Scenarios of Future Climate and Land-Management Effects on Carbon Stocks in Northern Patagonian Shrublands

Abstract

We analyzed the possible effects of grazing management and future climate change on carbon (C) stocks in soils of northern Patagonian shrublands. To this aim, we coupled the outputs of three (HadCM3, CSIRO Mk2, and CCSR/NIES) global climate models to the CENTURY (v5.3) model of terrestrial C balance. The CENTURY model was initialized with long-term field data on local biome physiognomy, seasonal phenologic trends, and prevailing land-management systems and was validated with recent sequences of 1-km Normalized Difference Vegetation Index (MODIS-Terra) images and soil C data. In the tested scenarios, the predicted climate changes would result in increased total C in soil organic matter (SOMTC). Maximum SOMTC under changed climate forcing would not differ significantly from that expected under baseline conditions (8 kg m⁻²). A decrease in grazing intensity would result in SOMTC increases of 11% to 12% even if climate changes did not occur. Climate change would account for SOMTC increases of 5% to 6%.

Appendix 1

Baseline Precipitation and Temperature Simulation Algorithm

A simulation algorithm was used in this study to generate expected monthly precipitation and air temperature series at the Patagonian Monte during the current century under the assumption that no climate change would occur. This represented a baseline condition for comparison of the various changes predicted by the AOGC models. The algorithm uses EPRA meteorologic station data gathered during 1900 to 2003 to infer the characteristics of their Probability Distribution Functions (PDFs.) Then the PDFs were convoluted to derive data realizations that mimic the observed time series.

Baseline monthly precipitation (P, mm) series were constructed (Equation 1) with the function:

$$ {P\,\, = \,a_{0} \, + \,a_{1} \,{\rm sin}\,(2\pi \,x/a_{3} - \,a_{2} )\, + \,a_{4} \,{\rm sin}\,(2\pi \,x/a_{6} - \,a_{5} ),} $$

(1)

(x = 1,.. 12: month no.), which was fitted to EPRA data. We wrote computer code to generate realizations of baseline P series by selecting normally distributed random values of the parameters a ₀ − a ₆ within their p ≤ 0.05 confidence intervals. Furthermore, the relative frequencies (f) of single monthly precipitation log values were fitted to the function:

$$ {f\, = \,a_{7} \, - \,a_{8} y\, + \,a_{9} \,exp ( - 0.5\,(y\, - \,a_{{10}} )/a_{{11}} )^{2} } $$

(2)

(y : frequency of ln (P) class in data), and expected stochastic baseline realizations of monthly precipitation were obtained by convoluting Equation 2 (Press and others 2004).

Baseline estimates of monthly minimum average air temperature (T _min ) were also based on EPRA data (1982 to 2002). Series were fitted to a single annual harmonic (Equation 3):

$$ {T_{{min}} \, = \,a_{{12}} \, + \,a_{{13}} \,{\rm sin}\,(2\pi x/a_{{14}} \, + \,a_{{15}} ).} $$

(3)

Maximum average monthly air temperature (T _max ) was then estimated based on the relation between the average minimum temperature (T _min ) and the monthly average amplitude (A = T _max − T _min ) (Equation 4):

$$ {T_{{max}} \, = \,10.88\,( \pm0.37)\, + \,0.10\,( \pm0.02)\,T_{{min}} \, + \,0.18\,( \pm0.03)\,A,\,(p\, < \,0.05,\,n\, = \,240),} $$

(4)

and stochastic realizations of the temperature regime were then generated through Equations 3 and 4.

Regional Variability of Precipitation and Temperature

We used NOAA-AVHRR 1-km, 10-day composites of the area (Kidwell 1995) obtained during the period April 1992 to April 1996 to estimate regional spatial fields of average air temperature (T _air ). The radiance values at Band 4 (T ₄ ) were rescaled with respect to the average values of 20 image pixels encircling the area of coverage of EPRA data, corrected to °C, and correlated to EPRA air temperature data. Altitude correction was achieved based on reported regional temperature–altitude profiles (Coronato 1992) and a 1-km Digital Elevation Model (EOS Data Gateway; http://edcdaac.usgs.gov/gtopo30/gtopo30.asp). These resulted in a time series of 65 regional thermal images during 1992 to 1996. Time profiles of the image series were obtained at 73 grid-corner points covering the region, and each profile was fitted to a sinusoidal trend (wavelength 365 days). Maximum (max) and minimum (min) fitted average, amplitude, and phase (Av, Am, Ph, respectively) values of the sinusoidal trends were used to construct combined estimates of minimum (minAv, minAm, minPh) and maximum (maxAv, maxAm, maxPh) values of the regional thermal ranges of variation. Similarly, regional time-serial fields of precipitation (P) over the Patagonian Monte were calculated through a series of NOAA-AVHRR images by applying the procedure described by Andersen (1996). This uses the zenithal instantaneous position of the AVHRR sensor (Zenith Band) and the recorded radiances at Bands 4 to 5 to generate images of the amount of atmospheric precipitable water (P _Water ). The procedure resulted in 65 regional P _Water image 16-day composite images during 1992 to 1996. These were rescaled to P values at ground level by calibration with 20 pixels surrounding the area of coverage of EPRA P ground data. The image-based P fields were then transformed to precipitation-ratio (P _r ) fields by dividing all pixel values by the average P at the 20 EPRA-calibrated pixels. This produced P _r maps describing the magnitude of regional anomalies of P with respect to simultaneous precipitation events at the EPRA site. The ranges of P anomalies were further stochastically sampled to generate an estimate of the regional variability of precipitation events.