The Importance of Model Resolution in Global Change Biology

Adam Wilson

Biologists increasingly realize that understanding the impact of global change on biological processes requires accounting for fine-grain environmental variability (Potter, Arthur Woods, & Pincebourde, 2013).  Similarly, climatologists have found that increasing the resolution of climate models typically produces better simulations of climate and precipitation (e.g. Jung et al., 2012; Kendon, Roberts, Senior, & Roberts, 2012; Mass, Ovens, Westrick, & Colle, 2002; Rauscher, Coppola, Piani, & Giorgi, 2010; Wehner, Smith, Bala, & Duffy, 2010).  But ‘fine-grain’ to a climatologist (which could be 10-250km), is more coarse than the grain relevant to an ecologist (0.01-10km).  Furthermore, in a recent evaluation of multi-model similarity across various resolutions, Masson and Knutti (2011)revealed stark variance between climate model predictions at resolutions finer than 2,000km.  The implications are dire for those who wish to understand how climate change affects ecological processes that occur at decision-relevant scales of meters to kilometers. There is an important gap in the spatial resolution of our understanding of climate change and ecological processes (Jetz, McPherson, & Guralnick, 2012; Potter et al., 2013).

Thus, biologists must account for fine-grain variability in climate (and other environmental variables) while acknowledging substantial uncertainty in the data.  This suggests two potential, and complementary, solutions: 1) improved high-resolution environmental datasets and 2) more sophisticated modeling strategies that allow for multi-scale data and the associated uncertainties.

Improved fine-grain environmental data

Advances in the capabilities of earth observing satellites over the past few decades have dramatically improved our ability to measure and monitor both physical and biological processes on the surface of the earth (Salomonson et al., 2011).  But incorporating these data into fine-grain ecological studies has been surprisingly slow.  Despite the acknowledged benefits of incorporating satellite data (e.g. Li et al., 2013; Wilson, Parmentier, & Jetz, 2014), many studies still rely on a ~1km resolution climatic data set (Hijmans et al., 2005) that was developed from station observations of temperature and precipitation alone. To help rectify this situation, the Environment and Organisms working group at the National Center For Ecological Analysis and Synthesis (NCEAS) is actively developing a global set of satellite-derived, 1-km resolution environmental datasets that will be made available for biogeographic analysis (Parmentier et al., in review; Wilson, Parmentier, & Jetz, 2013).  These will include, for example, cloud frequency, temperature, and precipitation over the past 30 years.  These data will be summarized to ecologically relevant metrics such as growing-degree days, growing season length, drought intensity, evapotranspiration, and other various extremes (Jackson, Betancourt, Booth, & Gray, 2009).  In addition to mean estimates of these quantities, we are also carefully quantifying the uncertainty in the predictions (arising from the interpolation to fine grains) to enable propagating those uncertainties through future analyses (e.g. Wilson & Silander, 2013).  However, global environmental data at spatial grains that truly matter to individual organisms (meters to kilometers) will continue to have notable uncertainties for the near future.  So in addition to continuing development of improved environmental data, we also need to develop analytical strategies that incorporate multi-scale data and can account for uncertainties in environmental datasets.

Thinking (and modeling) across multiple scales

It is common for scientists, even within a discipline, to focus on processes at a particular spatial or temporal scale.  For example, plant population demographers may collect and analyze data at the 0-1ha scale, community ecologists from 1-10ha, landscape ecologists from 102 -104 ha, and continental scale remote sensors from 104 -106 ha. There is a similar disconnect between ecosystem ecologists who ponder dynamics over 1-100 years, while evolutionary biologists reflect on changes over 104-108 years. These sub-disciplines typically publish in different journals, go to different meetings, and use different language despite the fact they are looking at the same processes, only at different scales.  This is reminiscent of the parable in which several blind men hold various parts of an elephant and exclaim they alone understand the nature of elephant-ness.

We need to expand our thinking (and modeling) to include processes and uncertainties that occur across multiple scales.  If the global (or regional) climate models are incapable of accurately predicting change at spatial grains directly relevant to ecological processes, we should incorporate that uncertainty into our modeling framework.  Hierarchical Bayesian models offer this kind of flexibility (Clark, 2005).  For example, in Keil, et. al (2013), we used multi-scale models to estimate the probability of occurrence across San Diego County for over one hundred bird species given relevant environmental attributes.  We have since explored using this procedure up to estimate the probability of occurrence for the American three-toed woodpecker (Picoides dorsalis) across the continental United States (Petr Keil, Wilson, & Jetz, 2014).  In those examples, our environmental data were finer than our biological information, but a similar framework could easily account for the opposite situation. The advantage of this sort of approach is that uncertainties are propagated across scales and can truly quantify our understanding (or lack of understanding) about the process (Clark, 2005).   

By combining finer-resolution data with more sophisticated multi-scale modeling strategies, we can bridge the resolution gap and facilitate probabilistic statements about the past and future impacts of climate change on biological systems.  Enabling fine-grain estimates of biodiversity in a changing environment will empower decision-makers with information at scales relevant to conservation (that of towns and reserves) rather than regional to continental scale studies currently available.  Furthermore, by propagating uncertainties introduced by the mismatch of spatial resolutions, we can quantify where we should be confident in our predictions and where we must be cautious due to overwhelming uncertainties.  This would be a significant advance over current projections of biodiversity under climate change that tend to either be “too coarse to be useful” for developing detailed conservation strategies or “too fine to be believed” due to underestimated fine-grain uncertainties. 

References

Clark, J. S. (2005). Why environmental scientists are becoming Bayesians. Ecology Letters, 8(1), 2–14.

Jackson, S. T., Betancourt, J. L., Booth, R. K., & Gray, S. T. (2009). Ecology and the ratchet of events: Climate variability, niche dimensions, and species distributions. Proceedings of the National Academy of Sciences, 106, 19685–19692. doi:10.1073/pnas.0901644106

Jetz, W., McPherson, J. M., & Guralnick, R. P. (2012). Integrating biodiversity distribution knowledge: toward a global map of life. Trends in Ecology & Evolution, 27(3), 151–159. doi:10.1016/j.tree.2011.09.007

Jung, T., Miller, M. J., Palmer, T. N., Towers, P., Wedi, N., Achuthavarier, D., … Hodges, K. I. (2012). High-Resolution Global Climate Simulations with the ECMWF Model in Project Athena: Experimental Design, Model Climate, and Seasonal Forecast Skill. Journal of Climate, 25(9), 3155–3172. doi:10.1175/JCLI-D-11-00265.1

Keil, P, Belmaker, J., Wilson, A. M., Unitt, P., & Jetz, W. (2013). Downscaling of species distribution models: a hierarchical approach. Methods in Ecology and Evolution, in press. doi:doi: 10.1111/j.2041-210x.2012.00264.x

Keil, Petr, Wilson, A. M., & Jetz, W. (2014). Integrating heterogeneous data, spatial autocorrelation and priors in multi-scale species distribution models. Diversity & Distributions, in review.

Kendon, E. J., Roberts, N. M., Senior, C. A., & Roberts, M. J. (2012). Realism of Rainfall in a Very High-Resolution Regional Climate Model. Journal of Climate, 25(17), 5791–5806. doi:10.1175/JCLI-D-11-00562.1

Mass, C. F., Ovens, D., Westrick, K., & Colle, B. A. (2002). Does Increasing Horizontal Resolution Produce More Skillful Forecasts? Bulletin of the American Meteorological Society, 83(3), 407–430. doi:10.1175/1520-0477(2002)083<0407:DIHRPM>2.3.CO;2

Masson, D., & Knutti, R. (2011). Spatial-Scale Dependence of Climate Model Performance in the CMIP3 Ensemble. Journal of Climate, 24(11), 2680–2692. doi:10.1175/2011JCLI3513.1

Parmentier, B., McGill, B. J., Wilson, A. M., Regetz, J., Jetz, W., Guralnick, R. P., … Narro, M. (in review). Assessing methods and remotely sensed derived covariates for interpolating high-resolution daily air temperature datasets.

Potter, K. A., Arthur Woods, H., & Pincebourde, S. (2013). Microclimatic challenges in global change biology. Global Change Biology, 19(10), 2932–2939. doi:10.1111/gcb.12257

Rauscher, S. A., Coppola, E., Piani, C., & Giorgi, F. (2010). Resolution effects on regional climate model simulations of seasonal precipitation over Europe. Climate Dynamics, 35(4), 685–711. doi:10.1007/s00382-009-0607-7

Salomonson, V., Abrams, M. J., Kahle, A., Barnes, W., Xiong, X., & Yamaguchi, Y. (2011). Evolution of NASA’s Earth Observing System and Development of the Moderate-Resolution Imaging Spectroradiometer and the Advanced Spaceborne Thermal Emission and Reflection Radiometer Instruments. In B. Ramachandran, C. O. Justice, & M. J. Abrams (Eds.), Land Remote Sensing and Global Environmental Change (pp. 3–34). Springer New York. Retrieved from http://link.springer.com/chapter/10.1007/978-1-4419-6749-7_1

Wehner, M. F., Smith, R. L., Bala, G., & Duffy, P. (2010). The effect of horizontal resolution on simulation of very extreme US precipitation events in a global atmosphere model. Climate Dynamics, 34(2-3), 241–247. doi:10.1007/s00382-009-0656-y

Wilson, A. M., Parmentier, B., & Jetz, W. (2013). Systematic landcover bias in Collection 5 MODIS cloud mask and derived products – a global overview. Remote Sensing of Environment, in press. doi:10.1016/j.rse.2013.10.025

Wilson, A. M., & Silander, J. A. (2013). Estimating uncertainty in daily weather interpolations: a Bayesian framework for developing climate surfaces. International Journal of Climatology, n/a–n/a. doi:10.1002/joc.3859