Climate Smart Pest Management

Abstract

This study examines the role of weather and pest infestation forecasts in agricultural pest management, taking into account potential correlation between weather and pest population prediction errors. First, we analytically illustrate the role of the correlation between weather and pest infestation forecast errors in pest management using a stochastic optimal control framework. Next, using stochastic dynamic programming, we empirically simulate optimal pest management trajectory within a growing season, taking into account correlation between weather and pest population prediction errors. We used lentil production in the Palouse area of northern Idaho and eastern Washington as a case study, where pesticide use was restricted due to environmental or health reasons. We showed that the benefit of applying pesticides later in the growing season can outweigh benefits of early application when pesticide use is restricted due to environmental or health regulations. The value of information is close to $9 per acre, approximately 8% of the expected net returns per acre, and close to $12 per acre ($106–$94), or approximately 13% of the expected net returns per acre for the baseline versus the climate change scenarios, respectively.

Using Time-Series Generative Adversarial Networks to Synthesize Sensing Data for Pest Incidence Forecasting on Sustainable Agriculture

“…The stored data are analyzed to identify potential problems and to advance agricultural technology. Big data time analysis plays a crucial role in this because agricultural sensors often produce huge amounts of time-series data [20][21][22]. As the method used for data processing can affect downstream results, the challenge is how to analyze past data, adjust the dataset appropriately, and come up with the best processing strategy.…”

“…Here, the generator input is a random sequence of size (2,22) as described in Section 3.2.1, and the generator and discriminator are both networks with three layers of GRU neurons stacked on top of each other. The number of GRU neurons cascaded in each layer is hidden_dim, the hidden dimension of the network.…”

“…Since we seek to expand the data in two-week increments, we set hidden_dim to 2, which causes the TimeGAN model to generate and identify two-week sample data. In addition, the generator input is a random vector of size (2,22), and the generator output and the discriminator input are also sequences of size (2,22), but the discriminator output is a discriminative score of size 1, used to determine whether the current generated sequence consists of true or fake data.…”

“…In this study, the embedded and recovered networks are trained with real data of size (2,22), as discussed in Section 3.2.1. As the size of the generated target data must be the same as the real data, the input size of the network is the same.…”

“…The hidden_dim is set to 2 to match the data size of the generator and the discriminator. In addition, the input of the embedded network consists of real data of size (2,22) and the output of the embedded network and the input of the recovered network are training sequences of size (2,22), but the output of the recovered network has a size of 22 features scored for the coding and decoding effect of the autoencoder.…”