Satellite Data for Version 2

Environmental monitoring using satellite data generates estimates of precipitation and was used to report monthly estimates for 23 inhabited atolls and islands using IMERG version 6 satellite data (National Aeronautics and Space Administration, 2023). In the initial study, we used 20 years (2001–20) of IMERG Final Run data to define baseline statistics for each of the 23 locations and IMERG version 6 Late Run data for the monthly reporting (Senay and others, 2023b). In this revision (version 2), we used IMERG Late Run data from the same sampling period to define baseline statistics, which ensured the monthly Late Run data were being compared with a baseline dataset that had the same descriptive statistical properties of central tendencies and dispersions. Additionally, in the initial version we filtered out two standard deviation outliers from the daily records in the 20-year baseline dataset before summing the daily data into monthly totals (Senay and others, 2023b). To fully reflect the variation of the baseline data, in this revision (version 2) we did not exclude statistical outliers from the IMERG Late Run baseline dataset before computing monthly sums and then statistics for each location. Instead, we identified six standard deviation daily peaks using Z scores computed from the same location and month, and, using a method similar to the “wet days” pre-processing method used by Beck and others (2019), replaced them with the local three-day average. This pre-processing also was applied to incoming data for the monthly reports. Finally, starting in June 2024 the National Aeronautics and Space Administration updated its IMERG data processing from version 6 to version 7 (Huffman and others, 2023). Thus, reports generated for January–May 2024 used version 6 data for both the baseline statistics and reporting month, and reports from June 2024 onwards used version 7 data for both the baseline statistics and reporting month.

Developing the Report Content for Version 2

The content of the monthly report consists of descriptive statistics assembled from the IMERG data plus visual representations of those statistics. The report produced each month was a summary of the estimated precipitation of the previous month. Because the monthly reports were designed to supplement existing precipitation, drought, and agroforest management products, the reports do not include predictive statistics or interpretation of the effect of the estimated precipitation on or potential risk to vegetation health or crop production. The monthly reports and included data are available and could aid with operational efforts for drought monitoring in the Marshall Islands as well as the U.S.-Affiliated Pacific Islands or all of Oceania.

During the development phase, the team agreed to have two rainfall indices that would serve as a frame of reference for how typical or unusual the monthly precipitation was (rationale and methodological details are provided by Senay and others [2023b]). One index used is the Pacific ENSO Application Climate Center (PEAC) 4/8-inch index (Pacific ENSO Applications Climate Center, 2015). No changes were made to the PEAC 4/8-inch index in version 2. A color ramp using the threshold values of 4 and 8 inches was applied; monthly precipitation less than 4 inches was colored red, 4–8 inches was colored yellow, and greater than 8 inches was colored dark green. The resulting PEAC map is presented in the lower panel of information graphic 1 in the USGS data release (Senay and others, 2023a) in version 2.

The second index used is the rainfall index. Precipitation was categorized along a nonparametric scale using quintiles derived from the 20-year IMERG baseline data. Methodological details are provided by Senay and others (2023b). The quintiles for the rainfall index are as follows: the bottom quintile (0–19 percent) consists of the lowest precipitation values (driest), the next quintile (20–39 percent) represents drier conditions, the middle quintile (40–59 percent) includes the median and spans the most typical precipitation values (typical), the next quintile (60–79 percent) is a little wet compared to the typical conditions, and the highest quintile (80–100 percent) consists of the highest precipitation values (wetter). Each of the quintiles is associated with a color as a quick indicator of how the current month’s precipitation compares to the statistics of the 20-year database, ranging from brown (drier quintile) to deep blue (wetter quintile). The resulting rainfall index map is presented in the upper panel of information graphic 1 in version 2.

In the initial approach, the management partners wished to focus on detecting unusually dry or wet conditions. Therefore, the precipitation of any given month was compared against the long-term (2001–20) data from that location and month. Thus, a set of descriptive statistics was calculated for each month at each of the 23 locations using the 20-year IMERG Final Run database. During the first year, the team identified two challenges created by the “same place same month” comparisons. The first challenge was that the long-term baseline dataset was created using IMERG Final Run data but compared with IMERG Late Run monthly estimates for timely reporting. Differences in the statistics of the Final Run versus Late Run data resulted in many, if not most, months being classified as “wet” or “wetter” than typical in the rainfall index, which rendered the rainfall index uninformative. Therefore, in this revision (version 2) we used IMERG Late Run data from 2001 to 2020 for the baseline data, as well as the monthly reporting data, which ensured the statistics of the baseline data (central tendencies, dispersions) would be most representative of the same modeling distribution from which the monthly precipitation estimates were derived.

The second challenge experienced by the team was lack of agreement between the two precipitation indices. The PEAC index (Senay and others, 2023b) reflected the seasonal and latitudinal variation in precipitation that is well-documented in the RMI (Stoddart and Walsh, 1992; Leong and others, 2014). In contrast, as described above, the rainfall index for most locations remained at “wet” or “wetter” than typical even during the winter/spring dry season. Although the team had discussed this possibility in the initial development process (refer to “Developing the Report Content” in Senay and others, 2023b), the users had not anticipated how challenging it would be to have two indices that did not seem to agree.

Therefore, we made a noteworthy change in the statistics used for the rainfall index. The team decided that communicating patterns of precipitation for each location, including “normal” seasonal dry/wet trends, would result in greater agreement between the rainfall index and the PEAC 4/8-inch index thresholds, which was desirable for our users. We pooled data for all months for each location and computed location-specific rank-order quintiles for each of the 23 locations. Pooling the data across months resulted in changing the protocol from a rainfall index that categorized monthly precipitation from the same place and same month to version 2, which categorized monthly precipitation from the same place as compared across all months from that place during the 20-year baseline data period. In version 2, the “Typical” category in the rainfall index graphics was re-labeled as “Local Median” in the upper panel of graphic 1 to communicate this change to same-place all-months baseline statistics (Senay and others, 2023a).

Developing the Layout for the Report Template Version 2

At the core of the monthly reports are two estimated precipitation values, two colorized rainfall indicators, and a bar graph with the precipitation history for recent months. This information is delivered in multiple ways that provide different quick-look summaries of the precipitation status. In this revision cycle, we modified the content and layout that was developed in the initial work (Senay and others, 2023a, 2023b). Information graphics in version 2 are numbered to facilitate the co-production process and referenced in this report.