Impact, Maintenance, and Sustainability of Irrigation Impact Evaluation Survey 2015-2018, Baseline, First, Second, Third Follow-up

For three of the sites that are being used for the spatial regression discontinuity analysis (K12, N23, and R34), we divided the site into 3 areas - CA buffer (BCA), CA Catchment buffer (BCAC), and CA terraces (TCA). BCA is the area inside of the CA (CA, below main canal) within 50m of the boundary of the CA. BCAC is the area in the CA Catchment (CAC, above main canal) within 50m of the boundary of the CA. TCA is the terraced farmland that is in the CA, but more than 50m from the boundary of the CA. The third site will be used for the within-CA experimental designs only, and as a result we focused our sampling in one area - the CA terraces (TCA).

We constructed our household sampling by dropping a uniform grid of points across the full site at 2-meter resolution, and then sampling points within the grid. After each point was sampled, we excluded any points within 10m of that point (to keep from selecting multiple points too close together).
Enumerators were then given GPS devices with the locations of the points, and sent to each point, with a key informant (often the village leader). For each point, they were asked to identify if the point was on cultivable land (this was to discard forest, swamps, thick bushes, bodies of water, or other terrain which would make cultivation impossible). They were asked to record, for points in cultivable land, in SurveyCTO, the following:

1. The name of the point visited (which was displayed on the GPS);
2. The name of the cultivator, the location of their residence, and their phone number;
3. A description of the plot detailed enough that the cultivator would be able to identify the exact plot described

Additionally, they were asked to save their GPS track at the end of the day, as a way of tracking the number of hours they spent checking points and to verify that they visited each point. We used the data from this listing to construct a roster of all the unique names of cultivators, clustering points together when the names seemed identical. This roster (which contained the name of the individual, their village and phone number, the descriptions of the plots, and the villages in which the plots were located (identified using village shapefiles) and were organized by village) were then used to contact village leaders and verify that the listed individuals in fact existed. Multiple follow-ups were sometimes needed when village leaders suggested that one individual lived in a different village, or multiple village leaders said an individual lived in their village.

Finally, a sample plot was selected for each verified 2689 households. To select this sample plot, one point was randomly selected for each household. The probability of selecting a particular point was weighted - a weight of 1 was assigned to points in the BCA and BCAC, and a different weight was assigned for points in the TCA, to balance the number of sample plots in these areas.