Laveesh Bhandari, Koel Roychowdhury Indicus Analytics Private - - PowerPoint PPT Presentation

Laveesh Bhandari, Koel Roychowdhury Indicus Analytics Private Limited New Delhi, India ,

Introduction Research objectives Research objectives Datasets used Methods

• Satellite Image Processing
• GDP data processing

Results Results

• Correlations
• Models

Maps Maps

• Total GDP
• Sectoral GP

R i f D l t

• Regions of Development

Summary

2

» DMSP-OLS night-time images primary source of data

for project for project

» DMSP-OLS used for a variety of applications (e g » DMSP-OLS used for a variety of applications (e.g.

environmental sustainability, urban mapping and light pollution etc)

» Problem of unavailability of data on economic

activities particularly for small administrative regions activities, particularly for small administrative regions Propose an approach to produce GDP data at the sub

» Propose an approach to produce GDP data at the sub-

national level using night-time satellite images

3

What is the utility

• f

DMSP OLS image for What is the utility

• f

DMSP-OLS image for accurately predicting Gross Domestic Product (Total and Sectoral) at a sub-national level? and Sectoral) at a sub national level?

Achieve this by:

Investigating statistical relationships between GDP and

information derived from DMSP-OLS images g

Development of prediction models Validation of models Application of models to derive GDP maps at sub national level Application of models to derive GDP maps at sub national level

for India

4

Satellite Images Satellite Images

• DMSP-OLS

Average Digital Number (DN) data (2008) Average Digital Number (DN) data (2008)

GDP D t t

GDP Datasets

• Indicus Analytics Pvt. Limited (2008)

5

Differences in average DN between satellites Differences in average DN between satellites Reference Image: captured by satellite F12 in 1999

• ver Sicily

Second order regression equation: Calculation of SUM of stable lights

Details can be found at:

Elvidge, C, Ziskin, D, Baugh, K, Tuttle, B, Ghosh, T, Pack, D, Erwin, E & Zhizhin, M 2009, 'A Fifteen Year Record of Global Natural Gas Flaring Derived from Satellite Data', Energies, vol. 2, no. 3, pp. 595-622.

6

Distributing the state level GDP of the sector into Distributing the state level GDP of the sector into

each district.

Creating two indices using principle inputs and Creating two indices using principle inputs and

• utputs.

Y(K, L, M)

• where Y is the output, K is the capital, L is labour and M is a

general variable related to land and natural resources.

The second index is an additive index based on The second index is an additive index based on

normalized sectoral data.

The natural log of the GDP at the district level was The natural log of the GDP at the district level was

calculated and was used in this study.

7

8

Ln (Y) = α + β1 Ln(X1) + β2 Ln(X2) + β3 Ln(X3) + β4 X4 + β5 X5 + β6 X6 + β7 X7 + β8 X8 + β9 X9 P di V i bl T l GDP GDP from Primary S GDP from Secondary S GDP from Tertiary S Predictor Variables Total GDP Sector Sector Sector (Constant) , α 0.86 ‐2.42 1.25 ‐0.97 Log Normal of Sum of Lights, β1 0.36 0.31 0.53 0.33 Log Normal of Area , β2 ‐0.11 0.04 ‐0.14 ‐0.14 Log Normal of Total Population, β3 0.47 0.57 0.24 0.58 Dummy for Metropolitan Districts, β4 1.49 ‐0.62 1.45 ‐1.73 Dummy for Suburbs of Metro cities, β5 0.90 ‐0.41 1.46 1.00 Dummy for Large Towns, β6 0.52 ‐0.09 0.79 0.63 Dummy for Capital Districts, β7 0.62 ‐0.57 0.86 0.85 Dummy for Mountainous Districts, β8 0.04* 0.17* ‐0.01* 0.12* Dummy for Snow‐Covered Districts, β9 0.10 ‐0.38 0.44 0.19 Adjusted R2 0.88 0.73 0.73 0.87

9

* Significant at less than 50% of the cases

Predicted Total GDP for India % Error in the predicted results India results

10

Predicted GDP for Primary sector for India % Error in the predicted results Primary sector for India results

11

Predicted GDP for Secondary sector for India

% Error in the predicted results

Secondary sector for India

results

12

Predicted GDP for Tertiary sector for India % Error in the predicted results Tertiary sector for India results

13

14

Strong correlations exist between sum of stable lights as

g g

• btained from DMSP-OLS night time images and GDP at the

district level for India. Non-linear correlations were noted at this level.

Total GDP was predicted for the districts of the country as a

whole from the models. The model was proposed with an adjusted r2 value 0.87. Sum of lights exhibited the highest impact on log of GDP than other predictor variables in the impact on log of GDP than other predictor variables in the model.

Models for sectoral GDP were also predicted. This included

the major sectors of the economy such as primary the major sectors of the economy such as primary, secondary and tertiary sectors. GDP in primary sector was mainly predicted on the impact of sum of lights and total

• population. Sum of lights also demonstrated to have the

p p g highest impact on the models predicting GDP from secondary and tertiary sources.

RMIT University 15

The urban centres were found to be more

developed with the highest contribution to GDP developed with the highest contribution to GDP in the secondary and tertiary sectors.

High, medium and less developed zones of the

g , p country were identified, the latter with very low contribution to GDP by the tertiary and d t secondary sectors. Thi l i l h h h i f i b i d This paper conclusively shows that the information obtained from the night time DMSP-OLS images can be successfully used to predict GDP at the district level and map areas on used to predict GDP at the district level and map areas on the basis of their economic development.

RMIT University 16