Model One: Visualization of Drought Conditions
Model Background
In China, drought is a long-standing climatic problem with enormous impacts on agricultural output, natural resources and socio-economics across the country. Drought can lead to crop yield loss, water scarcity, soil erosion, ecosystem collapse, and economic hardship for those engaged in agriculture. In the face of this problem, we embarked on a major research effort aimed at providing better drought monitoring and response tools through synthetic biology and data analysis.
Model Significance
Comprehensive understanding of drought: The drought map we have created not only provides an overall picture of drought across the country, but also digs deeper into the geographic differences between different regions. This gives farmers, governments and policy makers a clearer picture of the complexity and diversity of drought issues.
Optimize resource allocation: Governments can use map data to allocate resources, including water and emergency aid, more efficiently. This improves the efficiency of response to drought events and reduces the burden on society and the economy.
Ecological conservation: Understanding the distribution of drought-prone areas can help to implement measures to protect fragile ecosystems, preserve biodiversity, and promote the restoration of ecological balance.
Responding to climate change: Drought is part of climate change, and it will become more prevalent and severe in the future. Our work contributes to a better understanding of the impacts of climate change on drought and provides a basis for developing strategies to address climate change.
Modeling Results
From China Drought and Water Hazard Defense Public 2021 and Impact of Climate Change on Food Security in Taiwan and Response Countermeasures Yang Ming-Hsien, we obtained reliable data on the number of drought occurrences in 23 provinces, five autonomous regions, and four municipalities in China, and based on the pyecharts package, we visualized drought conditions in each region of China.
Model Two: Intelligent Plant Drought Recognition with CNN
Model Background
In the face of the threat posed by climate change to agriculture and ecosystems, our project is dedicated to the development of a revolutionary method to produce color changes under drought conditions by using the RUBY gene introduced into plants through synthetic biology and artificial intelligence techniques. At the same time, we apply the image recognition capabilities of Convolutional Neural Networks and combine this technology with the photo-taking capabilities of smartphones so that it can be used on farmers' smartphones. This innovative work will provide farmers with a convenient, real-time drought monitoring tool and provide feedback to Farmgate on solutions based on drought conditions, which provides strong support for the world's agricultural development and ecosystem conservation.
Model Purpose
The primary goal of this model is to eliminate the subjectivity and inconsistency associated with human eye-based recognition of leaf color, in order to inform farmers, agricultural practitioners, and decision-makers about specific drought conditions in a more standardized manner. The significance of this model lies in its ability to provide reliable and consistent data, unaffected by individual supervisor's subjective judgments or environmental conditions.
Model Parameters
Input Layer: 3-channel color images (RGB). Convolutional Layers: Two convolutional layers, each with 32 filters of size 5x5, utilizing the ReLU activation function, with the first convolutional layer followed by max-pooling. It also includes Batch Normalization to enhance convergence speed and robustness. The Sigmoid activation function is employed to introduce non-linearity. Two fully connected layers, each with 64 units, with the first fully connected layer using Dropout to mitigate the risk of overfitting. The output layer consists of 3 units, suitable for multi-class classification tasks. Input images are resized to 64x64 pixels and then converted to tensor form. The training and testing datasets are loaded from separate folders and are shuffled during loading. The optimizer is set as stochastic gradient descent (SGD) with a learning rate of 0.01 (adjustable). The loss function used is the cross-entropy loss. The model is trained for 500 epochs, typically converging after around 300 epochs. Both training and testing are performed with a batch size of 32. TensorBoard is used for visualizing training and testing losses and model accuracy.
Model Operation
Training Phase: In each training epoch, the model performs forward propagation on the training dataset, computes the loss, and then performs backpropagation to update the model's weights. The model continually optimizes to minimize the loss function, aiming to improve classification performance.
Testing Phase: After each training epoch, the model switches to testing mode and evaluates its performance using the testing dataset. Loss and accuracy are calculated and recorded.
Model Saving: During specific epochs (from 490 to 499), the model checkpoints are saved to disk for future use.
Model Significance
Agricultural Intelligence: Combining the shooting capabilities of smartphones with the image recognition capabilities of convolutional neural networks, we provide farmers with an easy-to-use tool for quantitatively sensing drought conditions. This provides farmers in rural areas with a quick and easy way to monitor the health of their crops without the need for expensive equipment.
Increased efficiency in agricultural production: Farmers can instantly access plant color data through the mobile app and know if plants are affected by drought. This helps to increase the efficiency of agricultural production and reduce losses.
Social Sustainability: By providing real-time drought monitoring tools, we help maintain the sustainability of agriculture and ecosystems. This not only benefits the economic situation of farmers, but also contributes to the stability of food supply and conservation of natural resources.
Model Principle
Our model synthesizes the principles of synthetic biology, artificial intelligence and mobile technology:
Input Data: Plant images taken by smartphones capture data on leaf color, including red (R), green (G), and blue (B) values at each pixel point.
Data Processing: We use the transform library in torchvision to first convert the image type to a tensor type and then process the image size to get a 3 by 64 by 64 3D tensor.
Neural network structure: our neural network consists of an input layer, three convolutional layers, two pooling layers, a Dropout layer and an output layer (a vector value representing the drought indication).The Batchsize is 32 and the model parameters are as follows:
Applies the element-wise function:
Model Results
Our model was trained and tested with satisfactory results: it achieved an accuracy of 97.96% on the overall test set.
Real-time drought monitoring: The model can output real-time drought conditions based on leaf color data provided by pictures taken by the phone, so farmers can instantly know the status of their plants.
Convenience: Farmers don't need expensive equipment or specialized knowledge; all they need is a smartphone with a camera function for accurate drought monitoring.
Analysis of Results
The success of this work demonstrates:
Equation of the model:
1. Integration of multidisciplinary technologies: combining synthetic biology, neural network technologies and smartphones provides farmers in rural areas with unparalleled drought monitoring tools.
2. Protecting farmers' interests: Farmers can better respond to the threats of climate change and drought by taking timely actions to minimize losses and maintain the sustainability of agriculture.
3. Technology-driven agriculture: This project represents the key role of technology in solving global challenges and opens up entirely new possibilities for agricultural production and natural resource management.
By integrating synthetic biology, smartphone technology, and artificial intelligence, we are providing farmers with an unprecedented way to address the risks of climate change and drought, contributing significantly to the sustainability of agriculture and ecosystems. This is not only a function of technological innovation, but also of social sustainability, guaranteeing the prosperity of rural communities and ecological balance.