Modeling Hydrology, Nitrogen Dynamics, and Crop Growth for Artificially Drained Cropland

Artificial drainage has helped in converting millions of hectares of flooded wetlands and swamps to highly productive soils. However, agricultural drainage effluent enriched with nutrients has resulted in severe ground and surface water contamination. Process-based simulation models of agricultural systems have been valuable tools for assessing the long-term impacts of different management practices on land productivity and water quality at both field and watershed scales. Upon entering the 21st century, however, more threats and challenges are confronting food security, land and water resources, in addition to the increasing concerns about climate change and global warming. This emphasizes the critical need for developing whole-system models which can help the design of effective water and nutrient management practices that reduce nitrogen (N) drainage losses and maintain high agricultural productivity under the anticipated scenarios of climate change and water availability. Most of the available models; however, emphasize certain processes in the agronomic ecosystems.;DRAINMOD and DRAINMOD-NII are widely used field-scale models for simulating the hydrology and water quality in artificially drained soils under wide variations of environmental conditions and management practices. However, they lack a mechanistic crop model that accounts for the effects of variations in seasonal weather conditions and soil nutrient status on crop phenological development and growth rate.;DRAINMOD-W, a DRAINMOD-based watershed-scale model, uses flow-based site specific regression equations to estimate edge of field loads that neglects the effects of nutrient and farming practices on N exports from fields to streams. In addition, DRAINMOD-W simulates N in-stream transformations using a fixed first-order decay equation ignoring the effects of spatial and temporal variability in stream characteristics on nutrient removal rate.;The first part of the current study focused on the development of a whole-system model for crop production on drained fields by integrating the deterministic widely-used crop growth modules of the DSSAT model (namely CERES-Maize and CROPGRO) with the hydrologic model, DRAINMOD, and the soil carbon and nitrogen model, DRAINMOD-NII on a daily time step. The second part of the current study focused on modifying the field and in-stream water quality components of DRAINMOD-W to produce a watershed scale model that is capable of modeling the hydrology and nitrogen fate and transport in artificially drained watersheds. Based on data availability, each model development was followed by a testing phase with different levels of detail to ensure that model integrity and simulation quality were properly maintained.;The modified version of the field scale DRAINMOD/DRAINMOD-NII, named DRAINMOD-DSSAT herein, was evaluated using a 10-yr (1996 -- 2005) data set collected from an artificially drained agricultural field located near Story city, Iowa, that planted to corn-soybean rotation with corn receiving low, medium, and high N fertilizer rates. The model was calibrated using data set collected from the high-N treatment, and validated for the other two treatments. For all treatments, predicted annual and monthly drainage flows, as well as annual and monthly nitrate (NO3-N) losses, were in very good agreement with respective measured values. Predicted and measured crop yields were in a very good agreement in most of the plot years. N removal in crop grains was adequately predicted; however, there were a consistent over-prediction of corn grain N receiving high N fertilization. Chapter 1 provides a detailed description of DRAINMOD-DSSAT and reports the model testing using the Iowa data set.;DRAINMOD-DSSAT was further evaluated using a 6-year (1985--1990) data set collected from a subsurface drained agricultural field at the Southeast Purdue Agricultural Center (SEPAC), Indiana. Subsurface drains were installed at three different spacings (5, 10, and 20 m). During the simulation period, the study site was planted to continuous corn receiving a high rate pre-plant N-fertilization. Data collected from the 20-m spacing plot was used for model calibration, and data sets collected from the other two plots were used for model validation. Predicted annual and monthly sub-drainage flows were in a very good agreement with measured volumes for both calibration and validation treatments. As well, the model was able to provide good predictions for monthly and annual nitrate losses. DRAINMOD-DSSAT was able to accurately predict the patterns of variation in corn yields in response to the variation in seasonal weather conditions, as well as the variation in soil water and nutrient regimes between different plots within each year. DRAINMOD-DSSAT testing using the Indiana data set is presented in Chapter 2 of this dissertation.;At the watershed level, modifications related to the in-stream processes component of DRAINMOD-W introduced a mathematical model that uses the mass transfer coefficient concept and considers the effects of stream water temperature and flow depth on nitrate removal from stream networks. In addition, the model was supported by a new mathematical formula to estimate daily stream water temperature as a function of daily air temperature. The modified model has been tested using a well documented 14-month data set (June 1998--July 1999) collected from a stream reach in a watershed in eastern North Carolina. Over the 14-month period, stream hydraulics and nitrate transport were well predicted by the model. Monthly net removal rates were adequately represented by the model. Net removal rate over the whole period was predicted with -0.18% error. The modification and testing of the in-stream processes component of DRAINMOD-W are presented in Chapter 3.;The second level of DRAINMOD-W model modifications allows for estimating edge-offield nitrogen loads with the physically-based model, DRAINMOD-NII. The modified model features were demonstrated through a simple case study that utilized limited data and surveys collected from a North Carolina Lower Coastal Plain watershed. Neither water yields nor nitrate load measurements at the watershed outlet were available. Previous DRAINMOD-W and DRAINMOD-NII applications in addition to available measurement from adjacent or enclosed sub-watersheds were utilized to justify model parameterization and predictions. Future work should focus on rigorously testing the modified DRAINMOD-W model using a more complete data set. The integration of DRAINMOD-NII into DRAINMOD-W is described in Chapter 4 of the dissertation.