Our Projects

Spatial factors are important for crops – how close are shade trees, how widely to fertilize, or how large should plots be, to name a few considerations. But the impacts can be tricky to analyze, and trickier still in an agroforestry context where farm plots can be complex.

Using spatial statistical methods we unpicked how factors like shade tree species and distance, soils, edge effects, and boundaries between plots all impacted yield quality and quantity, as well as disease.

Agrochemicals deal with all the environmental stressors found in the field. Even the most robust products become less effective in extreme temperatures or periods of plant stress. Pinpointing which conditions are responsible given the complexity of climate and local environments.

Using only survey data and locations from-end users, we can identify which extreme environmental changes impacted product performance by scraping climate data from the web. Working these models we can find critical thresholds specific to farming regions, such as temperature cutoffs.

Smallholder or family farms have limited space and often rely on a diverse base of saleable products. However, getting the best spatial positioning out of all profitable crops is a surprisingly difficult calculation - even if a computer is provided with perfectly accurate data.

Rather than trying to find the perfect layout, we have developed algorithms in which basic requirements such as crop diversity, annual budgets, and land surface can be used on a per-farm basis. These algorithms use an iterative, evolutionary process to return a range of options to inform planning.

Newly emerging pests require regionwide management strategies. Predicting the likely locations for spreading invasive pest species is critical for removal efforts.

Temperatures are often important for predicting pest outbreaks, especially in invertebrate pests. But weather dependence means these models can give too little warning of an outbreak for managers to intervene in time.

The insect emerald ash borer has ripped through forests in the eastern US, devastating native ash trees at an alarming pace. Current modeling efforts by academic and public scientists are critical for understanding how this invasive species impacts both forests and urban greenspaces.

Real world data is usually flawed and messy, not collected from carefully designed experiments. Answering questions with data that are missing controls or when treatments aren’t applied consistently is a serious challenge even for experienced analysts and requires bespoke solutions.

Managers can tap into a wealth of pest management knowledge from out-of-state and overseas experiences. But pests adapt to local conditions, such as climate, so it’s important to ground truth any external practices with local data from the field.

How are we doing, and where are we headed? These questions are key in an evolving situation like a new pest outbreak, where management resources are limited and failure could cripple a multibillion dollar industry.

It’s expensive and difficult to monitor range expansion in a novel disease, especially at the regional level. Surveys need to be informed by risk factors such as climate, human demographics, and transport vectors and corridors.

Ag professionals are interested in trying new practices to diversify farms. However, given the extremely large number of potential changes to crops and product usage, it's difficult to estimate long-term outcomes.

Experimentation is critical to successful product R&D, but sample size is expensive – too many samples is a costly waste, but with too few samples results can be inconclusive. In R&D, we want to find the sweet spot to balance statistical power against sample size.

Crops can often be pollination limited. To maximize yields, the correct timing and intensity of pollination services can lead to significant payoffs. In cacao production, hand-pollination can impact the size, number, and quality of fruit simultaneously. We have developed structural equation models to leverage large-scale, long-term data to determine the ideal hand pollination rates for cacao trees.

Before they launch, a business needs a complete picture of how a new product performs, including against their existing brands. While industry solutions are easy to use, their analyses are broad and can be hard to fine-tune to the quirks of each product.