Materials and methods

3.2.1 Instruments

The questionnaire survey is one of the effective instruments of data elicitation process.

Face-to-face interviews, telephone interviews, internet, and mail questionnaires are various modes of questionnaire survey. Of all modes, face-to-face survey delivers the most representative results. The questionnaire for this study was administered as a face-to-face interview by the researcher. In order to allow the researcher to collect the most accurate data from a target population, questionnaire must be unbiased. Bias is a problem in the design and administration of the questionnaire. It is a result of an unanticipated communication gap between the researcher and target population, which yields inaccurate results [75]. It can arise from the way an investigator questions or questionnaire as a whole is designed and administered. To avoid these potential biases various steps are suggested [75, 76, 77, 78, 79]

while designing and administering the questionnaire such as: The words used in the questions are kept simple, familiar and unambiguous to the target population. The length of the questionnaire is kept short in order to avoid response fatigue and skipping questions tendencies of respondents. Questions on a same topic are grouped together and transitional statements used to switch between different topics or sections. During the administration of questionnaire care was taken to avoid respondents’ sub-conscious reaction (tendency of trying to be conservative), conscious reaction (fake responses to seek sympathy), inaccurate recall, and cultural differences.

The questionnaire was designed referring to earlier studies by Habiba et al. [73], Manandhar et al. [80], Keshavarz et al. [70] etc. At the same time climate change, water resources and agriculture experts were also consulted. The questionnaire was designed to gather information about domestic water collection activities, crop production, employment and financial status of farmers to facilitate comparison between normal year (NY) and drought year (DY). The normal year in this study is considered as a year with timely and good amount of rainfall and during which expected or average crop production was obtained.

The drought year is defined according to the definition of IMD mentioned earlier. For better understanding and communication this survey was conducted in local language (Marathi).

The questionnaire was pre-tested with sub-sets of the target population (i.e. few farmers from two representative villages) to check the redundancy, missing information, relevancy as well as validity of the questions. The questionnaire was then revised based on pre-test results. The individuals included in pre-test were excluded from the sample considered in this study.

3.2.2 Procedure

In order to fulfill the objectives of this study (questions raised in introduction section) and to capture the impacts of drought 2012 on rural farming households, a structured

27

questionnaire survey (Appendix A) and focus group discussions (Appendix B) were conducted in the Upper Bhima Catchment in May 2013. A multi-stage stratified systematic survey sampling technique was used to select samples from the target population (villages considered as a penultimate unit and household as final unit) [81]. An individual household was considered as a primary sampling unit for this survey. In order to perform uniform sampling, the catchment was divided into three strata based on percentage of irrigation coverage - areas with less than 15% (low irrigated), between 15-30% (medium irrigated) and more than 30% (high irrigated) irrigation of the total cultivated area, averaging gridded irrigation percentages over each sub-district. It was assumed that the extent of irrigation coverage may influence the drought damage to the community. The gridded irrigation coverage in percent for the catchment was obtained from FAO Global Maps of Irrigated Area [82] and the list of villages and population data in the catchment were obtained from a website of Census of India [15].

To determine sample size for this survey, household survey sample design procedure by United Nations Statistics Division [83] was used. Total 223 households were included in this survey. In the first stage, the villages were selected by probability proportional to size (pps) sampling technique, while in the second stage households were chosen from selected villages by random walk technique. Considering the population proportion in all three strata, 76, 74 and 73-households were selected from less-, medium- and high- irrigated strata respectively from 23 villages. Each respondent household representative was interviewed face-to-face with the help of pre-tested questionnaire. The response rate of survey was almost 100% due to the respondents’ interest for discussing more about drought situation in the area, and their availability at home in off season for agriculture in the month of May. However, there was item-non response observed in few cases/questions due to inability of respondent to recall the requested information. Those cases were eliminated from the analysis.

In addition to questionnaire survey, we conducted 22 focus group discussions (FGDs) in 22 villages which were included in questionnaire survey. A group per village comprising of 8-12 persons (from households other than those involved in questionnaire) was involved in FGDs. In most of the cases a key informant or village leader (a Sarpanch or a local government member) included in respondent groups to have efficient conversation. The FGDs contained the information regarding village demographics, village water supply characteristics and the most concerning water issues according to the respondent groups during a year with normal rainfall (normal year) and drought year. The summary of the FGDs were noted in 22 questionnaires. In addition to this, 18 drinking (pot) water samples during drought were collected from a randomly selected household in a village. Samples were analyzed in laboratory with ion chromatography to determine the major anions and cations concentrations and compared with WHO drinking water quality standards. The primary and secondary data type used in this study and their sources are shown in Table 3.1.

28

Table 3.1 Data type and sources.

Sl.

No. Data Locations Timescale Source

1 Questionnaire Survey 23 2013 Researcher

2 Focus group discussions 22 2013 Researcher

3 Drinking water samples 18 2013 Researcher

4 Water supply

characteristics 36 sub-districts 2011 Census of India, Govt. of India [12]

5 Monthly rainfall 36 sub-districts 1998-2013 Dept. of Agriculture, Govt. of Maharashtra [13]

6 Monthly rainfall 39 sub-districts 1951-2012 India Meteorological Department, Pune

7 Monthly dam storage 3 irrigation

projects 2009-2013 Irrigation Dept. Govt. of Maharashtra [14]

8 Groundwater levels (pre-

& post- monsoon) 35 sub-districts 2005-2013 Central Ground Water Board, Govt.

of India [15]

9 Groundwater quality

data 7 locations 2011-12 Maharashtra Pollution Control

Board [16]

10 Irrigated area Upper Bhima

Catchment 2008 FAO [82]

3.2.3 Data analysis

The Standardized Precipitation Index (SPI) developed by McKee et al. [55] is used for the characterization of historical drought events. The assessment of water scarcity is done through a pilot study in the Bhima Catchment by using sustainability indicators proposed by the United Nations Organization for Education, Science and Culture (UNESCO) [84]. The detailed description of water scarcity assessment and drought characterization is given in chapter 4.

The primary data was processed and statistically analyzed using PASW SPSS 18. Various responses from farmers to open ended question were coded under similar answers of a question with coding 1 for “affirmative response” and 0 for “no answer”, to speed up data entry into SPSS. The affirmative response is expressed in percent. The 5 point Likert scale (very less, less, medium, high and very high) was used to get farmers’ response to the various close ended questions and coded 1 for “very less” to 5 for “very high”. The aggregate reliability of drought impact severity, Likert-type scale items was confirmed by Cronbach’s alpha for 20 items scale α = 0.80 (the acceptable limit is > 0.70). However, in case of Likert -type scale items for adaptations (6 items) and level of satisfaction from government mitigation measures (4 items), Cronbach’s alpha was 0.5 to 0.7. It is due to the reason that, with the few scale items (less than ten), it is common to get low Cronbach’s alpha (α = 0.5).

29

In this case, to confirm the reliability, mean inter-item correlation of these items was obtained 0.25 and 0.35 respectively (an optimal range 0.20 to 0.40 is acceptable) [85].

Descriptive and inferential statistics have been used to assess farmers’ perception of various drought impacts, coping/adaptive strategies being practiced to mitigate the effects of drought and government level administrative mitigation activities as relief measures (qualitative terms). To analyze the difference in perception of respondents, in addition to grouping based on sub-district wise irrigation strata [Low (<15%)-, medium (15-30%) - and high (>30%) - irrigated area of total cultivable area], individual farming system [rainfed, irrigated or mixed (rainfed and irrigation both) practices], population is grouped based on their land holding size [households with (marginal (<1ha), small (1-2ha), medium (2-4ha), large (>4ha) land holdings], annual household income [low (<INR 45,000 or US$828)-, middle (between INR45,000 or US$791 and INR180,000 or US$3,313)- and high (> INR 180,000 or US$3,313)] [86], education (illiterate-, primary-, secondary- and higher- education) and drought intensity (severe- and moderate- drought) faced. All these groupings are standards that are commonly used except irrigation strata. Data were analyzed using non-parametric significance testing, Wilcoxon Signed Rank test (NY and DY comparison), Kruskal-Wallis H-test (for comparison of 3 or more groups with multiple variables) and Mann-Whitney U-test (for comparison of two groups with multiple variables) at 5%

significance level [85, 87, 88, 89, 90, 91].

To analyze the drought impacts on domestic water supply, agricultural production, employment opportunity for unskilled laborers and financial status of the farmers (in quantitative terms), descriptive and inferential statistics have been used. The Shapiro-Wilk test as a numerical means of assessing normality is used for testing normality of continuous variables. It is found that the continuous variables used for analysis violated the assumption of normality, hence nonparametric test statistics were used for the analysis. The Wilcoxon signed ranks test (non-parametric counterpart to dependent t-test) is used for normal and drought year comparison of continuous variables (for example- time spent per trip to fetch water from the source, normal and drought year crop production etc.). The Man-Whitney U test (nonparametric counterpart to independent t-test) and Kruskal-Wallis H test (nonparametric counterpart to one way ANOVA) are used for comparison of continuous variables within two groups and more than 2 groups that may have different sample sizes respectively [85]. In addition to this, Jonckheere’s trend test is applied to find out the trend of significantly different cases against various groups. Results of various nonparametric tests were reported by using degrees of freedom, test statistics, p values and effect size r [85].

Drought impacts across different irrigation extent, household size and income, land holdings size, and drought intensity were analyzed by dividing respondent population into several groups as mentioned earlier.

30

Total 18 drinkable pot water samples were analyzed in laboratory with ion chromatography to determine the major anions and cations concentrations and compared with WHO drinking water quality standards. From qualitative and quantitative data analysis the drought vulnerability factors identified. The detailed description to obtain drought hazard, vulnerability and risk indices for sub-district of major district in the vicinity of catchment are given in chapter 8.