Land Cover Pattern and Case Fatality Rate of COVID-19

: Coronavirus disease 2019 (COVID ‐ 19) has already caused 1,405,029 deaths worldwide, as of November 25 th , 2020. Assessing whether land cover in people’s living environments affects COVID-19 health outcomes is an urgent and crucial public health problem. Here, we examine land cover data associated with the case fatality rate (CFR) of COVID-19 at the county-level, in the United States. A 1% increase in green space in the county is associated with a statistically significant 0.34% (95% confidence interval 0.13%-0.55%) decrease in the county’s COVID-19 CFR, and a 1% increase in emergent herbaceous wetlands are correlated with a 1.65% (0.19%-3.11%) decrease in the CFR. In addition, a 1% increase in high intensity developed area among the total developed area is related to a significant 3.63% (2.14%-5.12%) increase in the CFR, while a 1% increase in medium intensity developed area is associated with a 0.75% (-0.02%-1.51%) decrease. Our research highlights that governments could prevent similar pandemics in the future and even achieve some sustainable development goals by decreasing development intensity and increasing green space in living environments.


Introduction
Coronavirus disease 2019 (COVID-19) has raised serious and urgent concerns globally 1,2 . As of November 25 th , 2020, there were over 59.49 million confirmed cases and 1,405,029 deaths due to COVID-19 worldwide (Data from WHO COVID-19 Dashboard, see https://covid19.who.int/). In the United States, the cumulative numbers of confirmed and death cases owing to COVID-19 are 12,502,262 and 257,615, respectively, as of November 25 th , 2020 (Data from US Centers for Disease Control and Prevention). However, the COVID-19 case fatality rate (CFR, CFR definition from WHO, see https://www.who.int/news-room/commentaries/detail/estimating-mortalityfrom-covid- 19), which is the ratio of the numbers of deaths to confirmed cases, varies significantly among counties, ranging from 0 to 18.18%. A low COVID-19 CFR means that people are likely to recover, after they get infected, and that the local population may have few other chronic diseases 3,4 and relatively good mental health 5,6 . Cardiovascular disease is associated with an increased risk of death in COVID-19 patients 3 ; people experience less depression and anxiety, exposed to more green space during COVID-19 5 . Therefore, analyses on the associations between potential environmental factors and the COVID-19 CFR may help identify the high-risk areas and develop optimal land-use policies to deal with other similar public health emergencies in the future 7 .
Investigations regarding the relationships between COVID-19 health outcomes and geographical factors are urgently needed, to locate the high-risk areas, to slow the disease's devastation, and to lower the risk of similar infectious diseases outbreaks [8][9][10] .
People's living environments have effects on the severity of COVID-19 among individuals, because people with less greenness may have more medical conditions 11 , like cardiovascular disease [12][13][14] , which would ultimately exacerbate the symptom of COVID-19 3 . Numerous researchers point out that neighborhood greenness may have an effect on health outcomes, by promoting physical exercise and social connections, relieving stress, and removing air pollution, noise, and heat exposure 15 . Moreover, an increasing amount of neighborhood greenness is related to reduced risks of chronic diseases, such as respiratory, cerebrovascular, cardiovascular, among others 16 . It is hypothesized that living with more green land cover reduces the risk of death after infected, by alleviating the severity of COVID-19 symptoms. In other words, increased exposure to green spaces is associated with decreased risks of clinical diseases, which are related to a decreased risk of death after infection.
The positive effects of green land cover on COVID-19 are widely discussed, but urban land impacts remain inconclusive. In fact, an increase in urban land reduces green land cover, suggesting that the impacts of urban land on COVID-19 may be harmful.
Exposure to greenery is associated with improved mental health during the pandemic 5 .
Also, urban vegetation is related to a decrease in the prevalence of COVID-19 in the United States 17 . Many studies on this topic are available as preprints. However, due to the suddenness of pandemic, data insufficiency is always a severe problem hindering research. Specifically, individual-level data on COVID-19 health outcomes are still publicly unavailable or inaccessible to academics. Therefore, some models that generally provide more accurate estimation, like proportional hazards models, are unable to be utilized. In order to analyze the association between the COVID-19 CFR (shown as Figure 1) and land cover (illustrated in Figure 2, 3 and 4), linear regressions are employed using the county-level data, while controlling for other county-level variables. Here, we explore how multiple land cover factors (percentages of land types in the total area, percentages of development intensity in developed area) affect the COVID-19 CFR across counties in the U.S. Our results highlight the importance of green land cover and vegetation in the developed area to reduce the risk of death after infection.

CFR of COVID-19
The CFR in each county of the United States is used as the dependent variables in the analyses, which is the ratio between the numbers of death cases and confirmed cases.
The CFR is typically considered a measure of disease severity, where a higher CFR means that people are more likely to die after they are infected. Census Bureau, and then they are transformed into percentages for each land type.
Additionally, the percentages of green-blue-grey land cover in the counties are applied.
We divide the 16 land types into four categories: green land cover, blue land cover, grey land cover, and other land cover (Table 1). Finally, the percentages of developed land cover with different development intensities in the total developed area of each county are also calculated, which are the ratio of the area with certain development intensity and the total developed area in the county. Note: Wetland is considered green space in land use categories, but wetland is easier to confuse with water during remote sensing. Therefore, woody wetland and emergent herbaceous wetlands are classified as blue space, rather than green space, in this study.  Table S1: Data Statistic Summary, Table S2: Data Source, Figure S1.1 -S1.4: Linear Trends between Dependent

Other Demographic, Socioeconomic and Environmental Variables
Variable and Independent Variables)

Linear Regression Model
The following equations are built to analyze the effects of land cover variables on the COVID-19 CFR in each county, while controlling for other county-level demographic, socioeconomic, and environmental characteristics: where represents the COVID-19 CFR of county , represents a vector of land cover data of county , represents a vector of demographic, socioeconomic and environmental variables of county as control variables, and represents the error term. In this model 0 , 1 and 2 are parameters to be estimated.

Spatial Autoregressive Model (SAR)
The SAR model assumes that one observation's dependent variable is associated with its neighborhoods' dependent variable. In our study, the COVID-19 CFR of a specific county is related to the CFR of the counties surrounding it 24 . The SAR model is as follows: where represents a vector of spatial weights of neighboring regions of county , represents a vector of the CFR of neighboring county , and represents the spatial lag parameter to be estimated. To obtain the spatial weight vectors of each observation, we use the queen method 25 . In the queen method, two polygons are considered contiguous, if they share one point.

Quantitative Effects of Land Cover Factors on the COVID-19 CFR
To estimate the quantitative effects of land cover factors on the COVID-19 CFR, we build an equation to calculate the case fatality rate ratio (CFRR), which is the ratio of the change in the CFR after adjusting for land cover and the current CFR in the U.S.: where represents the ratio of the CFR after adjusting land cover and the current CFR, 1 represents estimated the parameter of land cover , ∆ represents the percentage of adjustment of land cover , and represents the COVID-19 CFR in the U.S. Assuming that ∆ is 1%, the equation is as follows: The CFRR for certain land types can be interpreted as a relative increase or decrease in the COVID-19 CFR associated with a 1% increase in this land type in a county. which is similar to the estimation of linear regression analysis.

CFRR of Each Land Cover Variable
The CFRR of each land cover variable is estimated, based on the parameters of the regressions and the COVID-19 CFR in the U.S., using Equation (IV). The CFRRs of land cover variables, their 95% confidence intervals (CIs), and the p-value are listed in To conclude, the percentage of green space is negatively associated with the COVID-19 CFR, and the type of green space also influences the degree of its impact on the COVID-19 CFR. In addition, low intensity development is associated with a low CFR. In other words, high intensity development aggravates the severity of infectious diseases, such as COIVD-19, which might increase both their morbidity rate and their CFRs. More greenness in the county is associated with a decreased risk of death from COVID-19 after infected among Americans. In contrast, more high intensity development is related to an increased risk of death, because of lack of green space.  11 . Therefore, these strategies would also prevent outbreaks of other diseases in the future. In this way, an increase in green space and reducing development intensity, at least, help achieve SDG 3 (good health and well-being) and 11 (sustainable cities and communities).
There are some limitations worth noting in this study. Firstly, some potential factors may be overlooked or unable to be obtained, although we have already controlled 27 county-level variables. Secondly, the resolution of land cover and the lag of these data increase the uncertainty of the results, because the land cover data are from 2016 with a resolution of 30 meters 34,35 . Thirdly, the COVID-19 data are county-level data possibly resulting in ecological fallacy. Future studies are better to use finer-scale data, or even individual-level data, to detect the casual interpretation of associations discussed in this article. Additionally, the specific benefits and costs of increasing green space and reducing development intensity to achieve SDGs need further estimation.

Conclusion
Our results indicate that a 1% increase in green space is associated with a 0.34% decrease in the county's COVID-19 CFR, and this relationship is influenced by the type of green space. In addition, a 1% increase in high intensity developed area among the total developed areas is related to a 3.63% increase in the CFR, while a 1% increase in medium intensity developed areas is associated with a 0.75% decrease. Our research highlights that governments could prevent other pandemics in the future and even achieve some SDGs by decreasing development intensity and increasing green space in living environments.