1. Remote Sensing for Agricultural Statistics Organisation, Resources and Competences
John Latham
DDNSD
24/11/2016

2. Summary
Chapter presenting fundamental requirements and criteria for an organization starting to use remote sensing for producing agricultural statistics
It describes the need for resources and the competences necessary for application of remote sensing systems in agriculture data collection and training needs
It highlights multi-disciplinarily team composition, its qualification, size and the budget involved
It gives examples of collaboration of the statistical services with mapping agencies
It also provides indication of budget and business plan requirements

3. Remote Sensing data and agriculture statistics
Remote Sensing data area used in agriculture statistics mainly for:
Land cover monitoring
Area Frame construction
Support field data collection
Crop area estimation
Crop monitoring and yields forecasting  3

4. Requirements for using RS in agricultural statistics
Fundamental requirements:
Sustainable access to satellite image collections
Resources and competencies for data collection and processing
Robust geospatial and statistical methodologies
Comprehensive capacity building programmes
Collaboration among statistical services and mapping agencies
Interaction with stakeholders
Budget and business plan 4

5. Background
Importance of temporal, timely and accurate information on crops during the growing season
Quality of available agricultural data (agricultural census, crop monitoring and yield estimation) generally can be inconsistent and erratic
Geospatial data and remote sensing processing offer robust new data and methods for strengthening data systems
Agricultural censuses are increasingly employing geospatial technologies
RS and GIS and derived land cover provide support for area sample frame (ASF) based approaches and Master Sample Frames (MSF)
Extensive development of these approaches within other sectors; opportunity for institutional sharing 5

6. Organization
Generally, public organizations produce the agriculture statistics
The data are available only at the end of the season and generally lack temporal and synoptic character
In a number of countries now, satellite remote sensing and GIS technologies have supported agriculture data collection successfully
Advantage of this system is that it is digital, temporal, and synoptic in character and can reach inaccessible areas
However, it requires multi-disciplinary institution to integrate information from satellite remote sensing, GIS, statistics, agronomy, agro- meteorology and economics. 6

7. Resources
Resources required by any organization to start producing agriculture statistics:
Qualified manpower: examples
Remote Sensing and GIS analysts are responsible to undertake the assignment of image processing and construction of Area Frame;
Statisticians are responsible for sample design, extrapolation and final estimation;
Image analysts calculates the area based on pixel/object based classification;
Field staff are responsible to carry out field survey and crop signature collection.
Hardware / Software
Input
Funding 7

8. Resources
Resources required by any organization to start producing agriculture statistics:
Qualified manpower: examples
Hardware / Software
A laboratory (mostly 20,000 sq ft) would be required to accommodate the manpower and equipment
Hardware for data processing (workstation/laptop), data collection (GPS/smartphone/tablet), input and output in digital format (scanner/printer), storage and dissemination are required
Software for Statistical analysis, GIS/RS processing, mobile based data collection, Computer Assisted Personal Interviewing (CAPI) and metadata / data dissemination will be needed.
Input data
Funding 8

9. Modis Aug 2010
Resources
Resources required by any organization to start producing agriculture statistics:
Qualified manpower: examples
Hardware / Software
Input data
The UN has developed a series of purchase agreements with key image providers e.g. MacDonald Dettwiler (MDA) (QuickBird, IKONOS, WorldView-1, WorldView-2, GeoEye-1, WorldView-3, KOMPSAT-2, KOMPASAT-3, ZY-3 and RADARSAT-2) and Airbus DS Geo (derived from TerraSAR-X, SPOT 6/7 and Pleiades)
In the context of humanitarian actions the image sources are available through the International Charter for Space and Major Disasters .
Integration of national agricultural monitoring with the FAO/WFP CFSAM assessment missions provides technical assistance to crop production forecasts.
Funding
Digital Globe, near Lahore
SAR image example 9

10. Resources
Work plan The acquisition time period of the satellite imagery depends upon their phenological stages 10

11. Resources
Resources required by any organization to start producing agriculture statistics:
Qualified manpower: examples
Hardware / Software
Input data
Funding
Integrating remote sensing into agricultural statistical will necessitate allocation of appropriate levels of funds
Optimization of imagery acquisition to processing and field data collection and validation. Potential to share costs with other applications
Costs of verification using high and very high resolution data may reduce the costs of field validation
The cost of hardware and software are mostly one time 11

12. Training Requirements
Monitoring agriculture using Remote Sensing and GIS requires a range of multi-disciplinary team and integrated skill set. Here is a summary of required training curricula:
Basic concept of RS, GIS, Statistics and Agronomy
RS and image processing, classification, analysis and reporting
Land cover classification approaches and database development
Integration of agri-environmental parameters and ground based information for yield forecasting
RS/GIS for Area Sampling Frame
GPS operating and field data collection for enumerators
Statistical sampling techniques 12

13. Training Requirements
Trainings should be regularly provided to staff by specialized national and international organizations The number of trainings is to be determined according to the requirements/ composition of the team. E-learning materials should also be developed to support the application in crop area and yield estimation 13

14. Case examples of implementation
FAO programmes supporting land cover mapping provide examples of the application of multiple scales of remote sensing data to agricultural statistics
Ethiopia
Pakistan (provincial and national)
Bangladesh
Rwanda 14

15. Application of Area frame stratification: Ethiopia
The Central Statistical Agency(CSA) of Ethiopia has used traditional list frame sampling as the basis for field and household based agricultural surveys
Approach with limitations
The CSA conducted a comparative evaluation of list and area frame based approaches in the 2008/09 ‘meher’ season (June – Oct rainy season).
West Shewa and then Oromiya were tested.
An high res land cover database was produced using the FAO Land Cover Classification (LCCS) methods.
The LCCS based approaches offer repeatability, a level of automation and consistency based on the adoption of ISO standards ( ).
GIS technology was used to develop the Secondary Sampling Units (SSUs).
Master sampling frames for agricultural and rural statistics
Oromiya: Mapped Enumeration Areas over land cover database
Enumeration area map subdivided into segments

16. Provincial Crop Reporting Services: Pakistan
Pakistan’s provincial CRSs provide agriculture statistics to the federal Government
The approach has evolved from revenue surveys to area frame surveys supported by geospatial processing and remote sensing
The service employs 1,611 professional staff with degrees in statistics, economic and mathematics
Resources from provincial normative finance supplemented by project level activities.
Capacity development forms part of the programme, with refresher courses at the beginning of the crop season
The overall system as it stands currently is combination of objective and subjective techniques: multiple cropping area frame for major crops, and sample frame for small acreage crops

17. Operative geospatial unit: Pakistan – SUPARCO
Sampling distribution in Pakistan
Operative geospatial unit: Pakistan – SUPARCO
Also within Pakistan, with Government funding, a FAO and SUPARCO collaboration has been developing a new, integrated Agricultural Information System (AIS) at federal level supported by remote sensing and geospatial technologies and dissemination tools.
Crop area estimation is based on image processing of satellite data acquired for specific time, ground truth surveys during cropping season, crops signature collection, lab processing, accuracy assessment and crop area estimation.
Punjab South Zone broken down into sampling units (red)
Sampling unit from satellite

18. Operative geospatial unit: Pakistan – SUPARCO
Extensive ground-truthing is undertaken with the support of GPS real-time navigation
Image classification and processing tools and methods are implemented
Area frame development has been based on imagery acquired in February and September
A critical examination of the data generated is made by a team of experts in the field of Agronomy, Remote Sensing and Statistics
Dissemination of information is a key component of the this crop monitoring system:
Monthly crop bulletins (digital distribution) providing information on crop conditions during the growing season
The Crop Information Portal providing historical trends and analytical tools on crop data and related indicators
The Satellite-driven Global Agricultural Monitoring for Pakistan (GLAM) providing real-time monitoring of vegetation conditions and analytical tools for historical trend analyses
Crop Portal
GLAM Pakistan

19. Multipurpose probability sample survey: Rwanda
High res ortho-photo
Based on very high resolution (0.25m) ortho- photos, the National Institute of Statistics of Rwanda (NISR) has designed and implemented multiple frame for the country combining an area and a list frame to support national agricultural statistics.
The objective for the Rwandan survey is to provide a national, seasonal and multi- purpose (crop, livestock, forest and commodities) agricultural survey, covering a range of variables
The area frame is based on land use strata, within the 30 Districts; strata are subdivided into non-overlapping ‘segment’ sampling units (SUs), each segment is subdivided into non-overlapping tracts.
PSUs in Bugesera district
Selected PSU and Segment

20. Pilot project: Monitoring of Rice Crop in Afghanistan
The Objectives
To test relevant agriculture methodologies based on recent medium and high resolution geospatial information: Proba-V, Aqua/Terra, Landsat-8, Sentinel-1, Sentinel-2, SPOT-5/6/7 and Pleiades 1A/ 1B imagery with focus on rice monitoring.
Rice crop area estimation and Rice crop mask development
Geographical Area: Sq. Km
Agriculture area: 4100 Sq. Km

21. Methodology
Following techniques have being used to estimate the area under rice crop:
Satellite image classification
Satellite based area frame sampling technique
Regression estimator

22. Development Satellite based crop Calendar Area Frame Strata ID: 32

23. Acquisition of Satellite Data
Pleaides 1A & 1B
SPOT-6/7
Sentinel-1 & 2
Landsat 8

24. Image Classification
Temporal satellite images were used for the classification to extract area under rice crop
Image Classification
Sentinel-2, 25July 2016

25. Final Rice Crop Area Estimates(Ha) 2016-17
Rice Crop Estimates
Final Rice Crop Area Estimates(Ha)
Province
District
*MAIL IC AF
Hybrid  
Baghlan
46,005
45,431
32,555
32,196
16,227
17,325
17,842
Kunduz
43,100
86,011
90,310
40,210
26,981
23,314
28,171
Takhar
23,320
32,628
40,523
35,532
15,041
14,238
16,263
Sub Total
112,425
164,070
163,388
107,938
58,249
54,877
62,276
Badakhshan
Keshem
1,837
2,425
1,986
6,000
5,500
4,850
Balkh
Sholgareh
1,361
1,390
1,472
10,500
2,100
1,900
2,000
1361
1390
Nangarhar
Shinwar
1,468
Beshud
998
1,079
Kama
1,064
1,151
13,410
8,410
21,958
24,371
2,062
2,230
Grand Total
63,509
60,160
67,964

26. District wise Estimates

28. Crop Mask

29. Conclusions
Integration of GIS and Remote Sensing has made crop monitoring simpler, quicker and more accurate
The use of GIS and RS in the area sampling frame methodology has changed the way crop estimation tasks are conducted
Satellite imagery and derived products such as land cover have made crop area estimation and yield forecast effective and cost efficient
Implementations such as the Pakistan Agriculture Information System are extraordinary examples of improvement of national ag statistical systems through the integration of geospatial technologies in traditional approaches and methodologies 29

30. Thank you