2015 Drought soil biogeochemistry and greenhouse gas emissions study at El Verde



Short name: 


Data set ID: 



We report the effects of the severe 2015 Caribbean drought on soil moisture, oxygen (O2), temperature, phosphorus (P), iron (Fe), pH, and GHG emissions (CO2 and CH4) across a catena sensor array field outside of El Verde Research Station, Luquillo LTER, Puerto Rico. Seven sensors of each type were installed at 12 cm depth along a ridge to valley catena; the entire catena transect was replicated five times for a total of 105 sensors. Within the sensor field we also installed nine automated gas flux chambers randomly located in each topographic zone (ridge, slope and valley). Soil carbon and nitrogen, extractable phosphorus (P) pools, iron (Fe) species, and pH were sampled before and during the drought as indicators of biogeochemical conditions.
Date Range: 
2014-11-14 00:00:00 to 2016-02-24 00:00:00

Publication Date: 

2018-02-28 00:00:00

Additional Project roles: 

Name: Leilei Ruan Role: Associated Researcher


Field location information

         Research was conducted in the Luquillo Experimental Forest (LEF), Puerto Rico, USA (Lat. 18° 18’ N; Long. 65° 50’ W).  The forest is congruent with El Yunque National Forest managed by the US Forest Service. The LEF contains approximately 11,500 ha of contiguous forest area, spanning an elevation gradient from approximately 350 to 1075 m above sea level.  The LEF has been well characterized geologically and ecologically as part of on-going Long-Term Ecological Research and Critical Zone Observatory projects6,26,55-57.  Soils in the LEF are derived from volcanoclastic sediments with quartz diorite intrusions58.

         El Verde Research Station, where this research took place, is located at ~350 m a.s.l. elevation.  Mean monthly temperatures range from 20.6 °C to 25.8 °C with an annual mean temperature of 23.0 ± 1.9 °C (means derived from 1975-2004 temperature record)59.  The forest can be classified as subtropical wet forest and the plant community is mature tabonuco (D. excelsa Vahl) forest60.  Soils at the field site are clay-rich Ultisols (Supplementary Tables 1 and 2) derived from volcanoclastic parent material.

Experimental design

         An automated sensor array was installed near the El Verde Research Station (Lat. 18° 32’ N; Long. 65° 82’ W). The array was composed of five replicate transects, each with seven topographic locations from ridge to valley; the sensor transects were located about 2.5 m apart over a 50 by 50 m plot on a slope of angle 25° over 5 m vertical distance (overall slope angle = 25°, steepest area (upper slope) angle = 55°; both measured using a standard field clinometer). The location was chosen to be generally representative of the vegetation, soils, and topographic variability of the larger watershed ecosystem55.

         Each of the 35 measurement points contained a sensor cluster with a galvanic O2 sensor (Apogee Instruments, Logan, UT, USA) and a time-domain reflectometry sensor (Campbell Scientific, Logan, UT, USA), both of which were installed in the top 15 cm of soil (Supplementary Fig. 2).  The moisture, temperature and O2 sensor clusters were installed at a random location within each of the seven zones along the catena; sensors were within approximately 5 to 10 cm of each other as root and rock placement allowed.  Nine automated gas flux chambers (Eosense, Nova Scotia, Canada) were installed within the sensor array; three chambers each were distributed randomly within ridge, slope or valley zones (Supplementary Fig. 2).  The chamber sites corresponded to the relatively flat ridge top, the mid slope, and the valley bottom of the sensor array transects. Continuous sensor measurements began in November 2014.  Trace gas measurements began in February 2015.

Soil moisture/oxygen and rainfall measurements

         Daily mean soil O2 and moisture measurements were compared with precipitation patterns.  Soil O2 sensors were installed in the top 15 cm of soil in gas-permeable soil equilibration chambers (295 mL, 5 cm diameter, 15 cm height) (sensu Liptzin et al. 201156) at each of the 35 sensor locations in the topographic array.  Data from these sensors were collected hourly using data loggers (Campbell Scientific, Logan, UT, USA) and multiplexers (Campbell Scientific, Logan, UT, USA).  Precipitation data was collected at a nearby rain gauge located at El Verde Research Station, approximately 500 m from the field site.  This gauge is administered by the Luquillo LTER as part of the long-term on-going climate monitoring program in the LEF.  Rainfall has been recorded daily or semi-daily since 1964 and this record was used to report precipitation during the study period as well as historical precipitation patterns (historical data catalogued on the Luquillo LTER datanet61).

Gas flux measurements

         To determine patterns in trace gas fluxes across the soil atmosphere interface we used 9 automated surface flux chambers deployed within in the sensor network plots (3 ridge, 3 slope, 3 valley; Supplementary Fig. 2, Supplementary Table 3, Supplementary Data 2).  Automated flux chambers were connected to a multiplexer, which dynamically signaled chamber deployment and routed gases to a Cavity Ring-Down Spectroscopy (CRDS) gas analyzer (Picarro, Santa Clara, CA, USA)62.  The automated chambers, multiplexer and CRDS gas analyzer were powered by a generator (Honda, Tokyo, Japan) with the generator, multiplexer and CRDS gas analyzer housed in a shed away from the array to avoid contamination from generator exhaust.  When chambers were not measuring a flux, lids were not in contact with the chamber base, and instead were held approximately 10 cm above and 5 cm outside of the chamber base circumference, in order to minimize impact on the sampling area and ensure that precipitation would reach soil within the chamber footprint.  Chambers were closed for a 10-minute sampling period with a 3-minute flushing period between chamber measurements.

         Trace gases were measured from one chamber at a time; a full cycle through the 9 chambers took approximately two hours and occurred continuously leading to a maximum of 12 measurements per chamber per day.  Sampling occurred daily unless instrument malfunction prevented sampling, which occurred because of instrument failure or debris (branches or large leaves) inhibiting chamber closure. The array was inspected at least twice each day to minimize these events.  Instrument related gaps in the data record are associated with instances in which the multiplexer, CRDS gas analyzer or generator needed repairs (Supplementary Fig. 3).  Automated chambers and the CRDS gas analyzer also recorded chamber and instrument temperatures, relative humidity values and pressure during the flux measurements.  Fluxes recorded during periods of high values of chamber and instrument temperatures, relative humidity values and pressure were removed from the data record.  After data cleaning and accounting for days in which data could not be collected, 6,479 CO2 flux observations and 6,379 CH4 flux observations remained from 150 unique days (out of 326 possible sampling days).

         Fluxes of CO2 and CH4 were calculated using software developed to work in tandem with an automated chamber-CRDS gas analyzer set up (Eosense EosAnalyze-AC v. 3.4.2).  For each measurement, two flux rates were calculated, one using a linear model and one using an exponential model sensu Creelman et al. 201363.

         Dataset quality assessment and control was subsequently performed in R (R v. 3.2.2).  Fluxes were removed from the final dataset if they were associated with anomalous temperature, moisture, or pressure readings, if the initial concentrations of CO2 or CH4 were substantially higher or lower than ambient values (potentially indicative of a malfunctioning flush period) or if the chamber deployment period was less than 9 minutes or more than 11 minutes.  The choice between the linear and exponential flux rate models was decided upon using the estimate uncertainty to estimate ratios, and in cases where both the linear and exponential models produced high uncertainty, the flux was eliminated from the dataset. Detailed results of all GHG fluxes are provided in Tables S3 and S5.

Soil variable sampling and processing

We sampled soils from ridge, slope and valley locations within the catena from 0-15 cm depth before the onset of the drought (April 2015) and during the height of the drought (July 2015).  Four replicate samples were taken from each topographic zone at each sampling time point.  Soils were transported to Berkeley, CA, USA and processed within 5 days of collection for pH and concentrations of Fe(II), Fe(III), organic P and inorganic P.

         We performed 0.5 mol/L HCl extractions on 5 g of wet soil, which were analyzed for concentrations of total Fe (i.e., Fe(II) + Fe(III)) and Fe(II) on a spectrophotometer (Thermo Scientific Genesys 20, Fisher Thermo Scientific) (sensu Liptzin and Silver 200964).  Fe(II) concentrations were measured directly while Fe(III) concentrations were calculated as the different between total Fe and Fe(II) concentrations. Soil pH was measured on samples of 1.5 g of wet soil in DI water using a pH probe (Dever Instrument Ultrabasic pH/mv Meter (UB-10)).

         We performed a Hedley phosphate extraction with two extraction steps65.  We first extracted 1.5 g of wet soil in 0.5 mol/L sodium bicarbonate to measure the concentration of organic phosphorus.  We followed this extraction with a 0.1 mol/L sodium hydroxide extraction to measure the concentration of inorganic phosphorus.  The extract was measured on a spectrophotometer (Thermo Scientific Genesys 20, Fisher Thermo Scientific).

         Several one-time soil variables were measured, in all cases with four replicate samples taken for each topographic location (ridge, slope, valley).  Soil bulk density was measured in August 2016 using standard volume cores (height 10 cm, diameter 6 cm) that were pounded into the soil, surrounded by an outer core to prevent soil compaction.  Samples were oven dried at 105 oC for 72 hours and then weighed.  Air dried soil was analyzed for percent C and percent N using an elemental analyzer (CE Elantec, Lakewood, NJ) in December 2016.  Soil texture was analyzed using the hydrometer method66 in April 2015. Detailed results of all soil variables are provided in Supplementary Table 4.


Seven sensors array: soil moisture, oxygen (O2), temperature, phosphorus (P), iron (Fe), pH, and GHG emissions (CO2 and CH4)



Time Period: