Multi-Year Trends in Chlorophyll-a and Sea Surface Salinity

This analysis investigates the relationship between Sea Surface Salinity (SSS) and Chlorophyll-a concentrations in the Mississippi River outflow into the Gulf of Mexico. The study utilizes monthly measurements from January 2012 to December 2022, across 25 locations. By looking at these measurements, we aimed to understand how these two factors change over time and how they might influence each other.

Sea Surface Salinity refers to the concentration of dissolved salts in the upper layer of the ocean and is a critical oceanographic parameter that influences marine ecosystem dynamics. Chlorophyll-a, the primary photosynthetic pigment found in phytoplankton, serves as a key indicator of phytoplankton biomass and overall ocean productivity. Recent literature acknowledges a strong relationship between chlorophyll-a and sea surface temperature but presents divergent findings regarding sea surface salinity at environmentally distinct locations. This study may benefit academia by enhancing the understanding of how chlorophyll-a and sea surface salinity interact spatially and temporally across different latitudes.

Data was sourced from NASA's Multi-Mission Optimally Interpolated Sea Surface Salinity (OISSS) dataset and the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Terra satellite. The OISSS dataset provides global monthly means of sea surface salinity on a 0.25-degree spatial grid, derived from missions including Aquarius/SAC-D, Soil Moisture Active Passive (SMAP), and Soil Moisture and Ocean Salinity (SMOS). The MODIS-Terra dataset offers chlorophyll-a concentrations at a spatial resolution of 4 km, facilitating detailed observation of phytoplankton distribution. The next step for beginning analysis was to create boundaries for analysis in which 480km of the watershed was selected. Within this boundary, 25 sites were systematically chosen, and monthly descriptive statistics (including mean, minimum, maximum, and median) were calculated for SSS and Chlorophyll-a from 2012 to 2022.

Visualizing measurement locations in the Gulf of Mexico region from Mississippi river outflow. To understand how the distance from shore impacts the relationship between salinity and chlorophyll-a, we plotted the measurement locations and calculated their distances from the coastline. This map helps us visualize where each point lies relative to the coast.

The median values of Chlorophyll-a and Sea Surface Salinity were calculated for the region of study at each month. Using the median helps reduce the influence of extreme values and offers a straightforward way to spot patterns and trends in the region of study before moving into more advanced analysis.

Here, we examine how salinity and chlorophyll-a levels vary over time. A double Y-axis chart, where both variables are scaled, allows for direct visual comparison:

Additional boxplots and seasonal plots detail month-to-month and year-to-year behavior, confirming a clear seasonality in both variables. Below, note that box plots on the left are aggregated from data points which can be seen on the right.

The data reveal distinct seasonal patterns in chlorophyll and salinity, with some locations showing stronger variations than others. For example, chlorophyll levels often peak during summer months and sea surface salinity levels drop in summer months.

Both monthly Chlorophyll-a and Sea Surface Salinity distributions demonstrate pronounced seasonal variability.

Time series data can have patterns that persist over time, making statistical modeling challenging. To test for stationarity (whether the data fluctuates around a constant mean and variance), we:

Examined the Autocorrelation (ACF) and Partial Autocorrelation (PACF) plots. These show how past values influence current ones. Conducted the Augmented Dickey-Fuller (ADF) test, which confirmed whether the time series is stationary.

Augmented Dickey-Fuller Test chlor_a_median Dickey-Fuller = -4.2033, Lag order = 5, p-value = 0.01 alternative hypothesis: stationary

sss_median Dickey-Fuller = -5.6019, Lag order = 5, p-value = 0.01 alternative hypothesis: stationary

ADF Test Results: For both chlor_a and SSS, the p-values (<0.01) indicate rejection of the null hypothesis, suggesting stationarity, supporting their suitability for time series modeling without the need of differencing.

1. We analyzed the correlation between chlorophyll-a and sea surface salinity at each measurement location, grouped by their distance from the coastline.

• Correlations range from weak to moderate negative values across most locations, with some exceptions where the correlation is closer to zero or slightly positive.
• The relationships are predominantly negative (23 out of 25 stations show negative correlations), with correlation coefficients ranging from -0.516 to 0.192, indicating that higher chlorophyll-a concentrations generally correspond to lower sea surface salinity levels, with the strongest relationship (r = -0.516) observed at station 2816, located 200 km from the coast.
• Of the 25 stations analyzed, 17 showed statistically significant correlations (p < 0.05), with the strength of these relationships generally decreasing with distance from the coast - stations closer to shore (within 100 km) typically showing moderate negative correlations (-0.3 to -0.4), suggesting a stronger coastal influence on the chlorophyll-a and salinity relationship.

Analysis of chlorophyll-a and sea surface salinity relationships in the study region reveals strong seasonal patterns across 25 locations (2012-2022). At each location, monthly correlations were calculated using the 11 years of data for that specific month. Summer months show the strongest inverse relationships, with August having the most negative mean correlation (-0.452) and the highest number of locations (11 out of 25) showing significant correlations over the study period.

Building on the observed seasonal and temporal trends, I used a SARIMAX (Seasonal Autoregressive Integrated Moving Average with Exogenous Variables) model to predict Chlorophyll-a median, with SSS median included as an external regressor. I selected the final ARIMA(1,0,0)(2,0,0)[12] structure through a stepwise search based on the Bayesian Information Criterion (BIC). This model design captures immediate monthly autocorrelation through AR(1) and a 12-month cycle via Seasonal AR(2).

Model Context: Our SARIMAX analysis reveals that across the 25 locations in the study area, increases in regional median sea surface salinity are associated with decreases in regional median chlorophyll-a concentrations. Specifically, for every 1-unit increase in median SSS, median chlorophyll-a is expected to decrease by 0.0436 units (p < 0.01), while accounting for seasonal patterns and temporal autocorrelation in the data.

The negative relationship between SSS and chlorophyll-a represents the true association between these variables after, removing the influence of the previous month's chlorophyll-a values on the current month and accounting for recurring annual patterns in chlorophyll-a that repeat every 12 months. This ensures the relationship we're observing isn't just due to temporal dependencies or seasonal cycles in the data.

1 Year Forecast with SARIMAX Model:

Forecasting one year ahead shows the model's overall performance, though some summer months deviate more than expected.

This research offers an initial look at how the Mississippi River shapes ocean conditions in the Gulf of Mexico by comparing Sea Surface Salinity and Chlorophyll-a from 2012 to 2022. Strong spatial gradients emerge, with higher Chlorophyll-a values concentrated nearshore and rising salinity further offshore. The data also exhibit pronounced seasonal cycles: Chlorophyll-a typically peaks during summer, while salinity often drops in these months. Cross-correlation results confirm a largely negative relationship between the two variables and underscore the strength of this pattern in the warmer seasons.

By employing a SARIMAX model, I was able to quantify this inverse association and show that higher salinity reliably corresponds to lower Chlorophyll-a, even after accounting for previous observations and seasonal cycles. Although the model performed well, it relied on aggregated median values for the region, and applying it directly at each measurement location could uncover site-specific patterns and nuances.

Data for this project was gathered from NASA Earth Data, specifically the OISSS and MODIS datasets, available through the  NASA EarthData portal  and the  Ocean Data website .

OISSS Dataset: Earth and Space Research (ESR). 2023. Multi-Mission Optimally Interpolated Sea Surface Salinity Global Monthly Dataset V2. Ver. 2.0. PO.DAAC, CA, USA. https://doi.org/10.5067/SMP20-4UMCS.

MODIS-Terra Dataset: NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Ocean Biology Processing Group. Moderate-resolution Imaging Spectroradiometer (MODIS) Terra Chlorophyll Data; NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Ocean Biology Processing Group https://doi.org/10.5067/TERRA/MODIS/L3M/CHL/2022.

About the Author Jacob Spier is a senior at California State University, Northridge (CSUN), pursuing a Bachelor of Science in Computer Science with a minor in Data Science. Contact: jacob.spier.809@my.csun.edu

Acknowledgements: Sincere thanks go to Mario Giraldo (CSUN), Swany Cuc (CSUN), Jorge Vazquez (JPL), Latha Baskaran (JPL), Joe T. Roberts (JPL), Dustan Levenstein, and Adriano Zambom (CSUN) for their support and expertise.