Agricultural Water Survey (AWS)

Detailed information for 2012

The Agricultural Water Survey is conducted to gather information on irrigation water use, irrigation methods and practices, and sources and quality of water used for agricultural purposes on Canadian farms. The results will help farm operators, governments and the Canadian public gain a better understanding of the demand for water and how it is used on Canadian farms.

Data release - December 13, 2013

• Questionnaire(s) and reporting guide(s)
• Description
• Data sources and methodology
• Data accuracy

Description

The Agricultural Water Survey is conducted to gather information on irrigation water use, irrigation methods and practices, and sources and quality of water used for agricultural purposes on Canadian farms.

This survey is part of the Canadian Environmental Sustainability Indicators (CESI) program. The data collected will be used in CESI's reporting activities. The information will also be used by Agriculture and Agri-Food Canada to inform water use policy and development of programs for Canadian irrigators. Statistics Canada will also use the survey results to improve the modelling of irrigation volumes by type of crops and continue to report on total water use by sector in Canada.

Subjects

• Agriculture and food (formerly Agriculture)
• Environment
• Environmental quality
• Land use and environmental practices

Data sources and methodology

Target population

The target population for this survey is the Canadian farm operations that irrigate. The survey frame was created using information collected as part of the 2011 Census of Agriculture (CEAG). The statistical unit was the agricultural operation. Any unit which reported sales of $10,000 or more and reported either irrigating in 2010 or owning irrigation equipment was considered to be part of the initial survey frame.

A number of groups were removed from the initial survey frame:

- All institutional farms (for example, government, university and prison farms), Indian reserve farms and community pastures.
- All units which reported greenhouse, sod, nursery, mushroom or Christmas tree operations on the 2011 Census of Agriculture.
- All units that belong to Statistics Canada's Large Agricultural Operations Statistics program. These very large and complex units have special collection agreements with Statistics Canada concerning the surveys for which they will provide data.
- All units for which the 2011 Census of Agriculture irrigation data was completely imputed.
- All units which reported only irrigation area in the "Other" category on the 2011 Census of Agriculture and did not report owning any irrigation equipment.
- All units in the seven most northern of Canada's 25 drainage regions (DRs): Yukon (5), Peace-Athabasca (6), Lower Mackenzie (7), Arctic Coast-Islands (8), Keewatin-Southern Baffin Island (16), Northern Ontario (17) and Northern Quebec (18).

The remaining 12,055 units comprised the survey frame.

Instrument design

The 2012 AWS questionnaire was revised by Environment Accounts and Statistics Division. Questionnaire design specialists from Statistics Canada's Questionnaire Design Resource Centre (QDRC) were consulted. The revised questionnaire was tested in November and December, 2011. One-to-one telephone interviews were conducted for both English (6 respondents) and French (3 respondents) questionnaires. The questionnaire was updated based on feedback from testing. The questionnaire was further revised based on recommendations from questionnaire design specialists.

Sampling

This is a sample survey with a cross-sectional design.

The survey frame came from the 2011 Census of Agriculture and included all farm operations that reported irrigated area or ownership of irrigation equipment and that corresponded to the criteria mentioned in the target population section. The sampling unit was the farm operation.

A stratified sample design was used. Geographic strata were defined at the drainage region (DR) level or, when there were small populations within an individual DR, groups of DRs. This resulted in 14 geographic strata. Within each of these strata, the population was divided into four sub-strata based on their predicted water use for irrigation. This predicted value was derived from a model which used data from the 2011 Census of Agriculture, the 2007 Agricultural Water Use Survey and the 2010 Agriculture Water Survey. Units were categorized into one of four sub-strata of zero, low, medium and high predicted water use. The thresholds for these sub-strata varied from one geographic stratum to the next.

The sample was allocated in order to meet predefined coefficient of variation targets for predicted water use at the geographic stratum (DR group) level while at the same time not exceeding a maximum sample size of 2,000 units to be sent for collection. The targeted coefficients of variation were not consistent from one DR group to the next. In those DR groups where greater irrigation was anticipated, the targets were lower than those used in other areas. The total sample size was 1,994 units.

In order to reduce the response burden on those farmers who had been selected for recent Statistics Canada surveys, a sample coordination method known as the microstratum approach was used. Within a sub-stratum, the units which had recently been selected by other Statistics Canada agriculture surveys were less likely to be selected for the Agricultural Water Survey (AWS).

Data sources

Data collection for this reference period: 2012-10-15 to 2012-12-14

Responding to this survey is voluntary.

Data are collected directly from survey respondents.

Prior to collection, a letter explaining the goal and objectives of the survey, along with a description of the information asked in the questionnaire, was sent to the respondents. Paper questionnaires were mailed in mid-October and respondents who had one month to complete and return the questionnaires.

A Computer Assisted Telephone Interview (CATI) data collection application was developed for follow-up for this survey and some telephone interviews for the AWS were conducted from Statistics Canada's regional office in Sturgeon Falls. The follow-up period was from November 19th to December 14th, 2012.

View the Questionnaire(s) and reporting guide(s) .

Error detection

The paper questionnaires are captured into an electronic format at Statistics Canada's Head Office. The responses from the CATI interviews are downloaded directly to this format. An initial set of critical edits is run against the data to identify inconsistencies. Statistical methods are also used to identify units which appear to have questionable reported values. When important inconsistencies are identified, Statistics Canada personnel attempt to contact the respondent by telephone for clarification and correction if necessary.

The CATI application is programmed with a number of the capture edits so that inconsistencies are identified and resolved at the time of the telephone collection rather than during data processing at Head Office.

Imputation

Data imputation is used for some important fields when respondents provide incomplete or inconsistent data. Data on irrigated area or water volumes used for irrigation are imputed using an automated nearest neighbour approach. Data for the other fields are not imputed, but rather left with a "don't know" response. Complete non-respondents are not imputed, but rather accounted for in the estimation step.

Estimation

Because the AWS is a sample survey, sampling weights are applied to individual respondents according to the number of units in the population that they represent. The initial or design weights are calculated based on the probability of the unit being selected for the sample. As with all surveys there is non-response, so an adjustment is made to the weights of the respondents to account for the non-responding units. Units with very high water use that are considered to be unique are assigned a weight of one. In order to estimate a characteristic for the entire population, this final weight is multiplied by the response value and summed up over the entire population. Direct variance estimation is used to measure the precision of the estimate.

Quality evaluation

A comparison of AWS's water volume and irrigated area estimates is made with estimates obtained from the previous survey cycle. Data are also compared with administrative data sources where available.

Disclosure control

Statistics Canada is prohibited by law from releasing any information it collects which could identify any person, business, or organization, unless consent has been given by the respondent or as permitted by the Statistics Act. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.

Tabular data suppression techniques are used to prevent such disclosure. Cells in published tables which are considered at risk to identify an individual's data are suppressed (the value replaced with an x). Additional cells in the table may be suppressed in order to eliminate residual disclosure. The at-risk cells are either those having a small number of respondents contributing to a cell estimate, or those identified by an automated approach known as tabular data cell suppression. This technique measures the sensitivity or the risk of disclosure in each cell, identifies the ones at risk and determines if any other cells also need to be suppressed in order to maintain confidentiality.

Estimates of area are rounded to the nearest ten hectares. Volume estimates are reported to the nearest thousand cubic metres. A random rounding approach is used for estimates containing counts, where cell estimates are randomly rounded up or down to a multiple of five. This means that the sums of rounded values and the rounded marginal totals may not necessarily be equal.

Revisions and seasonal adjustment

This methodology does not apply to this survey.

Data accuracy

The statistics contained in this publication are estimates derived from a random sample of Canadian farms and, as such, are subject to sampling and non-sampling errors. The quality of the estimates thus depends on the combined effect of these types of errors.

Sampling errors:
These errors arise because observations are made only on a sample and not on the entire population. The sampling error depends on such factors as the size of the sample, the variability of the characteristic of interest in the population, the sampling design and the method of estimation. For example, for a given sample size, the sampling error will depend on the stratification procedure employed, allocation of the sample, choice of the sampling units and method of selection. In sample surveys, since inference is made about the entire population covered by the survey on the basis of data obtained from only a part of the population, the results are likely to be different than if a complete census was taken under the same general survey conditions. The most important feature of probability sampling is that the sampling error can be measured from the sample itself.

Typically the sampling error is measured by the expected variability of the estimate from the true value, expressed as a percentage of the estimate. This measure is expressed as the coefficient of variation (CV). Coefficients of variation of the final estimates were computed for the Agricultural Water Survey and are indicated on the statistical tables. The quality of the estimates was classified as follows:

A. Excellent CV is 0.00% to 4.99%
B. Very good CV is 5.00% to 9.99%
C. Good CV is 10.00% to 14.99%
D. Acceptable CV is 15.00% to 24.99%
E. Use caution CV is 25.00% to 49.99%
F. Unreliable CV is > 49.99% (data are suppressed)

Non-sampling errors:
These errors are present whether a sample or a complete census of the population is taken. Non-sampling errors may be introduced at various stages of data collection (non-response, differences in the interpretation of questions, incorrect information from respondents) and data processing (such as coding, data entry, editing, weighting, tabulation, etc.). All efforts are undertaken to minimize non-sampling errors through questionnaire testing, extensive edits, quality control steps and data analysis, but some of these errors are outside the control of Statistics Canada.

The 2012 survey estimates for both irrigation volume and area of land that received irrigation showed growth rates that varied widely across regions. Differences in weather patterns, crop types and farming practices can all lead to these variations. Readers are also advised that this is only the second iteration of the survey and that, given the length of the time series, comparisons from a time series perspective should be made with caution.

Response rate:
After performing the editing and imputation steps and excluding the out-of-scope units, the resulting response rate was 75.5%.