1. Sampling Methodology for Rice Crop Cutting Pilot Survey
David J. Megill
ADB Statistical Consultant
Ha Noi, Viet Nam
November 2015

2. Objectives of pilot survey
Test methodology for improving quality and timeliness of rice crop area and production estimates and forecasts
Includes rice crop-cutting component and farmer’s response survey
Predominant rice-producing province in each of 4 countries (Viet Nam, the Philippines, Lao PDR and Thailand) selected for pilot survey

3. Sampling Frame
Sampling frame for the Crop Cutting Survey in each country is an area frame defined using satellite images that are stratified into different categories related to the concentration of rice-growing areas
Primary sampling unit (PSU) is a 200 m by 200 m square “mesh” which is spatially defined on a digitized satellite image map.

4. Stratification Four strata defined for sampling frame
(1) Rice area maps obtained by the International Rice Research Institute (IRRI) using 2015 MODIS data and from land use maps produced by the European Space Agency (ESA) in 2009 under its GLOBCOVER initiative, which provides the spatial extent of all crops growing in the selected pilot provinces 

5. Stratification (continued)
(2) Rice areas identified from IRRI area maps
(3) Crop areas identified by GLOBCOVER
(4) remaining areas where presumably rice is not grown (urban, forest, water, etc.)

6. Distribution of Meshes by Stratum for Each Pilot Province
Country/Province
Lao PDR/
Savannakhet
Philippines/
Nueva Ecija
Thailand /
Ang Thong
Viet Nam /
Thai Binh
IRRI
4,650
157
280
589
GLOB
154,227
66,127
2,777
4,846
IRRI+GLOB
80,839
68,514
22,105
36,376
Others
322,391
77,292 34
1,815
Total
562,107
212,090
25,196
43,626

7. Sample size and allocation
Stratified random sample of 120 meshes selected for the pilot survey in each province
Random sample of reserve meshes for possible replacement for each province
Sample allocation concentrates sample in strata with more rice

8. Allocation of Sample Meshes by Stratum for Each Pilot Province
Replacements
Total
IRRI+GLOBCOVER 80 10 90
IRRI 20 30
GLOBCOVER 15 5
OTHER
TOTAL
120
150

9. Actual Number of Sample Meshes Surveyed
Stratum
Country/Province
Lao PDR/
Savannakhet
Philippines/
Nueva Ecija
Thailand/
Ang Thong
Viet Nam/
Thai Binh
IRRI+GLOBCOVER 79 75
IRRI 19 20
GLOBCOVER 15 14
OTHER 5 6
Total
118
113
120
119

10. Final Distribution of Sample Meshes with Rice by Province and Stratum
Country/Province
Lao PDR/
Savannakhet
Philippines/
Nueva Ecija
Thailand /
Ang Thong
Viet Nam /
Thai Binh
IRRI+GLOBCOVER 56 66 69 63
IRRI 9 7 13 2
GLOBCOVER 10 8
OTHER 1 5
Total 77 75 91

11. Listing operation
For each sample mesh, listing of plots with at least part of the area within the boundaries of the mesh
All plots where rice would be harvested that season eligible for selection at the second sampling stage

12. Second sampling stage
At second sampling stage, 4 plots with rice selected within each sample mesh using random systematic sampling
Selection table used for selecting sample plots
Area of each sample plot measured with GPS

13. Third sampling stage
At the third sampling stage, one 2.5m by 2.5m “sub-plot” is randomly selected in each sample plot for rice crop-cutting
First random corner of sample plot selected, then random length selected for each direction to identify random sub-plot

14. Quality control and accuracy of estimates from crop cutting data
Probability sampling makes it possible to calculate unbiased estimates and measure reliability (sampling error)
Quality and operational control in crop cutting and measurement techniques reduce bias in the estimation of rice yield

15. Weighting procedures Probabilities depend on all sampling stages
where:
phij = overall probability of selection of the sample sub-plot in the j-th sample plot with rice within the i-th sample mesh in stratum h

16. Weighting procedures (continued)
nh = number of sample meshes selected in stratum h
Nh = total number of meshes in frame for stratum h
nhi = number of sample plots selected for crop-cutting in the i-th sample mesh in stratum h (generally equal to 4)
Nhi = total number of plots with rice planted this season that are at least partly located within the sample mesh

17. Weighting procedures (continued)
ahij = total area of the j-th sample plot with rice within the i-th sample mesh in stratum h (in m2)
Weight for each sub-plot

18. Estimation of Total Rice Production
Measurement of plots with rice used for estimating area planted in rice
Crop-cutting data used for estimating yield per m2
Use of ratio estimation

19. Farmer Recall Survey
Return to sample farms corresponding to the four plots selected for crop cutting
Farm operator asked about all plots planted in rice, both inside and outside the boundaries of the sample mesh
Estimation of farm characteristics based on weighted segment approach
Determine proportion of famer’s rice area inside mesh

20. Weights for Farmer’s Recall Survey data
WFhij = basic weight for the sample farm with the j-th selected rice plot in the i-th sample mesh in stratum h

21. Weighting procedures (continued)
A’hij = area of plots planted in rice inside the sample mesh for the sample farm with the j-th selected rice plot in the i-th sample mesh in stratum h
Ahij = total area of all plots planted in rice for the sample farm with the j-th selected rice plot in the i-th sample mesh in stratum h