A High-Resolution Global-Scale Model for COVID-19 Infection Rate
收藏资源简介：
This dataset contains all information to reproduce our experiment to produce a high-resolution global map (0.1°) of infection-rate risk for COVID-19, based on temperature, precipitation, and CO2. The produced risk index map predicts most of the areas with an actual high risk (87% accuracy), which are characterized by a moderate-high level of CO2, moderate-low temperatures, and a moderate level of precipitation. With respect to our previous model (https://zenodo.org/record/3945495#.YG7UEugzaUk) - which had a coarser 0.5° resolution - this new model is much more accurate at predicting real-world scenarios that reported both high and low infection rates in 2020 (80% accuracy). <strong>Explanation of data and images:</strong> comparisonvert.png -> Visualisation of the output produced by our model: (a) distribution of high-infection-rate areas using the MaxEnt balanced threshold (0.008), (b) probability peak areas (0.13 threshold), (c) overlap between low infection rate countries extracted from real data and our risk map, and (d) highlight of low infection rate countries not predicted by our model<br> countries_high_rate.csv-> high-infection-rate countries<br> countries_low_rate.csv-> low-infection-rate countries<br> countries_low_rate_mispredicted.csv-> low-infection-rate countries mispredicted by our model<br> covid_derivatives.csv-> extracted average derivatives of world countries covid_risk.csv->Risk Map dataset<br> gp.asc-> MaxEnt distribution raster<br> LowDerivativeRegions.png->low-infection-rate countries - image<br> MaxEnt distribution.png->distribution of high-infection-rate areas using the MaxEnt balanced threshold (0.008) - image<br> MaxEnt peaks.png-> MaxEnt probability peak areas (0.13 threshold)<br> Precipitation.png->Average precipitation 2000-2005<br> RiskMap.png-> New high-infection-rate risk map based on a 0.1° resolution MaxEnt model<br> RiskMap05.png->our previous risk map based on a 0.5° resolution MaxEnt model<br> riskmapcomparison.png-> Visual comparison between (a) our new high-infection-rate risk map based on a 0.1° resolution MaxEnt model and (b) our previous risk map based on a 0.5° resolution MaxEnt model.<br> RiskMapOverlap_mispredicted.png->highlight of low infection rate countries not predicted by our model<br> Temperature.png->Average Surface Air Temperature 2000-2005<br> time_series_covid19_confirmed_global.csv->World COVID-19 reports up to April 2021
本数据集涵盖复现本研究实验所需的全部信息，可基于气温、降水量与二氧化碳（CO2）浓度，生成分辨率为0.1°的高精度全球新冠感染风险分布图。所生成的风险指数地图可精准识别绝大多数实际高风险区域（准确率达87%），此类区域普遍具备中高CO2浓度、中低气温与中等降水量的特征。相较于本团队此前采用0.5°粗分辨率的模型（链接：https://zenodo.org/record/3945495#.YG7UEugzaUk），本全新模型在预测2020年同时存在高低感染率的真实场景时精度更优，准确率达80%。 数据与图像说明： comparisonvert.png → 本文件为模型输出可视化结果：(a) 使用MaxEnt平衡阈值（0.008）绘制的高感染率区域分布，(b) 概率峰值区域（阈值0.13），(c) 真实数据提取的低感染率国家与本研究风险地图的重叠区域，(d) 未被本模型正确预测的低感染率国家高亮展示 countries_high_rate.csv → 高感染率国家数据集 countries_low_rate.csv → 低感染率国家数据集 countries_low_rate_mispredicted.csv → 被本模型误判的低感染率国家数据集 covid_derivatives.csv → 全球各国新冠相关平均衍生指标数据集 covid_risk.csv → 风险地图数据集 gp.asc → MaxEnt分布栅格文件 LowDerivativeRegions.png → 低感染率国家分布图 MaxEnt distribution.png → 使用MaxEnt平衡阈值（0.008）绘制的高感染率区域分布图 MaxEnt peaks.png → MaxEnt概率峰值区域（阈值0.13）分布图 Precipitation.png → 2000-2005年平均降水量分布图 RiskMap.png → 基于0.1°分辨率MaxEnt模型生成的全新高感染率风险地图 RiskMap05.png → 本团队此前基于0.5°分辨率MaxEnt模型生成的风险地图 riskmapcomparison.png → 可视化对比结果：(a) 本研究基于0.1°分辨率MaxEnt模型生成的全新高感染率风险地图，与(b) 本团队此前基于0.5°分辨率MaxEnt模型生成的风险地图 RiskMapOverlap_mispredicted.png → 未被本模型正确预测的低感染率国家高亮展示图 Temperature.png → 2000-2005年平均地表气温分布图 time_series_covid19_confirmed_global.csv → 截至2021年4月的全球新冠确诊病例时间序列数据集
