Just Add Water and the Colorado River Still Reaches the Sea

Abstract

A recent article in Environmental Management by All argued that flood flows in North America’s Colorado River do not reach the Gulf of California because they are captured and evaporated in Laguna Salada, a below sea-level lakebed near the mouth of the river. We refute this hypothesis by showing that (1) due to its limited area, the Laguna Salada could have evaporated less than 10% of the flood flows that have occurred since 1989; (2) low flow volumes preferentially flow to the Gulf rather than Laguna Salada; (3) All’s method for detecting water surface area in the Laguna Salada appears to be flawed because Landsat Thematic Mapper images of the lakebed show it to be dry when All’s analyses said it was flooded; (4) direct measurements of salinity at the mouth of the river and in the Upper Gulf of California during flood flows in 1993 and 1998 confirm that flood waters reach the sea; and (5) stable oxygen isotope signatures in clam shells and fish otoliths recorded the dilution of seawater with fresh water during the 1993 and 1998 flows. Furthermore, All’s conclusion that freshwater flows do not benefit the ecology of the marine zone is incorrect because the peer-reviewed literature shows that postlarval larval shrimp populations increase during floods, and the subsequent year’s shrimp harvest increases. Furthermore, freshwater flows increase the nursery area for Gulf corvina (Cynoscion othonopterus), an important commercial fish that requires estuarine habitats with salinities in the range of 26–38‰ during its natal stages. Although flood flows are now much diminished compared to the pre-dam era, they are still important to the remnant wetland and riparian habitats of the Colorado River delta and to organisms in the intertidal and marine zone. Only a small fraction of the flood flows are evaporated in Laguna Salada.

Since Lake Powell, the most recent large impoundment constructed on the Colorado River, filled in 1980, a series of flood flows have been released from the United States to Mexico during wet years in the watershed. These flows are related to El Niño events, and since 1981 excess water has flowed to Mexico in half the years, totaling 15% (82 m³ sec⁻¹) of the mean flow of the river (IBWC 2006). Many studies report that these flows have improved the ecology of the riparian zone, the intertidal zone, and the marine zone in the Upper Gulf of California (Glenn and others 1996, 2001; Galindo-Bect and others 2000; Pitt and others 2000; Nagler and others 2005).

In a recent paper in Environmental Management, All (2006) proposed that these flood releases for the most part do not actually reach the sea but instead are captured and evaporated in Laguna Salada, a below-sea-level lakebed near the mouth of the river (Fig. 1). He then concluded that future flood flows would not improve the ecology of the marine zone and the fisheries in the Upper Gulf of California because they would fail to reach the sea.

Fig. 1

Locator map of the study area in the delta of the Colorado River. The map shows the main path of the Colorado River to the Gulf of California and the point at which some of the flows enter Laguna Salada, a below-sea-level depression near the mouth of the river

We believe that All (2006) is mistaken. We point out apparent flaws in his analyses and show that his conclusions are contradicted by studies that demonstrate that the river does, indeed, reach the sea during flood years.

Laguna Salada (or Lake Macuata in the older literature) is a shallow depression that fills with water during large flood releases. When inundated, the lake is shallow, ranging from 20 cm to 4 m depth; in 1981, following 3 years of flood releases from the United States to Mexico, it had a mean depth of 1.6 m and a surface area of 40,000 ha [Compean-Jiménez and others, no date (erroneously cited in All (2006) as Jiménez and others 1984)]. All (2006) contended that flood flows in the Colorado River flow preferentially into Laguna Salada, where they evaporate, rather than into the sea, and that “even during flood events, since 1983, little Colorado River water actually reaches the Gulf” (All 2006, p. 11).

Flows in the delta are measured by the International Boundary and Water Commission at the Southerly International Boundary (SIB), below the last diversion point for water on the river (Morelos Dam, at the Northerly International Boundary) (IBWC 2006). Below Morelos Dam, the river is contained within flood control levees until it joins with the Rio Hardy, which then flows to the sea. We agree with All (2006) that little evaporation occurs during the short time (2 or 3 days; Cohen and Henges-Jeck 2001) the water takes to reach the intertidal zone, hence the SIB gauge data are a fair measure of the flood water potentially reaching either the sea or Laguna Salada. We calculated the volume of water that could be captured and evaporated in Laguna Salada in each flood year since 1989 using the SIB flow data and the surface area and evaporation rates for Laguna Salada used by All (All’s Figure 7, reproduced here as Fig. 2). As an example of the calculations we made, in 1993 the annual volume of SIB flow was 4.2 × 10⁹ m³, the mean surface area of Laguna Salada was 220 km² (2.2 × 10⁸ m²), and the evaporation rate was 1.4 m m⁻² yr⁻¹ (from All 2006, and Compean-Jiménez and others, no date). Therefore, the evaporative capacity of Laguna Salada in 1993 was 0.3 × 10⁹ m³, or approximately 7% of the 1993 flows. Similar ratios were obtained for other flood years (Fig. 3). Clearly, Laguna Salada does not have sufficient evaporative surface area to account for the large flows that have occurred since 1981. If all the flows had been diverted into Laguna Salada, it would have been significantly larger and deeper; but in that case it would have spilled back to the sea.

Fig. 2

Reproduction of Figure 7 from All (2006) showing his calculated water surface area for Laguna Salada. Note the x-axis has two 1990s. The original figure caption reads, “Flooded area of Laguna Salada in square kilometers. The number of flooded pixels in Laguna Salada for the given data set are shown. Each pixel is 1 km². This figure was derived from AVHRR data”

Fig. 3

Volume of annual flow in the Colorado River at the Southerly International Boundary compared to the evaporative capacity of Laguna Salada, calculated from surface area and evaporation rate data in All (2006)

All (2006) cites Compean-Jiménez and others (no date) in support of his evaporation estimates but does not note the previous paragraph in the cited reference, which offers an estimate of Laguna Salada’s volume and extent (presumably in 1981, the apparent date of publication, since the much larger floods of 1983 are not mentioned): 730 × 10⁹ m³, at 40,700 ha. Compean-Jiménez and others state, “With a discharge of 6 m³ sec⁻¹, it is possible to maintain an inundated area of 7604 hectares during the month of June, with a maximum of 31,984 hectares for December.” In December 1998, mean daily discharge at SIB was 244 m³ sec⁻¹, suggesting that either the Laguna Salada should have greatly exceeded 32,000 ha in extent (but by Fig. 2 it was only 15,000 ha) or that most Colorado River water bypassed the Laguna Salada and flowed instead to the Gulf of California.

The x-axis in All’s (2006) Figure 7 contains a mistake (it has two “1990s”), creating confusion about the area estimates of Laguna Salada for particular years. We assumed that data for the second 1990 were actually for 1991 and that subsequent data were also off by 1 year. This better aligns the years of peak surface area with years with river flow. In either case, his Figure 7 shows a substantial water area in Laguna Salada in 1990 through 1992 (or 1991–1993 by our count of years). However, water ceased flowing in 1988 and did not resume until 1993; hence, Laguna Salada should have been dry in 1992. In fact, inspection of a June 30, 1992, Thematic Mapper (TM) satellite image shows that Laguna Salada was dry on that date (Fig. 4), even though All’s Figure 7 showed it had either 37,500 or 30,000 ha of water surface, depending on how the years are counted.

Fig. 4

Thematic Mapper image of Laguna Salada from June 30, 1992, at a time when All (2006) estimated it contained 30,000–37,500 ha of water surface. The bottom of Laguna Salada in this image is dry and covered with a white salt crust; a fringe of vegetation is seen along the perimeter of the lakebed from previous flows. The near infrared band (NIR) is shown in red; vegetation reflects NIR and appears red, whereas water absorbs NIR and appears blue to black (e.g., the Cerro Prieto Ponds). The blue arrow shows the pathway of water from the Colorado River to Laguna Salada during flood events

All apparently did not calibrate or validate his image classification method using ground reference data, and the analysis may have been flawed. In particular, he used a histogram method to identify the NDVI values for the “soil peak” and the “water peak,” then classified each pixel in his subset image as either soil or water based on the midpoint between the peaks. However, salt crusts, which are present on Laguna Salada when it is dry, generally have lower NDVI values than desert soil (Bradley and Mustard 2005), though not as low as water. On TM images, salt crusts have NDVI values of −0.1 to −0.2, whereas soil is approximately 0.0 (based on digital number rather than reflectance band values) (P. Nagler, personal observation). All (2006, p. 121) found the variability in the spatial extent of Laguna Salada during his study period “somewhat surprising,” attributing it to the rapid evaporation of sporadic shallow inflows. However, the saw-toothed pattern of surface area in Figure 2 cannot be explained using realistic evaporation and flow rate estimates. For example, in December 1991, the surface area was 450 km², but it was depicted as dropping to 200 km² 2 months latter. With an average depth of 1.6 m at 400 km² (Compean-Jiménez and others, no date), the required evaporation rate would be on the order of 13 mm day⁻¹, but Arizona Meteorological Network (AZMET) potential evaporation rates for Yuma Valley during that period were on the order of 3 mm day⁻¹ (AZMET 2006). Then it is depicted as refilling almost as rapidly, even though there were no flows in the river during that period (or in the subsequent year). We believe that his pixel classification scheme provides an unreliable estimate of true water surface area.

All (2006) assumes that water flows first into Laguna Salada and that flows to the sea resume only after Laguna Salada fills. In fact, only large flows produce spillage into Laguna Salada. All shows a TM satellite image from 1985, which he interprets as showing that the main pathway of flow during that flood year was into Laguna Salada. Close inspection of a similar image available to us (June 14, 1986) from the University of Arizona’s Arizona Regional Image Archive (ARIA) tells a different story (Fig. 5). In the zoomed out view of the floodplain, as presented by All, the main path of water movement does appear to be into Laguna Salada. However, a closeup of the area of floodplain where flows diverge shows that three branches of the Colorado River join together to flow in a single channel to the sea, whereas the broad channel that leads to the Laguna Salada does not appear to carry water, although the soil appears wet. The flow at the SIB for the 5 days previous to 1986 image acquisition was 410 m³ sec⁻¹, among the highest recorded since flows resumed in 1981. Flows at the SIB in 2000 were in the range of 100–200 m³ sec⁻¹, and inspection of 2000 and 2001 TM images from the ARIA collection (not shown) showed that these failed to enter Laguna Salada; hence, there is apparently a threshold flow needed for water to spill from the main channel into Laguna Salada. River water will flow to the Gulf before any spills into Laguna Salada.

Fig. 5

The Colorado River delta at the intake region for water into Laguna Salada during a major flow event in June 1986. SIB flows refer to the gauging station at the Southerly International Boundary between the United States and Mexico. The top panel appears to show that most of the water is diverted into Laguna Salada; however, the bottom panel shows that the connecting channel is nearly dry (A), whereas the main river channel (C) conveys water to the sea. A man-made channel (B) to divert water from the river to Laguna Salada is also visible; this channel reportedly had a capacity of approximately 6 m³ sec⁻¹ but it has since filled in with sand (authors’ personal observations, 1997)

Furthermore, there is evidence that indicates that Gulf water may flow into Laguna Salada during high spring tides (Payne and others 1992; Compean-Jiménez and others, no date). Such flows counter All’s (2006) assumption that river water is the only source of water in Laguna Salada and will also affect his calculation of evaporation rates. Other factual errors also detract from All’s argument. For example, All states that Morelos Dam has a maximum diversion capacity of 100 m³ sec⁻¹, but the actual capacity is 226 m³ sec⁻¹ (Cohen and Henges-Jeck 2001).

Several studies have documented that flood flows reach the Gulf of California. We measured salinity in the intertidal portion of the river on an incoming tide on January 13, 1998, during a flow event of 213 m³ sec⁻¹ (Fig. 6). In the absence of river flow, the mouth of the river and Upper Gulf behaves as an inverse estuary (Lavín and others 1998), with salinity at a maximum near the head of the river and decreasing toward the Gulf. Winter salinity in the head of the river’s estuary (approximately 10 km upstream from the mouth) was 38–40‰ during the nonflood years of 1989 and 1990 (Hernandez-Ayon and others 1993), exceeding the 37‰ observed by Lavín and others (1998) and Alvarez-Borrego and others (1975) seaward of Montague Island. However, during the January 1998 flow event, significant dilution of seawater was observed at the midriff of Montague Island seaward of the river mouth, and essentially fresh water was observed 10 km upstream from the mouth of the river.

Fig. 6

Salinity (‰) in the estuary of the Gulf of California during a flow event of 213 m³ sec⁻¹ on January 13, 1998. Readings were taken on an incoming tide (water was entering the river from the sea). Salinity was measured with a hand-held refractometer and locations were determined from a Garmen Global Positioning System unit

In 1993, a controlled release of river water during April maximum daily flow at the SIB reached 450 m³ sec⁻¹ and was observed to reduce surface salinity to approximately 23‰ southwest of Isla Montague and reduced surface salinities along the western coastline as far as 70 km distant from the river’s mouth (Lavín and Sánchez 1999).

Geochemical evidence from isotopic studies of live-collected shells also documents that Colorado River flow reaches the Upper Gulf of California. Colorado River water is isotopically distinct from the seawater of the Gulf of California. Dettman et al. (2004) document that the typical δ¹⁸O composition of Colorado River water is −12‰, whereas open Gulf water is typically 0‰ (compared to Vienna Standard Mean Ocean Water, the reference seawater). When mollusks and fish precipitate CaCO₃, they do so in equilibrium with the ambient water. Thus, the deviation of the δ¹⁸O in skeletal carbonates below the range produced by high temperatures is the result of the mixing of ¹⁸O-depleted river water with relatively enriched seawater (Dettman and others 2004). Controlled releases of river water in early 1993 and early 1998 are reflected in anomalously low δ¹⁸O values in live-collected shells of the bivalve mollusks Chione cortezi collected from Isla Montague and Isla Sacatosa (approximately 20 km south of Isla Montague) (Dettman and others 2004) and Mulinia coloradoensis from Isla Montague (Rodriguez and others 2001). In addition, anomalously low δ¹⁸O values in the otoliths (aragonitic earbones) of the Gulf Corvina (Cynoscion othonopterus) record controlled releases of river water in 1993 and 1999 (Rowell and others 2005). Maximum daily flows at SIB in 1999 did not exceed 278 m³ sec⁻¹. Colorado River flows are recorded in the skeletons of organisms living in the Upper Gulf of California.

Other effects of Colorado River flow on organisms in the Upper Gulf are well documented. Aragón-Noriega and Calderón-Aguilera (2000) show a statistically significant positive correlation between river flow and the relative abundance of postlarvae of the shrimp Litopenaeus stylirostris in the years 1993–1997. All (2006) questioned Galindo-Bect and others’ (2000) evidence for a positive relationship between Colorado River flow and the shrimp catch in the Upper Gulf, and advanced an alternative explanation based on varying fishing effort. However, Galindo-Bect and others explicitly took into account varying fishing effort in their analysis.

All (2006) cites Rodriguez and others’ (2001) study of the bivalve mollusk Mulinia coloradoensis as evidence that freshwater flows do not reach the sea. This clam is not commercially harvested, so if river flow is beneficial, then this brackish clam’s population should increase during flood years. However, Rodriguez et al. did not conduct any surveys to determine if this species had increased following flood periods.

In summary, (1) the volume and surface area of Laguna Salada are not sufficient to capture and evaporate more than approximately 5 × 10⁸ m³ yr⁻¹ of water, a volume that has been exceeded during most of the releases; (2) low-volume flows do not preferentially enter Laguna Salada but remain in the main channel; (3) direct observations of low salinity in the mouth of the river contradict All’s (2006) conclusion that little to no river water reaches the Gulf of California during controlled releases; (4) the isotopic signal of Colorado River water from flood years is recorded in the carbonate hard parts of species of mollusks and fish; and (5) previous studies have documented that shrimp populations and shrimp catches (even after accounting for variation in fishing effort) increase following controlled releases of river water.

All (2006) correctly draws attention to the problem of overfishing in the northern Gulf of California and the need for greater management flexibility in the timing of flood releases to Mexico. However, we disagree with his conclusion that most of the flood releases are captured and evaporated in Laguna Salada. We greatly appreciate the opportunity to summarize the wealth of evidence that Colorado River water can and does reach the Gulf of California, and that Colorado River water can and does have a beneficial effect on the species living there.