Crop growth and irrigation interact to influence surface fluxes in a regional climate-cropland model (WRF3.3-CLM4crop)

Abstract

In this study, we coupled Version 4.0 of the Community Land Model that includes crop growth and management (CLM4crop) into the Weather Research and Forecasting (WRF) model Version 3.3 to better represent interactions between climate and agriculture. We evaluated the performance of the coupled model (WRF3.3-CLM4crop) by comparing simulated crop growth and surface climate to multiple observational datasets across the continental United States. The results showed that although the model with dynamic crop growth overestimated leaf area index (LAI) and growing season length, interannual variability in peak LAI was improved relative to a model with prescribed crop LAI and growth period, which has no environmental sensitivity. Adding irrigation largely improved daily minimum temperature but the RMSE is still higher over irrigated land than non-irrigated land. Improvements in climate variables were limited by an overall model dry bias. However, with addition of an irrigation scheme, soil moisture and surface energy flux partitioning were largely improved at irrigated sites. Irrigation effects were sensitive to crop growth: the case with prescribed crop growth underestimated irrigation water use and effects on temperature and overestimated soil evaporation relative to the case with dynamic crop growth in moderately irrigated regions. We conclude that studies examining irrigation effects on weather and climate using coupled climate–land surface models should include dynamic crop growth and realistic irrigation schemes to better capture land surface effects in agricultural regions.

Appendix: Dynamic crop module in WRF3.3-CLM4crop

We incorporated the dynamic crop growth module from CLM4CNCrop into the coupled regional model WRF3.3-CLM4. The dynamic crop growth module is based on AgroIBIS (Kucharik 2003) and described in detail in Levis et al. (2012).

1.1 Modifications

We made several modifications to the dynamic crop module to better fit into the coupled regional model framework. First, we fixed the soil carbon and nitrogen state variables. In the original CLM4CNCrop model, crop growth is linked to the carbon and nitrogen model, which updates multiple soil and plant carbon and nitrogen variables at each time step based on crop phenology and environmental changes. It requires a long spin-up time (over 1000s of years) to enable the soil carbon and nitrogen to reach current steady states. For a high-resolution regional climate model, such long spin-up simulations are difficult with current computing resources. Further, even though soil carbon and nitrogen are simulated in CLM4CNCrop, these values would not be routinely coupled to atmospheric carbon and nitrogen in a regional model. Because our regional scale focus is on biogeophysical, not biogeochemical feedbacks, between land and atmosphere, we assumed that for crops, the soil carbon and nitrogen could be maintained at optimum levels year–year.

Second, at this stage, we consider WRF3.3-CLM4crop able to simulate C3 and C4 crops, not specific crop types. The current version of CLM4CNCrop simulates three crops (summer cereal, soybean, corn). The growth of these crops is strongly dependent on photosynthetic pathway. We assume that at a regional scale, it is inappropriate to expect the model to simulate specific crops across the domain with validation only at one or several grid cells where observations are available. Therefore, we used C3 and C4 crop types to represent the potential growth of major crops (e.g., C3 crops: wheat, soybean, and C4 crops: corn, sorghum). The next phase of our work will aim to gather more observations and validate growth parameters for more specific crop types.

Third, we made changes to crop phenology and carbon allocation to better suit the regional coupled model framework and applications. In the planting phase, we changed the 20-year running mean growing degree days into 5-year running mean growing degree days to better match our simulation period. In the harvest phase, we assumed harvest occurs when the crop reaches 1.5 times the GDD required for maturity rather than occurring as soon as the crop reaches maturity as in CLM4CNCrop, since some crops such as corn (Nielsen 2011) are left in the field after maturity to dry. We also modified the carbon allocation to better reflect environmental stress on crop growth as described in section A3 of the appendix.

1.2 Phenology

1.2.1 Planting

The thresholds for planting, and thus initiation of the crop development cycle, are defined as:

$$\begin{aligned} & T_{2m} > T_{p} \\ & GDD_{8} > GDD_{min} \\ \end{aligned}$$

where T _2m is the instantaneous 2-m air temperature (°C), T _p is a crop-specific planting temperature (7 °C for C3 crop and 10 °C for C4 crop), GDD ₈ is the 5-year running averaged growing degree days (base 8 °C) from March to September, and GDD _min is the minimum growing degree day requirement (50 degree days for both C3 and C4 crops). C3 crop must meet the planting temperature requirement between March 1st and May 14th, and C4 crop between May 1st and June 14th.

At planting, some initial values are assigned, including leaf area index (0.1 m²/m²), stem area index (0.01 m²/m²), leaf carbon (3 gC/m²), stem carbon (3 gC/m²), and fine root carbon (4.5 gC/m²). The growing degree days value necessary for the crop to reach vegetative and physiological maturity, GDD _mat , is updated:

$$\begin{aligned} & GDD_{mat}^{c3crop} = 0.85GDD_{8} \\ & GDD_{mat}^{c4crop} = 0.85GDD_{10} \\ & GDD_{8} = GDD_{8} + T_{2m} - 8, \quad 0 \le T_{2m} - 8^{{\circ }} \le 30^{{\circ }} \,{\text{days}} \\ & GDD_{10} = GDD_{10} + T_{2m} - 10, \quad 0 \le T_{2m} - 10^{{\circ }} \le 30^{{\circ }} \,{\text{days}} \\ \end{aligned}$$

where GDD ₈ and GDD ₁₀ are the 5-year running averaged growing degree days from March to September.

1.2.2 Leaf emergence

Leaves emerge when the growing degree days for soil temperature (0.05 m depth soil, third layer of CLM) since planting ($GDD_{{T_{soil} }} ,$ base 0 and 8 °C for C3 and C4 crop) reaches 3 % of GDD _mat . At this phase, available carbon is allocated to leaf, live stem, and fine root according to constant allocation coefficients. Leaf area index generally increases and reaches a maximum value, which is prescribed as 6 m² m⁻² for C3 and 5 m² m⁻² for C4 crop. Also, the stem area index is updated as stem carbon gain or loss.

1.2.3 Grain fill

Grain begins to fill when the growing degree days since planting (GDD _plant ) reaches 70 % for C3 and 65 % for C4 crop of GDD _mat . The leaf area index and stem area index decline and transfer some amount (defined in A3) of leaf and live stem carbon to grain.

1.2.4 Harvest

We assumed harvest occurs when the crop reaches 1.5 times the GDD required for maturity (GDD _plant  > 1.5GDD _mat ) rather than as soon as the crop reaches maturity as defined in CLM4CNCrop, because crops, such as corn were left in the field after maturity to dry (Nielsen 2011).

1.3 CN allocation

Initial leaf carbon and nitrogen is assigned at planting. We adjusted the value from 1gC/m² in CLM4CNCrop to 3 gC/m² because the small initial leaf carbon generated a too small leaf carbon, resulting in low LAI compared to observations and too little gross primary production (GPP) for carbon allocation. The initial leaf nitrogen was calculated using leaf C:N ratio from Levis et al. (2012). C and N allocation starts with leaf emergence and ends with harvest. Carbon allocation is based on allocation coefficients and the nitrogen is assigned based on the tissue (leaf, stem, root, and grain) C:N ratio.

1.3.1 Leaf emergence to grain fill

The allocation coefficients to each C pool are defined as:

$$\begin{aligned} & a_{grain} = 0 \\ & a_{froot} = 0.7(1 - \beta_{p} ) \\ & a_{leaf} = 0.5(1 - a_{froot} ) \\ & a_{livestem} = 0.5(1 - a_{froot} ) \\ \end{aligned}$$

β _p is a plant functional type dependent variable that indicates the root water stress and varies from near zero (dry soil) to one (wet soil). We used β _p to better inform carbon allocation between root and shoot. When the soil is dry (small β _p ), more carbon is allocated to the root (Ericsson et al. 1996) to a maximum of 0.7. The rest of the available carbon is allocated to leaf and live stem in equal amounts.

1.3.2 Grain fill to harvest

During the grain filling period, fine root carbon allocation is still controlled by β _p , while the maximum C allocation to fine root is changed to 0.2. 80 % of the remaining carbon is allocated to grain and the other 20 % to tissues that are not explicitly simulated in the model, such as corn silk, flowers, etc. We assume the leaf and live stem carbon decline in this stage, and some portion of the carbon is transferred to grain

$$\begin{aligned} & a_{froot} = 0.2(1 - \beta_{p} ) \\ & a_{grain} = 0.8(1 - a_{froot} ) \\ & a_{leaf} = 0 \\ & a_{livestem} = 0 \\ & tran = c_{timestep} \left( {tan\frac{{GDD_{plant} }}{{GDD_{p} }}} \right) \\ \end{aligned}$$

where tran is the transfer coefficient of leaf and live stem carbon to grain carbon, c _timestep is an adjusted coefficient for each timestep, GDD _plant is the soil growing degree days since planting (base 8 °C for C3 crop and 10 °C for C4 crop), and GDD _p is the 5-year running averaged soil growing degree days from April to September (base 8 °C for C3 crop and 10 °C for C4 crop).