Research

Research Interests:

Land surface processes in the Climate System, continental water cycle, water-vegetation-atmosphere interactions, hydroclimatology, droughts, terrestrial ecosystems, land-use change, global agriculture, land surface modeling, climate change.


Within the realm of land-climate interactions, my research has focused on several major themes:

Impact of soil-moisture atmosphere interactions in the mean state of the West African Monsoon in two models participating in GLACE-CMIP5 (Berg et al. 2017, JCLIM). Mean JJAS precipitation (mm/day) in (left) SM_INT and (center) SM_FIX simulations over 1971–2000 and (right) SM_FIX minus SM_INT difference. The gray contour line over land delineates the 1 mm day21 isohyet. Black contour lines over oceans represent 2-m air temperature. Stippling on the right-hand plots indicates significant SM_FIX - SM_INT differences (at 5%). Black contours represent the 1 and 5 mm day21 isohyet from SM_INT for each model. SM_FIX is a simulation without interactive soil moisture, instead a mean seasonal cycle is prescribed. SM_INT has interactive soil moisture.

Role of soil moisture feedbacks in the Climate System

As a postdoc at GFDL, I coordinated GFDL's participation in the GLACE-CMIP5 multi-model project, for which climate model simulations with and without interactive soil moisture were performed. I used these model experiments to show the critical role of soil-atmosphere feedbacks in key aspects of global land surface climate, including temperature and precipitation variability and extremes. To do so I applied novel diagnostics to these simulations, such as changes in the shapes of probability distribution functions of hydroclimatic variables, and changes in multivariate relationships.

I also used the experiment to demonstrate the crucial role of soil moisture feedbacks in the dynamics and seasonality of the West African Monsoon, in its response to climate change, and model uncertainties.

These experiments continue to be the basis for ongoing studies in which I am involved, regarding in particular the coupling between the water and carbon cycle, and soil moisture feedback on atmospheric circulation.

Beyond feedbacks on climate associated with the moisture state of the surface, I am also interested in issues related to feedbacks associated with land-use/land cover change.

Evaluating land-atmosphere and hydrological processes in climate models

I investigate how climate models represent the coupling processes between the land and the atmosphere: at the surface, how soil moisture controls land evaporation, how evaporation is represented as the sum of different components (canopy evaporation versus soil evaporation), or how precipitation is partitioned into evaporation and runoff. Above the surface, this also includes the impact of land-atmosphere fluxes on the boundary layer and precipitation.

To the extent that observations are available with the necessary spatial and temporal coverage, I try to compare model behavior against observations to evaluate model performance. I also investigate how model uncertainties propagate to the climate change projections provided by these same models.

Evaporative regimes in CMIP5 models (Berg and Sheffield 2018, JCLIM). CMIP5 multi-model mean correlation between summer-mean soil moisture (SM, top-10cm) and evapotranspiration (ET) over 1950-2005. This shows how, in models, evaporation is controlled by available soil moisture in the subtropics and midlatitudes, and by other factors (i.e., energy) in the Tropics and high-latitudes.

Analysis of CMIP5 projections of soil moisture at various depths under climate change (Berg et al. 2017, GRL). CMIP5 multimodel median relative change (in scenario RC8.5) (%) in (left) summer and (right) winter (top row) surface soil moisture, (middle row) 3m-soil moisture, and (bottom row) zonal-depth soil moisture (i.e., latitudes on the x-axis, depth on the y-axis; longitudes are averaged). Summer is JJA in the Northern Hemisphere and DJF in the Southern Hemisphere and vice versa for winter. The stippling indicates where more than three quarters of the models agree on the sign of projected changes. The full line represents the multimodel mean ratio of frozen over total soil moisture; the dotted line represents the global fraction of land by latitude (for both lines, values are on the right-hand y axis).

Response of the continental water cycle to greenhouse warming

Much of the literature on the future of the land water cycle under greenhouse warming has been dominated by the notion that warming will drive up land evaporation and thus dry out the land (i.e., reduced soil moisture, runoff), leading, in particular, to more droughts and increased background climatic aridity. Oftentimes, the warming projected by climate models is used (in combination with precipitation) in metrics and impact models to infer future impacts on hydrology and vegetation.

Relying on analysis of climate model simulations (CMIP5) and targeted climate experiments (e.g., GLACE), my research questions this paradigm by highlighting the role of land-atmosphere coupling and vegetation-water-climate interactions in the hydroclimatic trends projected by climate models: fundamentally, because air warming over land is both a forcing and a feedback on/from the land surface, near-air surface warming cannot be used as a proxy for hydrological changes under climate warming.

I am interested in exploring how such results can help inform more regional and applied hydrological predictions - predictions that could, on the one hand, represent the critical land-atmosphere processes (e.g., vegetation behavior, land-atmosphere coupling) that I've found to be essential for hydrological trends, but are often neglected in conventional hydrological studies - while, on the other hand, also providing high-resolution, locally-relevant water resources information.

Climate change impacts on crops

My background is in agronomy, and I have always been interested in the impacts of climate change on agriculture. During my PhD, I collaborated with agronomists to include one of the first representations of cultivated vegetation in a land surface model (here, tropical C4 cereal crop in IPSL's ORCHIDEE). This provides an innovative modeling framework in which to perform large-scale, yet process-based and spatially-explicit, assessments of climate change impacts on agricultural yields - in contrast with most crop impact assessment tools, which are either more statistical or more local in nature.

Compared to field trials, land-atmosphere measurements from eddy-covariance flux towers, and national yield census data, we demonstrated key improvements of the new version over the default (crops as grasses).

With sensitivity experiments we showed that, at regional scale, the two principal characteristics of rainfall forcing essential for high model skill in simulating interannual variability in productivity are cumulative annual variability and daily frequency - interestingly, the exact chronology of rainfall was not found to improve the model skill.

Using climate model outputs, we provided a large-scale, multi-model/scenario assessment of 21st century climate change impacts on crop yields in Africa and India. Results indicated a modest but robust detrimental impact of temperature rise, leading to a slight domain-wide average decrease in yield, albeit with a large spread in individual model projections that reflected the uncertainties in precipitation projections over arid regions in the Tropics (still a current issue with climate projections!).

From Berg et al., 2011: Standard version of ORCHIDEE (crops as grasses, dotted line) and modified version with crops (dashed). Full line is the original plot-scale crop model, validated with observations at this particular site and year (in Senegal, 1997).
Change (%) in simulated crop yields (C4 cereal) over Africa between 1970–1999 and 2070–2099 under the CMIP3 A2 scenario, over the different Köppen zones (see below) and for different climate models (in shades of grey).
Simplified Köppen classification, over the study domain, from the CRU monthly data over 1961–2000 (Mitchell et al., 2004). T, temperate; A, arid; d, desert; Eds, equatorial with dry season; Eh, equatorial fully humid.