*2.2. Data Acquisition and Methodology*

A Keyhole Markup Language Zipped (KMZ) file locating the wild blueberry fields of Maine was produced based on a field survey carried out by the University of Maine Cooperative Extension. The polygons of the 89 wild blueberry fields (area > 0.06 km2) in the major wild blueberry production region, Maine (Figure 1a,b) were acquired from the KMZ file. The Airport and Baxter field polygons were acquired from a KMZ file based on a field survey by Jasper Wyman & Son, Deblois, Maine.

In this study, the SPEI was used as a drought index, which is calculated based on precipitation, temperature, and potential evapotranspiration data. The SPEI was chosen over another popular drought index, the PDSI, because the PDSI has a fixed temporal scale (between 9–12 months), which prevents the understanding of drought severity on different temporal scales [28]. In addition, the SPEI was chosen over the widely used SPI because the SPI only considers precipitation data and does not include air temperature and evapotranspiration data, which could also significantly influence the understanding of drought impacts on agriculture [19]. The SPEI's multi-scalar character enables it to detect, monitor, and analyze droughts more effectively, as it can quantify the drought severity according to its duration and intensity [19]. The SPEI allows the comparison of drought severity through time and space, since it can be calculated over a wide range of climates. In this study, the SPEI data were collected from the readily available open access database "SPEI Global Drought Monitor" (https://spei.csic.es/map/maps.html, accessed on 10 July 2020) in netcdf format. The netcdf files containing the SPEI data were transferred to ArcGIS Pro Version 2.7 (ESRI, Redlands, CA, USA) to acquire the SPEI data for the study sites (Airport, Baxter, and major wild blueberry region of Maine) over 71 years from 1950 to 2020, using the zonal statistics tool in ArcGIS Pro (Figure 2). The SPEI data were acquired on different temporal scales ranging from 1 month (SPEI\_1) to 48 months (SPEI\_48). These data were provided on a per-pixel basis at a 4 km spatial resolution. The SPEI (SPEI\_6 of September) of only the growing season (April–September) was considered in this study. SPEI\_6 of September represents the water conditions of a growing season (April–September). To understand the long-term (multi-year) impact of water conditions on the vegetation indices (EVI and NDVI) and yield of wild blueberry crops, the average

SPEI (SPEI\_6 of September) of two, three, and four consecutive years was also calculated. A positive SPEI value represents wet conditions, whereas a negative SPEI value indicates dry conditions.

**Figure 2.** A flowchart showing the steps of data acquisition and analyses for this study.

The dataset of climate variables, such as total precipitation and mean temperature, during the growing season (May to September) over 71 years from 1950 to 2020 for the study sites (Airport/Baxter and the major wild blueberry region of Maine) was acquired from the online tool "Climate Engine" (https://clim-engine.appspot.com/climateEngine, accessed on 17 July 2021) of the Desert Research Institute, University of California, USA. Here, total precipitation refers to an average of monthly total precipitation (mm), and mean temperature refers to an average of the monthly mean air temperature at 2 m from the ground surface for the growing season. The original data sources for the climate variables were obtained from the AN81 m dataset of the PRISM Climate Group (https: //prism.oregon-state.edu/explorer/, accessed on 17 July 2021). These data were provided on a per-pixel basis at a 4 km spatial resolution for the conterminous United States with a temporal resolution of one month (daily mean temperature and total precipitation were averaged monthly). This AN81 m dataset is available from January 1895. The extracted data were transferred into Microsoft Excel (Microsoft, Redmond, WA, USA) to calculate the average total precipitation and the mean temperature of the summer months (May to September) of each year (Figure 2).

In order to quantify vegetation responses to drought, satellite-based remotely sensed EVI and NDVI data for 21 years (2000 to 2020) of the studied wild blueberry fields were acquired from Google Earth Engine. These data were originally obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS) product MOD13Q1 (https: //lpdaac.usgs.gov/products/mod13q1v006, accessed on 25 July 2021). The MOD13Q1 dataset is preprocessed as well as readily and freely available. The MOD13Q1 Version 6 data

have a spatial resolution of 250 m, generated every 16 days. For the development of the MOD13Q1 product, an algorithm was used to select pixels with low clouds, low view angle, and maximum index value to obtain the best available pixels over the 16-day-period image acquisitions [29]. The MOD13Q1 product has two vegetation layers: NDVI and EVI. The NDVI is the most common one used for characterizing canopy leaf chlorophyll content based on the reflectance contrast between the red and the near-infrared (NIR) wavebands [30]. However, the NDVI has some limitations, such as (1) it saturates in dense vegetation, (2) it does not consider the canopy background noise, and (3) its ratioing properties to eliminate noise [31]. These limitations were improved in the EVI to some extent, and thus the EVI has several advantages over the NDVI as it has improved sensitivity over high biomass regions [31]. This dataset is readily available in the Google Earth Engine. The vegetation indices values over the summer months were extracted for the study sites using a JavaScript-based API in the Google Earth Engine (https://code.earthengine.google.com/, accessed on 25 July 2021) using the extraction command "ui.Chart.image.seriesByRegion". The extracted data were transferred into Microsoft Excel (Microsoft, Redmond, WA, USA) to calculate the average EVI and NDVI of the summer months (May–September) of each year (Figure 2).

The historical yield data of Maine were collected from the United States Department of Agriculture (USDA), National Agricultural Statistics Service using a Quick Stats Ad-hoc Query Tool (https://quickstats.nass.usda.gov/, accessed on 26 July 2021). Historical yield data of the entire state of Maine (million lbs.) were available from 1924 to 2020, but the yield per production area data (lbs./acre) were only available from 2012 to 2020 (except 2013). It should be noted that the yield data were considered from all over the state of Maine, where ~90% of the yield was typically from the major wild blueberry production region (Washington and Hancock counties). The historical yield data of the Airport and Baxter fields at Deblois, Maine were provided by Jasper Wyman & Son, Maine. The yield (lbs./acre) data for the Airport and Baxter fields were available from 1993–2019 for every alternate year (except 2001 for the Baxter field).
