Retired farmland recovery in the Borrego Springs Subbasin: Supporting data for: Disturbance legacies in arid environments shape recovery of former agricultural land
Data files
Apr 23, 2026 version files 117.83 KB
-
LandscapeFunctionalAnalysis.csv
4.74 KB
-
LandUseHistory_Salinity.csv
1.34 KB
-
README.md
8.69 KB
-
Task2_Species_list.csv
5.69 KB
-
Task2_Vegetation_transects.csv
97.08 KB
-
WindRows_DistancetoNaturalLand.csv
285 B
Abstract
Increasingly, arid agricultural land is being permanently taken out of production due to more frequent drought and higher temperatures, alongside the overconsumption of groundwater. We investigated a chronosequence of time-since-cultivation of former agricultural sites in the Colorado desert to assess the degree and trajectory of recovery of vegetation community and soil characteristics and to identify key environmental drivers that shape this process. Recovery timescales varied widely for different variables; soil salinity was estimated to recover to a non-saline level within approximately 22 years, while native species richness could require an estimated 130 years to approach reference conditions. The primary drivers of recovery included a decline in soil salinity and a corresponding decline in non-native species richness, and the influence of episodic water flow, which enhanced landscape complexity and supported perennial vegetation establishment. These findings emphasize the slow pace of natural recovery in arid systems and the benefit of active restoration to overcome persistent abiotic barriers on former agricultural land to hasten recovery.
Dataset DOI: 10.5061/dryad.73n5tb3c0
Description of the data and file structure
These data were collected to assess the recovery trajectory of retired agricultural land in Borrego Springs, CA, USA located in the Colorado desert.
Files and variables
--Site-Level Data--
Methods
To assess salinity, we collected two 100 - 200 g surface soil samples to a depth of 10 cm across patches and interpatches identified on the gradsects. Soils were dried, sieved to 2 mm to separate any gravel or organic matter, and weighed. To determine soil texture, subsamples of soil were measured for particle size using the hydrometer method to determine the percent of clay, silt, and sand (Bouyoucos, 1962). Subsamples of soil were then measured for electrical conductivity (EC) using a YSI Pro Quatro conductivity probe (YSI Incorporated; Yellow Springs, OH, USA) to estimate soil salinity using the EC 1:5 (one part soil to five parts water by weight) extract methods. We then converted EC 1:5 to EC extract (ECe) using soil texture conversion factors (Department of Primary Industries and Regional Development, 2024). To estimate salinity at the site level, we calculated a weighted average based on patch or interpatch cover (see Landscape Functional Analysis below). Site-level salinity values were categorized as non-saline (0 - 2.0 dS/m), slightly saline (2.1-4.0 dS/m) moderately saline (4.1 - 8.0 dS/m), and strongly saline (8.1 - 16.0 dS/m) (Smith & Doran, 1996).
File: LandUseHistory_Salinity.csv
Description:
Variables
- Site: site identifier
- LandUseHistory: whether a site was formerly agricultural or an unfarmed reference
- Latitude: Latitude
- Longitude: Longitude
- EC_1_5_mS.m: EC 1:5 (one part soil to five parts water by weight) extract method for electrical conductivity. Units: millisiemens/meter
- ECe_dS.m: converted EC 1:5 to EC extract (ECe) using soil texture conversion factors. Units: decisiemens/meter
- ECe_interpretation: salinity interpretation (see methods for classes)
--Vegetation Composition--
Methods
Along each gradsect we measured plant community composition using the point-intercept method. A 1.3 cm diameter dowel was placed vertically each meter along the gradsect and any intercepting plant species touching the dowel were recorded. To measure species richness, all species encountered 1 meter to either side of the meter tape were recorded along the length of the gradsect to form a 2 m x 40 m belt transect. Species only detected in the belt transect and not the point-intercept sampling were termed “present” and not included in analyses other than richness. Where annual plants were identified at the species level, but dead or dormant, these species were included in the vegetation analyses to account for sampling during a single season.
File: Task2_Species_list.csv
Description: Species list to accompany the vegetation transect data.
Variables
- Code: 6-letter species code
- Species: Latin name of the species
- Native_NonNative: Origin of the species
- Life_Form: forb, shrub, cactus, graminoid
- Life_Cycle: annual, perennial, NA (not relevant because shrub or unknown if species identified to genus only)
File: Task2_Vegetation_transects.csv
Description: Vegetation transect data
Empty cell description: Empty cells found in columns Hit2, Hit3, Hit4 or Hit5 indicate that only one hit (Hit1) was recorded at this point-intercept. Empty cells are found in the Present column because these data are collected at the transect-level during the belt transect rather than at each individual point-intercept. If there are no values indicated in the Present column for a given transect, no additional species were detected in the belt transect other than those counted during the point-intercept sampling.
Variables
- Site: site identifier
- Date: date of data collection
- Gradsect: three gradsects per plot
- Point: data collected at 40 points (0 - 39)
- Hit1: the top hit on the dowel
- Hit2: the second hit on the dowel
- Hit3: the third hit on the dowel
- Hit4: the fourth hit on the dowel
- Hit5: the fifth hit on the dowel
- SoilSurface: the type of groundcover
- Present: if a species was not hit during the transect but was present in a 1 m belt on either side of the transect
--Landscape Function Analysis–
Methods
We used LFA to assess landscape complexity (Tongway & Hindley, 2004). The core of the methodology involves the classification of landscape elements into two main categories: patches, representing long-lasting landscape elements like perennial vegetation, litter piles from perennial vegetation, and topographic features that accumulate resources (such as sediment, water, litter, seed), and interpatches, representing various types of bareground or seasonal herbaceous covers that are susceptible to resource loss. Along each gradsect, we noted the type and length of patches and interpatches and additionally measured the height of patches. We categorized patches by the dominant cover type (i.e., mounds, shrubs, shrub mounds, cactus, cactus mounds, litter patches, and litter mounds). Mounds were defined as an increase in soil height of at least 10 cm. Detailed methodology can be found in Tongway and Hindley (2004).
File: LandscapeFunctionalAnalysis.csv
Description:
Variables
- Site: site identifier
- Patch_identity: type of patch
- Gradsect: three gradsects per plot
- Distance_occcupied_meters: the patch length in m
- Patch_Height_cm: the height of the patch in cm
- Patch_length_cm: the length of the patch in cm
- Distance_on_tape_meters: the locations on the tape measure that bound the length of the patch
- Percent_cover_on_gradsect: percent cover of the gradsect accounted for by this patch
--Agricultural Sites Only--
Methods
To assess the role of the distance to intact natural land, we calculated the shortest distance from the edge of each 100 m x 100 m former agricultural site to the nearest intact natural land. The intact natural land was identified in the previously described aerial imagery exercise for site selection and included private, never-developed land and Anza-Borrego Desert State Park land (Fig. S1b). We restricted the distance to intact natural land such that only land in the prevailing wind direction was considered (northwest: 270° to 360°). The distance of former agricultural sites from natural land ranged from 300 m to 525 m.
To test the role of windbreaks, which could mitigate scouring, we recorded a site-level binary variable that indicated whether a windbreak was present in the direction of the prevailing wind (northwest: 270° to 360°.
File: WindRows_DistancetoNaturalLand.csv
Description: For agricultural sites, the presence of windrows and the distance to natural land
Variables
- Site: site identifier
- Windrow: binary yes (Y) or no (N) for windrow presence
- Dist_natural_land_m: the distance to intact natural lands. Unit: meters
- Yr_fallowed: the year the site was taken out of irrigated production
- Yr_since_fallowing: the years between the fallowing time and field sampling
References
Bouyoucos, G. J. (1962). Hydrometer Method Improved for Making Particle Size Analyses of Soils 1. Agronomy Journal, 54(5), 464–465.
Department of Primary Industries and Regional Development. (2024). Measuring soil salinity. https://library.dpird.wa.gov.au/nrm_factsheets/26
Smith, J. L., & Doran, J. W. (1996). Measurement and use of pH and electrical conductivity for soil quality analysis. In Methods for Assessing Soil Quality (pp. 169–185). Wiley. https://doi.org/10.2136/sssaspecpub49.c10
Tongway, D., & Hindley, N. (2004). Landscape function analysis: a system for monitoring rangeland function. African Journal of Range & Forage Science, 21(2), 109–113. https://doi.org/10.2989/10220110409485841
Code/software
All statistical analyses and visualizations were performed in R ver. 4.3.3 (R Core Team, 2024).
R Core Team. (2024). R: a language and environment for statistical computing. R Foundation for Statistical Computing.
Access information
Other publicly accessible locations of the data:
- NA
Data was derived from the following sources:
- NA
We established seven former agricultural and seven unfarmed reference sites in the northern agricultural area of the Borrego Springs Subbasin. We used historical imagery to assess the land use status between 1953 (the earliest date of aerial imagery) and 2022 in order to assign a category of developed, formerly agriculture, presently agriculture, or intact natural land to nearly 500 parcels in Borrego Springs. Land that was found to be former agricultural land served as the pool from which we selected former agricultural sites. We aimed to select sites that span a range of times since cultivation and the latitude and longitude of the agricultural area, but we were somewhat limited by access constraints. Using aerial imagery and community knowledge, we were able to assign a general crop type to all sites. While most sites cultivated citrus, our two oldest sites cultivated grapes. The second oldest site additionally cultivated an herbaceous crop until 1990 following the cessation of grapes in 1968. Grapes generally have a lower irrigation requirement than citrus crops and thus our oldest sites were likely less intensively cultivated, providing a more conservative benchmark. Parcels that were not developed between 1953 and 2022, including privately owned land and Anza-Borrego Desert State Park, were considered intact natural land and served as the pool from which to select unfarmed reference sites.
Former agricultural sites were taken out of production between three and 57 years ago. Each site consisted of a 100 m x 100 m area in which we established three 40-m gradient-directed transects (gradsects). Gradsects are line transects which are oriented in the direction of a primary abiotic force of interest, which in this case is the prevailing wind. We oriented our gradsects using wind data (Iowa Environmental Mesonet, https://mesonet.agron.iastate.edu/sites/locate.php?network=CA_ASOS) and by assessing evidence of dominant wind direction in the environment, such as sand ripples, erosion abrasions on branches and the bases of stems, and the shape of sand hummocks in each site. The orientation of the gradsects ranged from 300° to 330° azimuth.