We also discovered that both water source and crop questions needed to be simplified. We achieved this by only asking for water source rank and name, which could be used to search external data for total dissolved solids, water price, and percent of supplier revenue generated from selling water to agricultural consumers. We simplified the crop question to only ask for the top 3 highest value crops, based on feedback from survey respondents. The final survey consisted of 28 questions . Questions associated with farm, farmer and water source characteristics were straightforward to construct , and we used either fill-in or close-ended questions. Water management practices required reliance on previous surveys and help from agricultural extension and resource conservation experts.Two questions on water scarcity perceptions were also included . The original data collection method was an on-line survey via the Survey Monkey platform that would be disseminated by the respective Agricultural Extension agencies or Farm Bureaus for the four counties. The greatest strength of an on-line vs. mail-in survey is the time saving both in disseminating the survey and inputting/coding data. One also has the ability to “force” responses for the most critical data, thus reducing incomplete surveys . The Agricultural Extension agencies, Imperial County in particular, were hesitant to host a survey on their websites. Ultimately, the Farm Bureaus in Riverside and San Diego counties hosted our survey. After being hosted on the Riverside Farm Bureau site for one month, the survey only received two complete responses . San Diego Farm Bureau did not directly host the survey on their website,vertical grow rack system but agreed to disseminate the survey via an email newsletter. This yielded two responses after one month. The pilot mail-in survey had a higher response rate than the on-line final version. Thus, we made a decision to implement a mail-in survey using the same dataset as from our pilot survey.
We mailed the survey to 1277 potential respondents in a staggered sequence following the timeline in Figure 3.1. We mailed the entire list in Imperial and Riverside counties due to the relatively smaller number of growers. For San Diego and Ventura counties, we used a random number generator to randomly select 300 recipients from each county. The mailing packets contained: invitation letter, informed consent, survey, and self-addressed return envelope. We also offered a $25 incentive if we were to receive the survey by the two-month deadline stipulated in the informed consent. We had a dedicated team of three undergraduate students to assist with the initial and follow-up mailings. These students later assisted with data entry. We received 221 responses, of which 187 were valid, resulting in a 14.6% response rate. The observations in the original dataset, used for the pilot and final surveys, were obtained from the County Agricultural Commissioners’ Offices in Imperial, Riverside, San Diego, and Ventura counties via a Public Information Request Form. There is no formal name for this dataset, though it is informally called “Pesticide Permit Data”. It represents agricultural entities who applied for Restricted Materials Permits at a given time. Conservative estimates suggest that these data represent roughly 75% of growers in each county . We used 2014 as our reference year, because we wanted to ask growers about the prior year when we started our analysis in mid-2015. The dataset includes contact information, location, and commodity information at the agricultural field level for each grower. Importantly, these excel data are linked to geospatial data. These data are not linked to email addresses, preventing us from using the dataset in an on-line survey. As discussed earlier, our primary reason for abandoning on-line survey dissemination was the poor response rate. Additionally, the Agricultural Commissioner has GIS field boundaries for each farm, which allow us to accurately link land value, climate, and soil to specific agricultural fields. This will subsequently be explained in more detail.
Riverside County is relatively more complex in agricultural water distribution, where two districts hold senior water rights to the Colorado River via the Seven Party Agreement. Other districts represented in the survey include Western Municipal Water District, Rancho California Water District, Riverside Public Utilities, Eastern Municipal Water District , and Lake Hemet Municipal Water District. Though it is not considered a water agency, Gage Canal also represents an important source of agricultural water from the Santa Ana River for citrus growers in the city of Riverside.There are 3 major wholesalers in the county: United Conservation District , Casitas Municipal Water District , and Calleguas Municipal Water District . Ventura County also has a proliferation of smaller entities that are either Mutual Water Companies or Private Water Companies. The former are commonly owned by their shareholders, while such ownership is not necessary in the latter. Mutual Water Companies are loosely regulated by the Public Utilities Commission, while Private Water Companies are not. The majority of mutual water companies receive water from the nearest groundwater basin , thus we derive total dissolved solids information based on the nearest groundwater well, as explained in the proceeding section on groundwater data. We still code mutual water companies as water districts because, even though these are smaller than the other water districts represented in this study, they still represent institutions with governing rules for members . Mutual water companies represented in our survey include Farmers Irrigation, Del Norte Mutual Water Company, Fillmore Irrigation Company, Crestview Mutual Water Company, South side Improvement Company, and La Loma Ranch Mutual Water Company. In addition to wholesalers and mutual/private water companies, mid-sized water districts also operate in Ventura County. Those represented in our survey include Camrosa Water District, Ventura County District 1, Ventura County District 19, and the City of Simi Valley. Total dissolved solids in ppm is used as a measure of water quality, where higher tds values imply lower water quality. It is calculated in ArcMap using USGS Groundwater Ambient Monitoring and Assessment reports for Riverside, San Diego, and Ventura counties 15.
Maps of sample wells from these reports were converted to ArcMap documents, and TDS data were linked to each sample well. The centroid location of respondents using groundwater as their primary water source was linked to the sample well maps using inverse distance weighting. Inverse distance weighting is a spatial interpolation technique in ArcMap that averages the values in the neighborhood of each data point, giving a decreasing weight as distance increases. This resulted in a given respondent’s TDS value equal to a weighted average of surrounding sample wells represented in the USGS data . Following the practice of previous hedonic property studies in California, climate data were obtained from the Parameter-elevation Regressions on Independent Slopes Model Group housed in Oregon State University . These are spatially interpolated datasets collected from a range of climate monitoring networks, which provide more intra-county variability than the limited number of California Irrigation Monitoring Information System weather stations18 . We used a 30-year average monthly climate normals for maximum daily temperature and minimum daily temperature measured in Celsius. To clarify, tmax implies the monthly average of the daily maximum temperature,clone rack and not the maximum temperature recorded each month. We also include the 30-year normal for precipitation , which represents the total monthly rainfall and snow melt in millimeters averaged over the 30-year period. 19 GIS files of the national climate data were downloaded at a resolution of 800m, and clipped to represent the 4-county region of analysis. The polygon data in the Agricultural Commissioner files was first converted to points. Using the Extraction tool in ArcMap, climate data values were assigned to these points. Due to limited “within field” variability of climate variables, the centroid point value was assigned to each field. We then took a weighted average of the field-level data in order to determine the farm level climate normal. Constructing a complex dataset is a time-consuming endeavor. Using a paper survey compounded this challenge as we manually input all survey data. Some questionnaires were only partially complete when mailed back to us. Some growers were not producing commercially, others were coded as tree crops when in reality these were nursery trees. And, we ultimately excluded all nurseries from our dataset because these do not rely on the soil or, in the case of indoor nurseries, on climate, in the same way as conventional, open-space crops. In addition to missing or irrelevant data from the questionnaires, the spatial datasets also created a few challenges. First, the spatial data from the Agricultural Commissioner on field boundaries was quite noisy, including multiple permit years with slightly different acreages for identical crop fields. The same polygons were redrawn on top of one another in ArcMap several times. We corrected for this by selecting 2014 as the reference permit year. Even after doing this, several growers who were in the 2014 excel file originally requested from the Agricultural Commissioner were missing 2014 spatial data. To correct for this, we used the next chronological year of spatial data. We experienced a similar challenge with the Assessor Data. Only Assessor Parcel Numbers were included in the spatial data, and not land values. Thus, we had to separately purchase land value data for agricultural parcels from the respective Assessor offices. These data also introduced missing values since not all APNs in the spatial data were available in the land value data.
Agricultural Commissioner spatial data for San Diego County presented the greatest challenge. Several growers in the original dataset were missing completely from the spatial data. We used the physical address of the farm/grove/vineyard to find the centroid latitude and longitude in Google Earth prior to geospatial analysis. Additionally, several respondents were not represented in the soil shape files. We supplemented this data with the Web Soil Survey as discussed in the soil moisture data section of this chapter. Overall, we addressed these data collection challenges in the most efficient and robust means at our disposal in order to minimize respondent attrition. We begin with our first of three empirical chapters. The Ricardian framework allows one to exploit cross-sectional differences in key production variables to approximate the marginal impact of climate, at a given time period. Previous studies have aggregated data at the county level, which involves strong simplifying assumptions of county homogeneity in farmer characteristics, water source, soil type, irrigation technology, and other key inputs. Perhaps most relevant for an analysis on climate impacts, such data aggregation assumes a homogeneous county climate. Using data from our farm questionnaire, we study the impacts of change in microclimate on agricultural productivity in Desert and Southern California regions while controlling for other micro-level effects such as grower, farm, and water source characteristics. We define microclimate as the average climate of all cropped fields on a given farm.20 Figure 4.1 illustrates our survey region and respondent distribution. What is the marginal impact of these micro-level variables on productivity after long-run adaptations by the grower are taken into account? To what extent do such variables, which provide more accurate representation of both the grower and farm, help explain the variation in gross revenue per acre in Desert and Southern California agriculture? We explore these questions using two, non-linear specifications of the farm-level value function with respect to climate: log-log and quadratic transformation. The Ricardian specification allows us to implicitly capture long-run production decisions, many of which are the result of complex historical relationships with local climate, water, and soil conditions. This is particularly true in Southern California where growers in the hottest regions tend to have access to senior water rights. Mendelsohn, Nordhaus and Shaw is one of the first studies to elegantly represent economic decision-making by the grower while analyzing the impacts of climate change on US agriculture.Cross-sectional differences in climate across 2933 US counties reveal the relative contribution of temperature and precipitation normals to farm productivity. Ultimately, these authors find that previous studies overestimate the loss in agricultural profits as they do not allow for reasonable adjustments to baseline production practices that incur substantial losses under a changing climate.Although previous Ricardian studies account for responsiveness to climatic change, most studies do not include characteristics of the decision-makers on farm, which are likely to directly influence productivity. This is not to say that socioeconomic variables are entirely excluded from earlier analyses.