NASA climate modeling The new NASA global data set combines historical measurements with data from climate simulations using the best available computer models to provide forecasts of how global temperature (shown here) and precipitation might change up to 2100 under different greenhouse gas emissions scenarios. Credits: NASA

 

What is a climate model?

A climate model represents the earth's atmosphere, oceans, ice, land surface, and vegetation as they interact and respond to changes in the sun, volcanos, land use, and atmospheric chemistry.  Climate models are an important tool for understanding the implications of societal emissions of greenhouse gases (such as carbon dioxide, CO2) that trap energy in our climate, raising global temperatures, shifting weather patterns, and altering the likelihood of extreme events.  Climate models are based on atmospheric physics and chemistry, simulating daily weather and long term climate shifts using some of the most powerful computers in the world.

Why are climate models different?

There is strong agreement in the global climate response to greenhouse gas emissions (such as carbon dioxide, CO2), however different levels of complexity and resolution cause unique regional patterns of climate changes that are important to agriculture.  Climate models simulate weather patterns and climate shifts over the whole globe, however they cannot represent every process and local effect in the climate system.  Each model has its own way of representing established atmospheric physics and chemistry, which can lead to subtle differences in how climate change drives global and regional impacts.  Climate models operate on global grids (the size of an individual grid box varies from model to model -- typically 50-200 km across), and differ in the way that they represent climate processes that are smaller than a grid box.  Models also vary in their representation of the upper atmosphere, ocean circulation, and land and ice processes. 

How was climate historical climate data assembled?

Historical meteorological station observations form the gold standard for historical weather and climate information used in the AgMIP Regional Integrated Assessments.  We work with national environmental and meteorological agencies to locate weather stations in our agricultural zones, then examine the data to identify any spurious values or gaps in the data.  These we replace with data from the NASA AgMERRA historical climate forcing dataset, adjusting for any monthly biases identified in a comparison between the available historical observations and AgMERRA.  The result is a daily time series of rainfall, maximum and minimum temperatures, solar radiation, wind speed, and humidity from 1980-2010 that may be used to drive crop and livestock models.

How were climate scenarios selected?

We cannot precisely predict the way that society, industry, and technology will develop in the coming decades, and thus we cannot precisely predict the amount of greenhouse gas emissions that will drive climate change (such as carbon dioxide, CO2).  Scientists therefore use climate models to project different scenarios of greenhouse gas levels in the atmosphere, exploring the regional temperature and rainfall impacts that affect agriculture and food security.

Regional climate scenarios produced for the AgMIP Regional Integrated Assessments were designed to sample the most important aspects of regional climate change for the local agricultural sector.  We first sampled regional projections of growing season temperature and precipitation changes from a set of 29 global climate models, then selected 5 models for analysis that best represented the types of climate changes seen in the full set.  Specifically, a climate model was selected to represent the middle temperature and rainfall change projections for each region, and then four were selected to represent relatively hot and dry, hot and wet, cool and dry, and cool and wet projections compared to this middle scenario.  Note that "relatively cool" scenarios are still warmer than today as climate model projections for the 2050s are all warmer than present conditions, however these models do not project as great a regional warming and are therefore cooler than most other models.  The combination of relatively cool/wet, cool/dry, middle, hot/wet, and hot/dry models allow us to explore how a broad range of plausible climate projections affect regional agriculture, and we can also track how many of the full 29 climate model set projects similar regional climate changes in order to understand the probability of a given risk.

How were climate scenarios generated?

For each selected climate scenario we compare the climate model projection of the 2040-2069 period (referred to as the "2050s" given the central decade) against the same climate model's 1980-2009 period.  We calculate monthly changes in rainfall, maximum and minimum temperatures, the number of rainy days, and the standard deviation of maximum and minimum temperatures.  These quantities then allow us to alter the distribution of temperature and rainfall for each month according to the climate model projections.  To do this, we adjust the original observational time series impose new mean temperatures and rainfall totals, shift the number of rainy days, and alter the frequency of extreme temperatures within the growing season.   The result is a time series that retains many of the day-to-day characteristics of the observed climate from 1980-2010 but also reflects the new averages and distribution that result from projected climate changes.

What is a crop model?

Crop models are computer programs that predict daily growth and development of crops in response to daily weather, soil conditions, crop management, and variety traits, through to the point of final yield and biomass.  The crop models are not regression models, but rather simulate daily processes such as photosynthesis, leaf area growth, and grain growth, with a daily time-step.  The models are written in various programming languages, and are coded and parameterized with relationships describing the sensitivities of processes to solar radiation, temperature, water stress, and N supply.  The models require inputs of crop management, crop variety traits, soil water and N supplying traits, as well as daily weather.  The model outputs include daily growth, crop evapotranspiration, N uptake, etc., as well as final yield and yield components.

Why are crop models different?

Crop models have been developed by many different teams throughout the world, and the structure and theories used in those models differ according to the concepts of the model developers-authors.  Thus the models have different coding and parameterization for sensitivities of growth processes to solar radiation, temperature, water, and N supply.  The models often have different temperatures sensitivities (base and optimum temperatures) for simulating various processes such as photosynthesis, leaf area expansion, and grain growth rates (in part based on how the model developers interpreted sparse published literature).  This can cause different model responses to temperature (and we have seen this for APSIM compared to DSSAT maize models).  Models may have different parameterization for the photosynthesis and transpiration sensitivity to elevated carbon dioxide, thus resulting in differences among models in response to elevated CO2.  While the weather and management inputs should be the same, the models typically have different ways of interpreting how the crop predicts soil water balance and evapotranspiration, how the soil organic C pools are available for mineralization of N, and how the crops handle N uptake for growth.  This can cause variation among models in their sensitivity to rainfall, water supply, and N fertilization.