31 lines
4.5 KiB
TeX
31 lines
4.5 KiB
TeX
\FloatBarrier
|
|
|
|
\section{Data}
|
|
The data used in the empirical analysis come primarily from \ac{CDSS}, through the HydroBase \ac{API} tools. This dataset is provided by the Colorado Division of Natural Resources and includes well attributes and pumping records, as available. \ac{SLV} wells were required by law to install pump monitors starting in 2009. This means that the pumping records for each well in the \ac{SLV} is available with two years of pre-pumping fee data available. Irrigated crop parcels are provided in geospatial files as part of this system, as well as \ac{CSV} files with the attributes of those parcels. The crop parcel is not a spatially fixed unit. Instead, it is defined as a contiguous piece of land using a unique crop type and irrigation technology combination in a given year\footnote{For example, if a particular legal parcel grows alfalfa on one half of the field and small grains in another the data set includes two different parcels for that year. If the legal parcel switches from flooding to sprinklers the next year, this too creates new crop parcel entries.}. However, wells are linked to crop parcels so the crop choice of wells can be assessed. There is some error in these estimates since the percentage of water applied to each crop is assumed to equal the crop distribution of the parcel.
|
|
|
|
Geospatial data is used to generate a distance matrix between each well, this is used to identify the distance of wells from those that serve parcels enrolled in \ac{CREP}. Maps of the \ac{SLV} were used to manually create geospatial shape files of the subdistrict boundaries. Overlaying this layer with the map of wells allows the subdistrict wells to be identified. The distance is used as a cutoff radius for determining if a well is a neighbor to \ac{CREP}, which is determined by the geology of the \ac{SLV}. \cref{FIG:MAP} provides a visual summary of the distances from \ac{CREP} calculated for each well.
|
|
|
|
%\FloatBarrier%
|
|
\begin{figure}
|
|
\centering
|
|
\includegraphics[width=0.85\textwidth]{CREP_DIS_MAP}
|
|
\caption{Distance of wells from an enrolled \ac{CREP} well}
|
|
\label{FIG:MAP}
|
|
\end{figure}
|
|
%\FloatBarrier%
|
|
|
|
A one-half mile distance was selected for the primary results. Previous analysis in Kansas had used a larger radius of two miles based on local geologic conditions \citep{rouhirad2021}. Using such a large distance is not feasible in the \ac{SLV} since nearly all subdistrict wells are within a two-mile radius. This could lead to downward bias in the neighborhood policy estimates since some of the \ac{CREP} spillover will be attributed to the general \ac{SBD1} effect. Mitigating this is the lower permeability and separate closed basin of the \ac{SLV}. Furthermore, any neighborhood effect should be strongest closer to the well. A robustness check running neighborhood effects at each considered radius is provided in appendix \cref{A:RAD}.
|
|
|
|
%\FloatBarrier
|
|
|
|
|
|
A list of wells tied to farms enrolled in \ac{CREP} is provided by the subdistrict, and the unique groundwater ID are used to link them with pumping data \citep{sbd12023}. The unit of investigation is the well, so well level fixed effects are included to remove confounding factors. In order to observe any potential selection issues, summary statistics for wells were acquired by linking permits in the \ac{CDSS} system to wells. These statistics are provided in \cref{SUMSTAT}.
|
|
|
|
%\FloatBarrier
|
|
\input{Tables/Summary_Stats_Wells.tex}
|
|
\FloatBarrier
|
|
|
|
Upon comparison, many attributes in \ac{CREP} and subdistrict wells are in the same range. The largest discrepancy is in the pumping rate, which is found by testing the capacity of the well. These tests are not uniform and can occur anywhere from before completion to months after operation. This contributes to the large standard deviation in the \ac{CREP} group. This may suggest that wells with higher operating costs were entered into \ac{CREP} before more productive wells.
|
|
|
|
Comparing subdistrict wells to other control wells suggests that the control group uses older and deeper wells, with larger production zones. This could create concerns that the control can respond asymmetrically upwards compared to the smaller subdistrict wells. However, due to the policy change bringing wells within the prior appropriations system, the legal well permits tend to create the pumping ceiling and not the maximum pumping rate. While there is still some potential for error, well level fixed effects will remove the effect of this capacity difference on the average, and such errors likely only occur in years with extremely high pumping rates.
|