Quality assurance of water levels in the Mississippi River Valley alluvial aquifer in Arkansas from the Arkansas Department of Agriculture Natural Resources Division

This data release provides several data files representing groundwater levels reported through driller's reports for the State of Arkansas Arkansas Department of Agriculture National Resources Division (https://wise.er.usgs.gov/driller_db/) (accessed January 10, 2024) within or near the Mississippi Alluvial Plain (MAP) and (or) associated with the Mississippi River Valley alluvial aquifer (MRVA). First, a retrieval of data from the State of Arkansas was made and manual preparatory filtering including complete information of location, date, water level (depth below land surface), well depth in feet, and general association with the MAP or MRVA. The land-surface datum for each record was derived from a 10-meter digital elvation model (DEM) based on decimal latitude and longitude (U.S. Geological Survey, 2017). Further manual inspection by software functionality was made to confirm that the data were not already within the U.S. Geological Survey (USGS) National Water Information System (NWIS) (U.S. Geological Survey, 2023). The agency code for the water levels has been assigned "AR008" (Arkansas Natural Resources Commission) in accordance with the https://help.waterdata.usgs.gov/codes-and-parameters/code/agency_cd_que… (accessed January 10, 2024). Use of the AR008 agency code is consistent with historical and current USGS storage practices in NWIS when in collaboration with the State of Arkansas. This first data file is titled "ARNRC_drillers_working.csv" (33,380 records) including a well identification of "123413204978" that occurs 2,785 times at locations spread throughout the MAP in Arkansas. Another 8 records lack a well identification number; these records were removed. The decimal latitude and longitude for each record in ARNRC_drillers_working.csv resided within the MAP boundary (Painter and Westernman, 2018). Second, the records with the "123413204978" identification are judged without further study as erroneous or spurious database records and were removed. Third, the intermediate data file (30,587 records) was then processed through data structure conversion software (infoGW2visGWDB) (Asquith and Seanor, 2019). The resultant but transient data structure of 30,587 records was given over to quality-control and assurance using statistical modeling (visGWDBmrva software) (Asquith and others, 2019, 2020). Some 394 records by horizontal position having a missing altitude of the bottom of the MRVA and 90 records having water-level altitudes below the bottom of the MRVA when digitally working with the Torak and Painter (2019) surface of the MRVA bottom. These 394 + 90 = 484 records are retained through the workflow described herein to avoid potential scientific interpretation of hydrogeologic framework. The statistical analyses result in formation of a regional statistical time-series models using generalized additive models (GAMs) and support vector machines (SVMs). In summary, for each of the 30,587 well-water-level records (or rather in detail, each unique well identifier), the visGWDBmrva software isolated all water levels for the MAP/MRVA from USGS (2023) within 16 kilometers radial distance. This means that the driller's dataset is being internally compared to itself and USGS MAP/MRVA data. The visGWDBmrva software computed a "pseudo water level" from a blending of GAM and SVM model predictions for the date of the driller's recorded water level. These computations are all created on-the-fly. A residual was computed from the pseudo water level (as altitude) to that water-level altitude reported for the well-water-level record of the driller's dataset. These statistical results are listed the file titled "ANRC_retained_levels.csv" (28,964 records) for which records were retained ARNRC_drillers_working.csv if the absolute value of the residual of the well-water-level record and the pseudo water level was less than or equal to 20 feet. This threshold resulted from exploratory review of the statistical computations and is consistent with Smith and others (2020) and Weber and others (2021) for a similar driller's reported dataset for the Missouri part of the MAP/MRVA. The results listed in file ARNRC_retained_levels.csv are deemed especially suitable for greater statistical modeling of groundwater levels in the MRVA (Asquith and Killian, 2022).