Time-series Analysis of Remotely-sensed Seawifs Chlorophyll in River-influenced Coastal Regions

The availability of a nearly-continuous record of remotely-sensed chlorophyll a data (chl a) from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) mission, now longer than ten years, enables examination of time-series trends for multiple global locations. Innovative data analysis technology available on the World Wide Web facilitates such analyses. In coastal regions influenced by river outflows, chl a is not always indicative of actual trends in phytoplankton chlorophyll due to the interference of coloured dissolved organic matter and suspended sediments; significant chl a timeseries trends for coastal regions influenced by river outflows may nonetheless be indicative of important alterations of the hydrologic and coastal environment. Chl a time-series analysis of nine marine regions influenced by river outflows demonstrates the simplicity and usefulness of this technique. The analyses indicate that coastal time-series are significantly influenced by unusual flood events. Major river systems in regions with relatively low human impact did not exhibit significant trends. Most river systems with demonstrated human impact exhibited significant negative trends, with the noteworthy exception of the Pearl River in China, which has a positive trend. INTRODUCTION Large river systems respond to natural and anthropogenic changes in the regional watershed that provides their source of water. River outflow in coastal regions influences the coastal marine environment in several ways through the delivery of fresh water, nutrients, dissolved organic matter, and suspended sediments. Monitoring of trends in such constituents in coastal regions influenced by river outflow can thus indicate or identify alteration of the regional environment, both inland and along the coast. Remotely-sensed chlorophyll a (chl a) from satellite sensors provides a method to examine trends in the coastal environment without the necessity of in situ sampling. Because chl a data in coastal regions essentially constitutes an indicator of the ambient ocean optical environment, identification of trends in this remotely-sensed variable can be potentially linked to terrestrial and human interactions in the river watershed (1,2). In order for such trends to be valid, the long-term accuracy of the data, by means of sensor calibration and ongoing data validation, must be reliable. It should be noted, however, that even though long-term accuracy of the chl a data parameter has been maintained, in coastal waters remotely-sensed chl a is a less accurate indicator of actual chlorophyll concentrations than in the open ocean due to increased turbidity (primarily coloured dissolved organic matter (CDOM) and suspended sediments) in coastal waters. Although human actions will affect the entire ocean through long-term processes, rivers are frequently sensitive indicators of human activities. Numerous studies show that anthropogenic factors affect river flow as well as the nutrient and sediment supply, which in turn affects composition of species and ecosystem elements such as water and nutrient cycling (3,4). Rural EARSeL eProceedings 8, 2/2009 115 inputs, such as deforestation and agricultural production (5); and urban inputs, such as sewage systems and vehicular exhaust (6) cause increases in nutrients in the river. In addition, climate change can be related to changes in river patterns and nutrient loads (7,8). Increases in nutrient loading can threaten the ecological conditions in coastal zones surrounding the rivers through increased phytoplankton productivity, leading to eutrophication, hypoxia (low dissolved oxygen in bottom waters), and anoxia (zero dissolved oxygen in bottom waters) (9,10). The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) provides a nearly continuous, consistent and reliably-calibrated record of remotely-sensed chl a now exceeding 10 years in duration. Maintenance of SeaWiFS calibration, which is necessary for the maintenance of consistent chl a retrievals, has been accomplished through a series of monthly lunar calibration maneuvers that utilize the lunar radiance as a constant radiance source (11). Ongoing data validation efforts using research cruise data and data from moored optical sensors are also utilized to maintain consistent chl a retrievals. Analysis of SeaWiFS data has shown a 4% global increase in chlorophyll from 1997 to 2003 with most of the increases occurring in the coastal zones (12). This study represents an initial systematic examination of different oceanic river outflow regions across the globe from 1998 to 2007. As Earth remote-sensing data becomes increasingly accurate and spans longer continuous periods, analysis of coastal zone trends there is a greater need for easy access to this data for both scientists and interested citizens. The GES DISC Interactive Online Visualization And aNalysis Infrastructure (Giovanni) allows data acquisition and application of basic analytical functions using only a Web browser (13). By eliminating the barriers between data access and scientific inquiry, Giovanni can enable further research into coastal ocean trends and related datasets such as precipitation, wind, and atmospheric conditions. The usefulness of Giovanni to examine various aspects of the coastal zone has already been described (14). Through its simplicity of use, Giovanni can provide the first step for scientific analysis of remotelysensed chl a data. Several studies have used Giovanni to map chl a in regional studies, such as the Red Sea (15) the Chesapeake Bay (16) and seasonal variations in the northern South China Sea (17). By demonstrating the ease of chl a time-series analysis near river outflows, research to determine the causative factors involved with coastal chl a trends may be inspired, as well as more detailed research utilizing both remote sensing and in situ data. Such research can be useful both for coastal zone monitoring and also to improve understanding of the relationship between remotely-sensed chl a and in situ chlorophyll measurements in optically complex coastal waters. River outflow influence on coastal waters Rivers integrate environmental change factors, both natural and anthropogenic, over their entire watershed, giving insight into the changes occurring on the land. The regional climate, hydrology and climate changes within the watershed determine the seasonality, size, and outflow of the river. Topography plays an important role in the discharge and the type of sediment transported. Humans can affect chlorophyll concentrations in the outflow region in two ways: via higher nutrient loading, particularly nitrogen and phosphorous, or by influencing the mean flow and sediment load. Increases in nutrients will commonly lead to increased phytoplankton productivity and thus an increase in remotely-sensed chl a. Direct nutrient loading usually occurs due to land-use changes. Increases in nitrogen-rich fertilizer, phosphorous-rich sewage from urban waste, and run-off due to land-use change and deforestation are a few examples of direct nutrient loading. It is often hard to fully characterize these anthropogenic factors and even harder to quantify their effects. Indirect nutrient loading is a result of changes in discharge and sediments, which carry nutrients. Discharge and sediment load provide insight into the nutrients entering the ocean from the river, especially through trends over time. In most cases, water discharge and sediment load exhibit a strong correlation. However, human-built reservoirs can skew this correlation. Reservoir dams can EARSeL eProceedings 8, 2/2009 116 block sediment transport and restrict discharge, causing decreases in nutrient delivery to the ocean. Our study focuses on examining river outflow trends for significance and outlining some watershed alterations that may be related to these trends. We present the time-series results in conjunction with brief general descriptions of each river and river outflow region. This information outlines average discharge and sediment load, seasonality, and any human interference through dams. These are some of the primary factors that can lead to observable significant trends in coastal waters influenced by river systems. METHODS SeaWiFS global Level 3 monthly data products are acquired as they are generated by the NASA Ocean Biology Processing Group (OBPG) and ingested into the Giovanni system. SeaWiFS Level 3 data products are binned and averaged to global 9 × 9 km resolution on an equal-area grid. Giovanni allows averaging of the mapped 0.083° × 0.083° grid values in any user-selected region. Average monthly chl a time-series are generated over the user-selected time period by generating the average chl a value for the selected region and then plotting each consecutive average value. Giovanni generates the arithmetic mean of the data values within the selected region. In this study, Giovanni was employed to generated chl a time-series for nine distinct river outflow systems (The Mississippi River has both a western and eastern time-series due to the strong influence of a western-flowing coastal current). Figure 1 shows the selected areas chosen. Table 1 provides the corner coordinates for each region. Each area is a 0.5° × 0.5° region as close as possible to the river mouth, but possessing an average chl a concentration over the entire time-series not exceeding 1 mg/m. This concentration range was selected to reduce the influence of turbidity and CDOM, although their influence cannot be ruled out, particularly during high flow regimes, due to the known difficulty of discriminating CDOM from chl a in SeaWiFS data. In the case of large rivers, the area of ocean influenced by the river plume can be extensive; we utilized Giovanni to examine yearly averaged regional images to establish the apparent spatial extent of river-influenced waters. The Congo River-influenced region, for example, extends to at least 0° latitude from the river mouth location at 12° E latitude, approximately 1300 km; the Congo River study area was 1100 km from the river mouth. Because the lower concentration zones are situated near the boundary of the river-influenced region, alterations of flow regime or nutrient content would be expected to more significantly influence transitional regions than deeper in the river-influenced zone where conditions are less variable. The Giovanni ASCII text output of time-series values was input into a MS Excel spreadsheet. Chl a time-series were generated over a nine-year time span. Linear regression was then applied to these complete monthly datasets using the least-squares linear fit method. This methodology is a simple way to perform trend analysis despite the potential sensitivity of the computational method to outliers. Significance of the linear model was determined by a 95% confidence interval (p <. 05) using ANOVA analysis (18). The significance of the overall trend (and whether it was linear) was evaluated using the f-test. The r-value (coefficient of determination) was used to indicate how well the regression line fitted the actual data. In addition, we created plots of the yearly average of monthly chl a anomalies to mirror the 2005 study of global trends in chl a concentrations (12). These trends were calculated by 1) subtracting the average mean monthly value from each month which created monthly anomalies, 2) creating yearly anomalies by averaging the monthly anomalies for each year, and 3) applying regression analysis (Figure 15). EARSeL eProceedings 8, 2/2009 117 Figure 1: Map of the nine river outflows that were evaluated for this study. Table 1: Latitude-longitude corner coordinates (in digital format) for the ten river-influenced 0.5° x 0.5° regions for which chl a time-series were generated. RIVER SYSTEM North South West East Amazon 8.00 7.50 -51.00 -50.50 Congo -5.00 -5.50 2.00 2.50 Eel 41.50 41.00 -125.90 -125.40 Ganges 20.30 19.80 89.40 89.90 Guadalquivir 36.50 36.00 -6.80 -6.30 Mississippi (East) 29.60 29.10 -86.75 -86.25 Mississippi (West) 28.75 28.25 -94.00 -93.50 Orinoco 12.50 12.00 -64.80 -64.30 Pearl 22.50 22.00 117.50 118.00 Po 44.50 44.00 13.50 14.00 EARSeL eProceedings 8, 2/2009 118 RESULTS The results of each time-series analysis are presented subsequently. A brief description of the river system and potential influential factors in the watershed is presented with each time-series plot and statistical summary.