Search for: All records

In the stochastic linear contextual bandit setting there exist several minimax procedures for exploration with policies that are reactive to the data being acquired. In practice, there can be a significant engineering overhead to deploy these algorithms, especially when the dataset is collected in a distributed fashion or when a human in the loop is needed to implement a different policy. Exploring with a single non-reactive policy is beneficial in such cases. Assuming some batch contexts are available, we design a single stochastic policy to collect a good dataset from which a near-optimal policy can be extracted. We present a theoretical analysis as well as numerical experiments on both synthetic and real-world datasets. 

We explore the impacts of tile drains in agricultural fields on the coupled age and concentration dynamics of nitrate, immobile ammonium, mobile ammonia and ammonium, and nonreactive tracers such as chloride. We implement two mobile interacting pore domains to capture matrix and preferential flow paths in a coupled ecohydrology and biogeochemistry model,Dhara. We apply this model to an agricultural farm that utilizes a corn‐soybean rotation in the Midwestern United States located in the Intensively Managed Landscapes Critical Zone Observatory. In general, we observe both low concentration and age of nitrate in the areas that are classified as topographic depressions even with the presence of tile drains. Also, an increase in the age of mobile ammonia/ammonium is observed after installing tile drains. This is in contrast to the cases for nitrate, immobile ammonium, and nonreactive tracer. These results arise because the depletion of mobile ammonia/ammonium due to tile drainage causes a high mobility flux from immobile ammonium to mobile ammonia/ammonium, which also carries a considerable amount of relatively old age of nitrogen from immobile ammonium to mobile ammonia/ammonium. These results illustrate how storm event scale dynamics impact spatial heterogeneity and temporal variability of the efflux, which helps in disentangling the complexity of nitrogen dynamics in the soil. This understanding can contribute to precision agriculture for nitrogen applications to reduce environmental impacts.

 

Soil organic carbon (SOC) is going through rapid reorganization due to anthropogenic influences. Understanding how biogeochemical transformation and erosion‐induced SOC redistribution influence SOC profiles and stocks is critical to our food security and adaptation to climate change. The important roles of erosion and deposition on SOC dynamics have drawn increasing attention in the past decades, but quantifying such dynamics is still challenging. Here we develop a process‐based quasi 3‐D model that couples surface runoff, soil moisture dynamics, biogeochemical transformation, and landscape evolution. We apply this model to a subcatchment in Iowa to understand how natural forcing and farming practices affect the SOC dynamics in the critical zone. The net soil thickness and SOC stock change rates are −0.336 (mm/yr) and −1.9 (g C/m²/year), respectively. Our model shows that in a fast transport landscape, SOC transport is the dominant control on SOC dynamics compared to biogeochemical transformation. The SOC profiles have “noses” below the surface at depositional sites, which are consistent with cores sampled at the same site. Generally, erosional sites are local net atmospheric carbon sinks and vice versa for depositional sites, but exceptions exist as seen in the simulation results. Furthermore, the mechanical soil mixing arising from tillage enhances SOC stock at erosional sites and reduces it at depositional ones. This study not only helps us understand the evolution of SOC stock and profiles in a watershed but can also serve as an instrument to develop practical means for protecting carbon loss due to human activities.