DEM: a digital 3-D height map of the land) for the whole country by 2030 [4].

Prelims Hooks

• LQ > 1 = regional specialisation (part of the economic base); LQ < 1 = under-represented; LQ = 1 = same as the national share.
• LQ is a ratio of two ratios (the regional share ÷ the national share), not a ratio of absolute totals. This is a common trap.
• Nearest neighbour index was given by Clark and Evans, 1954. They were ecologists, not economists.
• Rn ≈ 0 → clustered; Rn ≈ 1 → random; Rn ≈ 2.15 → perfectly regular (hexagonal).
• The maximum Rn (≈ 2.15) comes from hexagonal spacing, the same geometry as Christaller's central place theory.
• Kernel density shows where points cluster. Nearest neighbour index shows whether they cluster [5].
• Bhuvan is ISRO's geoportal, developed by NRSC [2]. The PM GatiShakti NMP platform was developed by BISAG-N [3].
• The National Geospatial Policy 2022 targets a nationwide high-resolution map and DEM by 2030 [4].

Mains Points

• Regional planning through LQ. District-level LQs from Census and NSS/PLFS (Periodic Labour Force Survey) employment data find clusters such as textiles and leather. These guide cluster programmes and "One District One Product"-type plans. The caveat: a high LQ also means the region depends on one industry and is exposed if that industry slumps.
• Settlement planning through NNA. NNA puts a number on the clustered vs dispersed contrast:
• dispersed hill settlements mean a higher cost of services per household, for roads, schools and power lines;
• clustered plain villages mean services are cheaper to reach, but pressure on land is greater.

• These figures help with the placement of facilities under rural infrastructure schemes.

• GIS joins statistics with maps. Platforms such as Bhuvan and PM GatiShakti (about 1,700 layers) let planners run spatial statistics on official data layers [2][3]. The National Geospatial Policy 2022 widens open access to such data [4]. This shifts analysis from description to evidence-based, location-specific decisions (GS-III: infrastructure).

• Limits to state in answers. The MAUP, boundary effects, and point-based simplification mean results must be read alongside fieldwork and local knowledge.