# Introduction

Changes in atmospheric CO$$_2$$ concentrations between 1600 and 1750 as observed in Antarctic ice cores (Bauska 2015) coincide with societal and environmental changes of the Columbian encounter in South America. Several studies (Nevle 2011, Kaplan 2010) suggested a link between population collapse, reforestation and an increase in carbon sink potential.
A 600-year long global HadCM3 simulation (Schurer 2013) which assesses the effect of solar forcing on Northern Hemisphere temperature over the last millennium has been used to determine the effects of volcanic eruptions, changes in greenhouse gas concentrations and land use on temperature and precipitation in different regions of Central and Southern America between 1400 and 1850. Scenarios used by (Schurer 2013) ranged from specific solar forcings to volcanic eruptions, greenhouse gas emissions (GHG) and simulations with land use held constant at 1400 values. Table \ref{tab:runs} lists the scenarios and their individual forcings used herein.

\label{tab:runs}

 Scenario No. of runs SOLAR VOLC GHG LUSE AER O$$_3$$ Orbital Control 1 850 800-850 800-850 825 0 PI 825 ALL 4 x x x x >1820 x x NoLUSE 4 x x x 1400 >1820 x x GHG 4 1400 0 x 1400 0 PI 1400

# Methods/Set up

The land use component of the model was forced by the land use simulation from (Pongratz 2009), which uses relatively conservative population estimates for pre-Columbian America and see less land use change than other land use simulations (Kaplan 2010). The NoLUSE simulation uses constant land use values from 1400 and aerosol concentrations from past 1820 while all other forcings act over the whole time period, i.e. it captures all natural and greenhouse gas forcings. In the GHG simulation all forcings except greenhouse gas emissions are constant at their 1400 or 0 levels respectively to identify the effects of changing greenhouse gas concentrations. The ALL simulation uses all forcings over the whole time period (except aerosols from 1820 onwards) and is used here to compare the impact of land use, as NoLUSE and ALL have the same forcing set-up apart from land use. The control scenario has all forcings set to their initial (i.e. AD 800-850) conditions and is used to determine effects from individual forcings in the forced ensemble runs by comparing the forced runs to the control scenario. As the focus lies on Central and South America the results of the global HadCM3 simulation were clipped to different areas of interest (AOI) (see figure \ref{fig:AOI}). The area averaged temperature and precipitation for the eight regions was extracted from every ensemble member. A 11-year moving average was then applied to all outputs to account for the solar cycle (Schurer 2013).

# Results

## Temperature

The temperature time series of the GHG scenario follows in general the course of the control run, i.e. the magnitude of the variations of both ensembles are similar. Six out of eight regions however show a temperature that is on average about 0.25 $$^{\circ}$$C higher than the control run over the whole time period from 1400 to 1850 (Figures \ref{fig:T_SDN_GHG} & \ref{fig:T_MON_GHG}). The remaining two regions (fully humid (temperate) & arid (S)) still have a 0.15 $$^{\circ}$$C higher mean temperature compared to their control run (see figure \ref{fig:T_ARS_GHG}). Two regions in Central America show a decrease in temperature at around 1600 when CO$$_2$$ concentrations abruptly decreased (figure \ref{fig:T_SDN_GHG} & \ref{fig:T_ARN_GHG}) .

The NoLUSE simulation shows a completely different pattern. The most prominent feature, visible in all regions, are five extrema located at 1460-1475, 1590-1610, 1690-1700, 1815 and 1835. The 1460-1475 and 1815 extrema are the most prominent minima with dips in temperature of up to 1 $$^{\circ}$$C in the regions arid (south), winter dry (north) and monsoonal (figures \ref{fig:T_ARS_LSE}, \ref{fig:T_WDN_LSE} & \ref{fig:T_MON_LSE}). While the 1815 event is unmatched by the control scenario in any region, the 1460-1475 as well as the less prominent 1590-1610, 1690-1700 and 1835 were matched by the control run in the arid (north) region (figure \ref{fig:T_ARN_LSE}). In all other regions the smaller extrema are clearly distinguishable from the control run or even have a different direction than the control scenario (e.g. figure \ref{fig:T_WDS_LSE}).

## Precipitation

Both, the NoLUSE and the GHG scenario show on average only little divergence from the control scenario. Notable exceptions are the regions winter dry (north) and winter dry (south) under the NoLUSE scenario (figures \ref{fig:P_WDN_LSE} & \ref{fig:P_WDS_LSE}). In both regions a 1450 maximum event, similar to the temperature event (but reversed) can be clearly distinguished. In the winter dry (south) region extrema at around 1700 and 1815 are also visible. Furthermore the monsoonal region in the NoLUSE scenario shows a peak at around 1590-1625 (figure \ref{fig:P_MON_LSE}), similar to the 1590-1610 event in temperature under the same scenario.

## Land use

A comparison between the ALL scenario with land use forced over the whole time period and the NoLUSE scenario with a fixed land use forcing at the 1400 level shows that precipitation and temperature follow the same path under both scenarios with no clear differences in magnitude or course.

# Conclusion

A comparison between the temperature GHG runs and the control scenario indicates that elevated greenhouse gas concentrations (while all other forcings are muted) have led to an average increase in temperature of around 0.25 $$^{\circ}$$C between 1400-1850. The increase in greenhouse gas concentrations during this period however appears to have no impact on the precipitation regime. The extreme events in the NoLUSE run coincide with major volcanic events (see table \ref{tab:volcans}). Since the ALL scenario and the NoLUSE scenario are in close agreement and the GHG scenario does not reflect the events observed in the NoLUSE scenario these forcings can be excluded as cause for the short term extrema.

\label{tab:volcans}

 Extreme event Volcanic eruption Reference 1460-1475 Kuwae, Vanuatu (1452-53); Bardarbunga,Iceland (1477) (Gao 2006), (Thordarson 2007) 1590-1610 Billy Mitchell, Papua New Guinea (1580); Huaynaputina, Peru (1600) (Palmer 2001), (Verosub 2008) 1690-1700 Long Island, Papua New Guinea (1660) (Zielinski 1994) 1815 Mt. Tambora, Indonesia (Zielinski 1994) 1835 Cosiguriu, Argentinia (1835) (Self 1989)

The absence of an effect of land use on temperature and precipitation may has several reasons. Firstly the land use scenario used to force this HadCM3 model shows no or only little land use change in Central and South America before and during the Columbian encounter (Pongratz 2009), thus the detection of a land use signal in this model configuration is unlikely. Secondly HadCM3 as a coupled atmosphere-ocean general circulation model may not detect short term changes in soil carbon and other carbon fluxes which are a substantial part of land use change and would have an impact on atmospheric greenhouse gases which then in turn would have an effect on temperature and eventually precipitation through feedback mechanisms. To detect these short-term fluctuations and their effect on climatic variables the use of an Earth System model is necessary, therefore the use of HadGEM2-ES as preferred model type can be justified.