Climatic Change and Water Availability in the Rio Grande and Pecos River Basins

Climatic change and water availability in the Rio Grande and Pecos River basins

Item Type Thesis-Reproduction (electronic); text

Authors Quinlan, Peter Thomas.

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 01/10/2021 18:08:04

Link to Item http://hdl.handle.net/10150/191762 CLIMATIC CHANGE AND WATER AVAILABILITY IN THE RIO GRANDE AND PECOS RIVER BASINS

by Peter Thomas Quinlan

A Thesis Submitted to the Faculty of the DEPARTMENT OF HYDROLOGY

In Partial Fulfillment of the Requirements For the Degree of MASTER OF SCIENCE

In the Graduate College THE UNIVERSITY OF ARIZONA

1982 STATEMENT BY AUTHOR

This thesis has been submitted in partial fulfillment of requirements for an advanced degree at The University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this thesis are allowable without special permission, provided that accurate acknowledgment of source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his judg- ment the proposed use of the material is in the interests of scholar- ship. In all other instances, however, permission must be obtained from the author.

SIGNED:

APPROVAL BY THESIS DIRECTOR

This thesis has been approved on the date shown below:

Z-rz 79T?-- Date ACKNOWLEDGMENTS

I am indebted to Charles W. Stockton, David Meko,

Irmgard Flaschka, and W. R. Boggess for valuable suggestions.

Mr. Boggess also provided important editorial advice.

Most importantly this thesis could not have been written without the support of Deborah Quinlan. TABLE OF CONTENTS

Page

LIST OF ILLUSTRATIONS

LIST OF TABLES vi

ABSTRACT vii

1. INTRODUCTION 1

2. PHYSICAL DESCRIPTION OF THE STUDY AREA 6

3. METHOD 9

Selection of Basins 9 Selection of Climatic Data 14 Selection of an Equation for the Model 19 Snowmelt 23

4. RESULTS AND DISCUSSION 27

Early Results 27 Description of Basin Specific Models 30 Comparison of Model Predictions with Streamf low . . . 32 Scenario Results 38

5. CONCLUSION 45

LIST OF REFERENCES

iv LIST OF ILLUSTRATIONS

Figure Page

1. The Rio Grande Basin 7

2. Study Areas in the Upper Rio Grande Basin 13

3. Precipitation-Runoff Pattern of the Rio Grande at Del Norte 15

4. Precipitation-Runoff Pattern of the Conejos River at Magote, Colorado 16

5. Annual Discharge for Rio Chama Below El Vado Dam, New Mexico (lower, smooth line) With Model Predictions (lower line with triangles) and Precipitation Input to Model (upper line) 33

6. Annual Discharge for the Conejos River at Magote, Colorado (smooth lower line) with Model Predictions (lower line with triangles) and Precipitation Input to Model (uppermost line) 34

7. Annual Discharge for the Rio Grande at Del Norte, Colorado (smooth lower line) with Model Predictions (lower line with triangles) and Precipitation Input to Model (uppermost line) 35

8. Cumulative Departures from the Present Mean for the Rio Chama 41

9. Cumulative Departures from Present Mean for the Conejos River 42

10. Cumulative Departures from the Present Mean for the Rio Grande LIST OF TABLES

Table Page

1. Observed Mean Annual Flows and Estimated Natural Flows in Acre-Feet for Various Locations in the Rio Grande Basin as Computed by the Water Resources Council . . 10

2. Parameters of the Basin-Specific Models for the Rio Grande, Rio Chama, and Conejos River 31

3. Comparative Statistics for Observed Streamflow Records from Study Watersheds and Models 36

4 • Mean Annual Discharge of Scenario Investigations and the Percentage Change from Original Model 39

vi ABSTRACT

Climatologists have speculated that increasing concentrations of atmospheric carbon dioxide deriving from fossil fuel combustion will result in a warmer, drier climate for many parts of the world.

One such area which is already facing serious water shortages is the upper Rio Grande Basin. A climatic water balance equation was adopted to model three representative drainages within the basin in order to

investigate the effects of this climatic change on streamflow. Results show that a 2 ° C increase in temperature and a 10% decrease in precipitation would result in a 30% decline in streamflow. Since demand already exceeds or approaches supply, such a diminution in water yield would have serious regional and national ramifications.

vii CHAPTER 1

INTRODUCTION

Development in the western United States to the year

2000 and beyond is tied to the availability of surface water for off-

stream uses including: agriculture, industry, municipalities, and power generation. In addition, there is increasing demand for water for instream use. As described by Anderson (1982) this category

includes, but is not limited to, recreation, the preservation of scenic reaches in accordance with society's aesthetic values, and the main- tainence of wildlife habitats. As the population of the West, and especially the Southwest, grows due to the influx of people from the

North and East, water demand in many of the sectors mentioned will increase. While it is doubtful that the streams and rivers of the

West have sufficient unallocated reserves to meet increasing demand, it is likely that the Rio Grande does not. According to the United

States Water Resources Council (WRC), the surface water of the Rio

Grande Basin appears to be totally allocated (WRC 1978). As noted by the Water Resources Council (WRC 1978), any increase in use by one segment must be accompanied by an equal decrease in another, or others.

In view of this total allocation, any dimunition in supply will have adverse consequences.

Ground water is not available as an alternative supply to make up any discrepancy between demand and surface water supply in

1 the Rio Grande Basin because it is already extensively mined. Since the 1930's ground water pumpage has lead to dramatic water table declines. Clearly, increased pumpage, if possible, is not a long term solution. Furthermore, it seems likely that in view of WRC (1978) projections of decreased pumpage by the turn of the century ground water will meet a smaller portion of demand in the future than it does at present. This decline will place even greater stress on surface water supplies.

As a result of the Rio Grande demand-supply ratio approaching, or perhaps exceeding, 1.0 there is increasing need for careful water resource planning. Any planning should assess probable demand in each use sector and probable available supply. Climate is usually treated as a constant in water resource planning, even long term planning (WRC 1977), but climatologists agree that climate is ever varying.

Accordingly, the purpose of this study was to investigate the effect of one element of climatic variation on water availablity. While not dealt with here, it seems clear that demand too is affected by climatic change; irrigation requirements increase during periods of drought.

Climate appears to vary in a number of natural frequencies.

Evidence of short term variation, 1-50 years, can be found in extremely long historical weather and streamflow records. Longer term,

10-1000 years, climatic variation is evidenced in tree-ring width records. Very long term, low frequency, variation is seen in geologic records, as in the cases of ice ages. ln addition to this natural variation, there is now speculation about the possibility of climatic 3 change arising from man-induced increases in atmospheric carbon dioxide concentrations. The focus of this work concerns the effects of this last, anthropogenic, climatic change on water availability in the Rio Grande Basin.

Since the nineteenth century, there has been a marked increase in atmoshperic carbon dioxide concentrations resulting from man's combustion of fossil fuels. There is gathering agreement that this carbon dioxide build up will lead to global warming (National

Defence University 1978). A modelling study (J. Hansen et al. 1981) predicts a dramatic, +2.5 ° C, global warming over the course of the next centruy. This paper further points out that the cooling trend that followed the 1880-1940 warming period stopped in the sixties, and that global temperature is now nearly at the 1940 level. Kellogg

(1982) cites other studies that posit a global warming of +3 ° C with a doubling of carbon dioxide concentrations. Revelle (1982) cautions that temperature increases over the last 100 years have not exceeded the noise interval of normal variation, and that verification of a direct causal link between carbon dioxide increases and global warming will only come with continued warming in the 1990 1 s. These studies stated that temperature changes would be most pronounced in the higher lattitudes of both hemispheres. Hansen (1981) and Flohn (1982) have both speculated that local climatic change may occur due to atmospheric circulation patterns being altered by global warming. Both have fore- seen increased and expanded aridity in the western United States. In view of these findings, a temperature change of ±2 ° C was deemed reasonable and, Perhaps in the long term, conservative for useininvestigating 4 the effects of this anthropogenic climatic change on water availability in the Rio Grande Basin. Three scenarios: warmer-drier

(+2 ° C and -10% precipitation); warmer-wetter (+2 ° C and +10% precipitation); and cooler-wetter (-2 ° C and +10% precipitation are examined in this study.

To accomplish the estimation of the impact of climatic change, a climatological water budget equation was adopted. This choice allowed the modification of climatic input to simulate the desired scenarios. Results were verified by comparison with records of observed streamflow. The sutdy goal was to develop a fifty-to- seventy-year time series of each scenario. Opting for a time series approach rather than using long term means permitted scrutiny of the extreme effects accuring from persistent drought.

The Rio Grande illustrates a number of problems encountered

in applying climatological water balance equation to basin analysis.

Three main problems concerned spatial resolution, snowmelt, and the availability of natural, non-regulated, flow for model verification.

Difficulties with spatial resolution centered on the fact that climatic

input to the water balance equation were from a specific point, or points, in a basin thus yielding a point value for runoff. This point measure had then to be extrapolated over the entire drainage area for comparison with observed streamflow. In essence, this

amounted to an attempt to generate a point value that represented

average conditions over the whole basin, but that might, in fact,

not reflect, in total, the conditions at any specific location. 5

Snowmelt, or rather snow accumulation, creates problems both in

simulating the temporal distribution of runoff and in energy budget

calculations of potential evapotranspiration. Agriculture, incorpor-

ating stream diversions for irrigating in the basin, was firmly

established before the Spanish arrived in 1540. Unrecorded diversions

distort gaged streamflow records reducing or totally negating the use-

fulness of fifty-plus year records. If detention and diversion records

exist for the drainage area above a gage, natural flow at that spot

can be reconstructed. All too often in the Rio Grande Basin, regulated

streams have no records of their regulation. The resolution of these

problems will be addressed later in the Methods Section.

The above constraints not withstanding, the Rio Grande

Basin appeared particularly appropriate for investigating the effects of anthropogenic climatic change on water availability on three counts.

First, because the Rio Grande's surface water supply is already

totally allocated, any climatic change resulting in a dimunition of

flow will have dramatic social and economic impact. Second, two

studies (Flohn 1981; Kellogg 1982) predict just such a change to

greater aridity in the area of the upper Rio Grande. Finally, the

problems encountered in investigating this basin, while perhaps more

severe than elsewhere, are likely to be present in the runoff producing

areas of other river basins in the western United States. CHAPTER 2

PHYSICAL DESCRIPTION OF THE STUDY AREA

The Rio Grande Basin incorporates parts of Colorado,

New Mexico, Texas, and Mexico (Figure 1), and comprises 230,000

square miles. Of that total, 93,000 square miles are in Mexico. Only

89,000 of the remaining 137,000 square miles on the American side of

the border contribute runoff. Closed basins encompass 44,000 square miles. The Rio Grande has the following major tributaries: the

Pecos River (draining 35,200 square miles); the Rio Conchos (25,400);

the Rio Salado (23,300); the Rio Puerco (7,350); and the Rio Chama

(3,144). Rising in the San Juan Range of the southern Rocky Mountains,

the Rio Grande flows 1,244 miles to the Gulf of Mexico near Browns-

ville, Texas. Elevations in the mountains above the headwaters

approach 14,000 feet. The annual natural flow at the mouth of the

river was roughly estimated to be 5.4 million acre-feet (WRC 1978).

The majority of this water derives from melting snow in the high

mountain areas (WRG 1978; New Mexico State Office of Engineering

1967). Thunderstorms and hurricanes contribute to flow in the lower

reaches of the river.

The geology of the Rio Grande Basin from its head in the

a rift-zone San Juan Mountains down to E . is characterized by

valley with mountains on either side. The rift zone, a contiguous

series of deep grabens, is filled with alluvium and lava flows

6 7

• ••••n•

MEXICO

Figure 1. The Rio Grande Basin. 8

(WRC 1978). Between El Paso and the Big Bend, the river enters a basin and range province. Thereafter, it flows over flatlying sedi- mentary formations. The valley fill is as much as 9,000 feet thick and has proved to be a productive aquifer. Recharge comes from streamflow (New Mexico State Office of Engineering 1967). The Pecos

River flows over beds of limestone, sandstone, shale, gypsum, and salt. Of these the limestone (when solutioned) and the gypsum are productive aquifers.

The basin has both spatial and temporal variation in climate, ranging from arctic in the mountains to arid in the middle reach basins to semi-tropic in the coastal plains (West and Broadhurst

1975). Following this pattern, mean annual precipitation varies from thirty inches to eight inches to thirty inches. Winter precipitation characteristically derives from frontal stroms. Hurricanes followed by intense rainfall occur in the coastal plains. CHAPTER 3

METHOD

There were three stages in the design of this study.

The first was the selection of gages (and, thereby, drainage areas) whose streamflow records would be employed in verification. The

second stage was the selection of meteorological records from stations within the chosen drainages. Following these two, the third stage

involved the selection of a climatological water budget equation that would be appropriate for the conditions in the selected basins and

for the available meteorological data.

Selection of Basins

As previously stated, the Rio Grande has been character-

ized by extensive, long-standing, and only partially recorded regu-

lation and diversion. A look at the vast discrepancies between

estimated natural flows (WRC 1978) and observed mean annual discharge

illustrates the magnitude of this problem (Table 1). At the Colorado-

New Mexico line, the natural flow of the Rio Grande mainstream is

over three times the observed. At the New Mexico-Texas border, the

estimated natural flow is five and a half times the observed discharge.

The Pecos River is even more extreme; the estimate of natural flow

is eight times greater than what is actually observed flowing into

Texas from New Mexico. Clearly it would be an overwhelming, if not

9 10

Table 1. Observed Mean Annual Flows and Estimated Natural Flows in Acre-Feet for Various Locations in the Rio Grande Basin as Computed by the Water Resources Council.

LOCATION ACTUAL MEAN FLOW NATURAL FLOW

Rio Grande

Colorado-New Mexico 299,000 950,000 Border

New Mexico-Texas 385,000 2,134,000 Border

Brownsville, Texas 1,380,000 5,400,000

Pecos River

New Mexico-Texas 137,000 1,090,513 Border 11 impossible, task to reconstruct the Rio Grande's natural flow at its mouth - or even at El Paso. This was particularly true below El Paso in view of the lack of available records for the Mexican tributaries, a situation which narrowed the search to basins above the confluence of the Rio Grande and the Rio Conchos. Therefore, the first priority became the selection of constituent basins that would be representative of the whole and whose gages either measured natural flow or could be adjusted to approximate natural flow.

Seven criteria determined whether a basin was chosen.

The first concerned record length; it was hoped that all basins would have fifty or more years of records. Two criteria defined representative, while the other four delineated acceptable levels of human flow-altering activities. Representative basins were defined as those that (1) had large drainage areas and (2) were major contribu- tors to the estimated natural flow at the New Mexico-Texas state line.

Acceptable gages were those that had no reservoirs above them, or those that had reservoirs with records of contents, or, failing that, those for which total reservoir capacity did not exceed the annual gage error. Annual gage error was considered to be the product of the U.S.G.S. assessment of gage error with the mean annual

Flow at the gage. For example, a gage designated good by the U.S.G.S. has an accuracy of ±10/0 . If the mean annual discharge at a good gage is 250,000 acre-feet, the total acceptable storage capacity for reservoirs with unreported contents above the gage would be 25,000 acre-feet. 12

The second record-related criterion was that there be no diversions for irrigation above the gage, or, failing that, that the total reported irrigated acreage multiplied by a locally reasonable column of water (three to four feet per year in southern Colorado and northern New Mexico) be less than the annual gage error. Similarly the third criterion was that there be no trans-basin diversions above the gage, or that there be records of them, or that they fall within the gage error. The last criterion was that there be records of ground water levels in the basin above the selected gage. In the absence of records, it was hoped that there would be at least indirect evidence of negligible water table change. The State of New Mexico has monitored ground water levels extensively since the 1940's; if a basin fell in an area that had not been monitored, I assumed that this indicated an area free from complaints, an area with stable water levels.

Four basins were selected for study and include the drainage areas above the following gages: the Rio Grande at Del Norte,

Colorado; the Conejos River at Magote, Colorado; the Rio Chama below

El Vado Dam, New Mexico; and the Pecos River near Pecos, New Mexico

(Figure 2). The Rio Grande at Del Norte has a drainage of 1,320 square miles. The elevation at the gage is 7,980 feet, and the mean annual discharge is approximately 660,000 acre-feet. Unlike the other watersheds, the Rio Grande gage did not meet the criterion for regulated storage capacity. Total storage capacity above the gage was 1 3

0 I 07°W SAGUACHE 106°

38°N

Figure 2. Study Areas in the Upper Rio Grande Basin, 14 approximately 20% of mean annual flow. The basin was included despite this failing because of its status as a major contributor to the ultimate flow.

The Conejos River at Magote drains an area of 282 square miles, is at an elevation of 8,272 feet and has a mean flow of 246,000 acre-feet/year. The Rio Chama below El Vado has a contributing drainage of 777 square miles, an elevation of 6,692 feet, and a mean discharge of 277,300 acre-feet. The Pecos River above Pecos has a drainage of 189 square miles and a mean annual discharge of 72,180 feet.

This gage is at 7,503 feet above sea level.

Taken together, the first three contribute 1.1 million acre-feet (MAF) per year to the Rio Grande. Thus roughly 50% of the estimated natural flow at El Paso derives from approximately 7% of the basin area. As can be seen from the monthly mean precipitation- runoff patterns (Figures 3 and 4), the majority of the precipitation occurs from July through October, whereas the majority of the runoff comes in May-June. Obviously snow accumulation and melt are important factors in the hydrology of these basins. All streamflow and gage data come from U.S.G.S. Surface Water Supply Papers (U.S.G.S 1960, 1964, 1970, 1974, 1971, 1972, 1973, 1974, 1975, 1976, 1977, 1978,

1979 and 1980).

Selection of Climatic Data

Two criteria governed the choice of climatic data. The first was length of record; desireable records should exceed fifty years. The second criterion concerned the need for representative climatic data. 1 5

RIO GRANDE AT DEL NORTE

precipitation

streamf low

at Figure 3. Precipitati on-Runoff Pattern of the Rio Grande and Del Norte, Colorado. Monthly Average Precipitation Annual Mean Streamflow are Presented as a Percentage of Precipitati on Data are from Hermit, Colorado. 16

CONEJOS RIVER AT MAGOTE

precipitation

streamflow

Figure L. Precipitation-Runoff Pattern of the Conejos River at Magote, Colorado. Monthly Average Precipitation and Streamflow are Presented as a Percentage of Annual Mean. Precipitation Data are from Chama, New Mexico. 17

Individual station records contain localized noise not

representative of general weather conditions over the surrounding

areas, even in flat plains. Thus extraploating from a point to des-

cribe the weather of an area introduces the posSibility, in fact, the

probability of error. This problem is further compounded when, as in

the case of the study areas, great topographical variety results in disparate weather conditions over the basin. One solution to the problem is to calculate an elevation-area weighted average of the

stations in a basin. Unfortunately, I was able to find only two

stations of sufficient length within the four basins. Clearly an elevation-area average was not possible.

The only alternative to using individual station records was to average data from stations over a larger area that included

the selected basins. The National Climatic Center has compiled monthly divisional averages of temperature and precipitation for the

period 1931-1980 based on the climatic divisions delineated in its monthly and annual climatological data publications. The problem with using these divisional data was that the divisions included much larger areas than the study basins, and that much of the additional area was low lying. Also there tend to be more towns and cities in lowland regions than in the mountains (hence more weather stations) and since the divisional average is a straight

arithmetic average, these divisional data have a lowland bias.

Results using divisional data were much less satisfactory than those

resulting from the use of individual stations. Accordingly the use 18 of divisional averages was, with the exception of the Pecos River where no individual station was available, soon dropped. It is reasonable that individual stations high in the mountainous runoff- producing areas, even with localized noise, would do a better job of predicting streamflow than a broader average with a lowland bias.

To ensure that individual station noise was not compounded by trends induced by urbanization or by station moves, double mass analysis was employed to test for homogeneity among the selected stations. In this test cumulative precipitation (or average temperature) values from two stations are plotted against each other. Any deflection or change in slope of the resulting line indicates heter- ogeneity, that a change in conditions at one station is not matched at the other. Five stations in the upper Rio Grande area were examined. They were: Chama, New Mexico; Silverton, Colorado; Hermit

7 ESE, Colorado; Saguache, Colorado; and Santa Fe, New Mexico. The stations selected for use - Chama, Hermit, and Sagauche - all did satisfactorily in this analysis. Ultimately the Chama temperature and precipitation records were used in modeling the Rio Chama, the

Chama temperature record and an average of Chama and Hermit precipitation records were used for the Conejos River, and an average of

Chama, Hermit, and Saguache precipitation was used along with the

Chama temperature record on the Rio Grande at Del Norte. These stations have only temperature and precipitation records. They have no published records of relative humidity, percent of clear sky, or atmospheric pressure. This limitation was a consideration in choosing a water balance equation. 1 9

Selection of an Equation for the Model

The atmospheric phase of the hydrologic cycle is a

process whose great complexity is matched only by the paucity of

measurements of the parameters involved. Accordingly climatic water

balance equations vary in complexity from simple to highly complicated

depending on how many parameters are taken into consideration. The

choice of an equation for this study was dictated by the twin con-

straints of large areal scale and long temporal scale. Given the

large and sparsely populated areas to be modeled, rain gage density was necessarily low. It was surprising, however, how low the density

became as a result of the long record constraint. This, along with

the lack of records of relative humidity, cloud cover, and potential

evaporation for stations within the study basin questioned the accuracy

of results from a complex, realistic model that has been fed approxi-

mate, extrapolated means as input for missing parameters. Accordingly,

a simple model was chosen. The equation adopted was a version of

Budyko's water balance equation (Budyko 1956) which had been simplified

by Sellers (1965). Other, more recent, snowmelt-runoff models based

on LANDSAT measurements of snow cover (Jones et al. 1981) were passed

over because of their short term forecasting focus and the lack of

long term LANDSAT records. The general equation used was essentially

a bookkeeping water budget and in this study was used to predict

runoff according to the general equation: 20

P - E - (W2-W1 ) = S, where; (1)

P = monthly precipitation

E = evapotranspiration

W2 = soil moisture content of present month W1 = soil moisture content of preceding month S = moisture surplus (including both surface

runoff and downward percolation).

In equation (1), the only value known initially is P. The others must be defined in terms of known values and few enough unknowns that they may be solved for empirically. This is accomplished by means of the following relationships. The relation between E and potential evapotranspiration (E ) is given by;

E = E when 0.5(W1+W2) -0'75Wmax , where

W = maximum soil moisture capacity max and

0.5(W +W )E 1 2 p (2) E = when 0.5(W1+W2) < 0. 75Wmax. max

for Sellers' W (critical soil The specific value chosen in this study c moisture) was 0.75 of the maximum. To make the model more sensitive, potential evapotranspiration was made a function of temperature, T, according to the relationship:

Ep E =T =Ta , where; 21

E = the computed value of evapotranspiration

for the specific month in question

(say June 1950)

T = the temperature of the specific month

in question

= the long term mean evpotranspiration for

the month (June)

"T = the long term mean temperature for the

month.

The moisture surplus, S, is defined by Sellers (1957) as; 2 0.8P [0.5(w +w )] 1 2 S - (3) (aT +'P)Wmax

It should be noted that this expression allows for runoff when the

soil moisture column is not saturated as is the case nature. Sub-

stituting equations (2) and (3) into (1), the full expression becomes; 2 1 +W )aT 0.8P [0.5(W + W )] 0.5(W 2 1 2 P - W +:W - 0.75W 2' 1 (aT + P)W max max where, by assuming an initial value for W1' everything is known except W2 . Solving for W2 yields;

1 W2 = 0.5aT + 1.0 + 0.8P2115.5) 0.75Wmax

a value for moisture surplus is found using' Once W2 is quantified, equation (3). 22

In order to solve for W as outlined on the previous page, 2 it was necessary to first compute values for mean monthly potential evapotranspiration. Two options presented themselves for this com- putation: use of the Penman energy budget equation or use of class

A pan evaporation data. The Penman equation requires first that an energy budget be calculated so that net radiation can be used as input and then that wind speed, vapor pressure, and relative humidity be known in order that potential evapotranspiration be calculated.

The general form of the energy budget is;

Q Q NR = QSW (1—cx) + LW where,

= net radiation QNR

Q = short wave (solar) radiation SW

"r= the albedo of the surface receiving the radiation

(infared) radiation. Lw = net long wave

As there were no records of these parameters for Chama and Hermit,

long term monthly means interpolated from a climatic atlas were used

(U.S. Dept. of Commerce 1968). Because of the aforementioned concerns with applying a precise model to approximate data and because of snowmelt related problems in the energy budget which will be dealt with

later, the model was run primarily with 0.7 of the long term mean monthly values from a class A pan at El Vado Dam, New Mexico (N.M.

State Engineering Office 1967). Being the results of actual obser- vations, the pan evaporation values engendered greater confidence. 2 3

Some runs were made using Penman-computed evapotranspiration values for informal comparision and, surprisingly, differed little from those using pan evaporation.

Snowmelt

Stating that snow accumulation and snowmelt runoff are very complicated processes, Mather (1978) suggests that modifications made to water budget to account for snow be done on a case by case basis. The prominence of snow accumulation and snowmelt in the hydrology of the study areas lead to concerns for the applicability of the Budyko equation and the validity of the energy budget values for evapotranspiration that result from Penman calculations. Sellers

(1965) states that the Water balance equation employed here is valid when snow accumulation is negligible. He further reports that Thornth- waite and Mather (1957) judge snow effects to be negligible when monthly temperatures are above 30.2 ° F (Sellers 1965). Just such conditions exist in all the study areas for up to four months a year.

To cope with this problem, the model was modified in the manner suggested by Thornthwaite and Mather (1957) for their model when snow accumulation was a decided factor (Thornthwaite and Mather 1957).

They introduced a new bookkeeping aspect to the model by keeping all runoff in storage until the monthly temperature exceeded 30.2 ° F. I experimented with releasing varying portions of the accumulated runoff in stages of up to four months, but ultimately followed the lead of Stone in releasing all of it during the first 30.2°F month

(Stone, Albrecht and Yoshioka 1971). The effects of this modification 24 were to increase evapotranspiration by keeping soil moisture at higher

levels thus lowering mean moisture surplus and to improve slightly the

temporal distribution of the runoff. However, since the aim of the

investigation was to generate annual discharge values, improved monthly

estimates were important only in so much as they improved the predic-

tion of the annual volume of flow.

The snow related problems with the energy budget are

twofold. First, snow cover can increase the albedo dramatically from

the desert upland norm of 0.1 to 0.2. Without snow cover measurements,

the proper modification of albedo is difficult to assess. Second, having somehow arrived upon a snow cover influenced value for net

radiation, an additional term must be added to the Penman equation to account for the energy consumed in melting snow and thereby not available for use in evaporation, as in the following equation; x (QNR- G M) + LE LE = "76- where:

1 L = latent heat of vaporization

E = evapotranspiration

G = change in soil heat

M = heat used in snowmelt

Ex = vapor pressure deficit, that is, a term

proportional to the difference between

saturation vapor pressure and actual vapor

pressure

A = the slope of water vapor pressure with

temperature 2 5

Y = psychometric constant, a function of

air pressure.

The quantification of the term M would be problematical even with snow

course measurements to aid in the estimation of the mass of snow to be

melted. Without them, one must resort to assuming that M is equal to

a general percentage of net radiation that varies with latitude. The

use of pan evaporation does not circumvent the problems with albedo

and energy loss to melting snow. The albedo for an old snow cover can

be from four to twenty fold greater than a water surface (Geiger 1965).

Pan evaporation for the month of May will clearly be much higher than

actual evapotranspiration from the higher parts of the watershed still

covered by snow where incoming radiation available for evaporation is

greatly reduced by reflection and snowmelt. In sum, however, I felt

that there were fewer problems associated with using pan evaporation

and proceded accordingly.

Having chosen a model and study areas, the next state of

the investigation was to make an initial run for each area and cor-

relate the output with observed streamflow and to compare the magnitude

of mean observed and predicted flows. One might expect that the soil

moisture surplus of the model ought to be lagged before correlation

with streamflow since it represents both runoff traveling on the sur-

face and water percolating through the ground. Mather (1979) reports

that the period of lag varies with basin, but that van Hylckama (1956)

found it unnecessary to introduce any lag for the Rio Grande above

Del Norte. Given the similarity of the basins, no lag was incorporated 26 in the study models. Various adjustments in soil moisture capacity, in the period of release for stored runoff, and, later, in temperature and precipitation to improve the correlation between model and streamflow and magnitude of flow lead to basin specific models. The goal of the study was to settle upon a combination of parameters describing a hypothetical point whose runoff best represented basin wide runoff through time. Therefore, a wide range of adjustments to input and parameter values was undertaken - always checking to ensure that they were reasonable in view of actual conditions in the basin. In the end, the model run that yielded the predicted series that was most highly correlated with observed streamflow and that had the most nearly identical magnitude was considered the best, most representative model of the basin. CHAPTER 4

RESULTS AND DISCUSSION

The presentation of results will have the following form. First, a brief chronological outline of the early results will show how the final parameter selections were made. Second, the final model for each basin will be described. These descriptions will be followed by comparisons of the predicted series and observed streamflow. Finally, the results of the scenarios will be presented.

Early Results

The initial results were disappointing but not dis- couraging. Using six inches of soil moisture capacity, Penman- compute a evapotranspiration (without modification for snowmelt), and

unaltered temperature and precipitation records, and the Budyko-

Sellers equation without any provision for snow, the predicted series

had one-eighth to one-quarter the magnitude of the observed discharge

and showed no improvement in correlation with gaged flow over straight

precipitation. The snowmelt modification was introduced into the model

at that time and a series of runs testing the release of snow storage

over one, two, three, and four months was made. The resulting series

had even smaller means and correlated considerably less well than

precipitation with the observed runoff. It was apparent that releasing

all stored runoff in one month tended to result in higher correlations

27 28 and larger means than multiple month releases. During this time smaller values of soil moisture capacity were assigned to the models thereby increasing the amount of runoff marginally. It was at this time too that the switch to 0.7 pan evaporation coefficient was made.

It clearly became apparent that using climatic input from divisions

2902 and 2907 would not produce satisfactory results in modeling the

Pecos River. In view of the small contribution to ultimate flow made by the watershed and the poor performance of the climatic data, it was decided to stop efforts to model the Pecos River.

At this stage a change in approach was introduced. Since the majority of runoff derived from snowmelt, but the model with the snow modification was not producing results that rivaled precipitation records, two conclusions were possible. One was that the modification made to reflect conditions of heavy snow accumulation was ineffective.

The other conclusion was that the climatic input was not representative of the major runoff producing areas. It seemed logical to investigate the second conclusion before discarding or changing the snow modification as it was widely used with the Thornthwaite equation.

There was also evidence that the second conclusion was correct in that the station with the highest mean precipitation had only 18 inches while numerous descriptions of the area referred to 25 to 30 inches.

Further, vast areas of the study watersheds were higher than the elevations of the weather stations.

In investigating the conclusion that the climatic input was not representative of the runoff producing areas, the temperature 2 9 and precipitation data were altered to simulate conditions of other, higher elevations. This alteration involved subtracting a constant from the temperature and multiplying the precipitation by another constant. The subtraction constant represented the rate of cooling with increased elevation. Arriving at a rate that described the conditions in the study area was difficult. The ambient lapse rate for the atmosphere is -0.6 ° C/100 meters or -3.4 ° F/1000 feet. The lapse rate between Del Norte (7,884 feet) and Hermit (9,000 feet) and between

Chama (7,850 feet) and Hermit is -7.6 ° F/1000 feet as derived from mean temperature records. It is very difficult to surmise the actual lapse rate for very mountainous terrain, where thermal inversions and cold air drainage and pooling are common, from a few, distant weather stations. In view of this difficulty, I settled on a lapse rate of

-5.5 ° F/1000 feet which split the difference between the ambient rate and the rate derived from temperature records. With a topographic map as a reference, a series of adjustments to the records were undertaken attempting to find the center of mass of the runoff in the basins.

Since precipitation varied with elevation in an unknown manner, it was not possible to measure the area of various elevation intervals and multiply them by a precipitation rate to find the centroid of precipitation. The rough areas of elevation intervals could be used to intuitively estimate where the center of mass of runoff might be.

These estimations were verified by statistical comparison of results with observed streamflow and by comparison of the monthly distribution of observed and predicted runoff. Having discovered what appeared to 30 be an optimum correlation for a watershed by temperature adjustment, the precipitation for that basin was multiplied by a constant to produce a predicted runoff series with a mean nearly equal to the actual.

Care was taken to ensure that this constand did not cause the modified precipitation to have a mean that exceeded thirty inches.

In this trial and error manner, parameter values were assigned and adjustments to climatic records made that produced basin specific models. Because there are four factors involved, other combinations of slightly different parameter values and climatic constants may produce similar results. Evidence of some one hundred trial runs suggests, however, that any other combinations that did as well as the ones settled on in this study would not differ drastically from the values chosen here.

Description of Basin Specific Models

Parameters of the three watershed models are presented in

Table 2. For the Rio Chama model, temperature records from Chama were adjusted using a lapse rate of -5.5'F/1000 feet to simulate conditions at roughly 9000 feet, the proposed center of mass of the watershed runoff. The Chama precipitation record was increased by a factor of

1.25. The Chama temperature record was also used for the Rio Grande and Conejos River models. In both cases, temperatures were adjusted to simulate conditions at 11,000 feet. The precipitation input to the

Conejos model was an arithmetic average of the records from Chama and

Hermit which was then multiplied by 1.67 to yield the same mean as the observed streamflow. The Rio Grande model employed an average of 31

Table 2. Parameters of the Basin-Specific Models for the Rio Grande, Rio Chama, and Conejos River.

RIO GRANDE CONEJOS RIVER RIO CHAMA

SOIL MOISTURE CAPACITY 2 2 4 (INCHES)

MINIMUM ELEVATION 7,980 8,272 6,692 (FEET)

MAXIMUM ELEVATION 13,830 13,180 12,760

ELEVATION OF ESTIMATED 11,000 11,000 9,000 CENTER OF MASS OF RUNOFF

ANNUAL MEAN OF ADJUSTED 22.33 30.16 23.73 PRECIPITATION (INCHES) 32

Chama, Hermit, and Saguache precipitation increased by a factor of

1.5. Averaged precipitation records were adopted because they better results than individual stations. In the case of the Conejos, no suitable weather stations were available within the watershed. A comparision of Conejos streamflow with the predicted series from Hermit to the north-west and from Chama to the south suggested that the rainfall experienced within the basin paralleled that to the north some years and that to the south others. Still other years were clearly different than either record. As for the Rio Grande model, the average of three well-spaced weather stations probably gave the best results because of the basin's vast, 1320 square miles, area.

Comparison of Model Predictions with Streamflow

The predictions of all three basin specific models agree quite favorably with the observed runoff. The predicted runoff is plotted with observed streamflow and precipitation in Figures 5, 6, and 7. Statistics for all three series are summarized in Table 3.

The model of the Rio Chama did very well in predicting actual streamflow. Small discrepancies such as 1537 and 1963 are attributable to areal variation in precipitation. The area around Chama was drier than other parts of the basin in those years. Similarly Chama was probably anomolously wet in 1967. The greater amplitude of the model

predictions probably reflects their genesis from point climatic input.

In contrast, streamflows are a natural average of conditions through- out the watershed with a resulting dampening of local extremes. 33

0-) CN

•-1 CO 31+

-o CNI z e•-n • CO 0 0 • 4_J 4J D eir) 1- CL CD M • —

C\J

S3-I3 \ I 35

LID D LID D LrD

..0 V) Ca.) L. 0 C •.'" 4- (r) (1) 4., Cn (7) -0 (!) •r-.1 1- a) 0 S- E _C CL. L- (1) 0- (1) 0- CI -17 7 0 **•••". CN 1:n C (1.) c 4-) -0 4-, 0 (0 3 4-1 • N. (1) L_ CT) CT)

LID D LID 0 LID Li_ Cr) CN CN

—1 J _7 nN 36

Table 3. Comparative Statistics for Observed Streamflow Records from Study Watersheds and Models.

RIO CHAMA RIO GRANDE CONEJOS RIVER

MEAN DISCHARGE (ACRE-FEET)

OBSERVED RECORD 266,648 640,160 239,582 MODEL 262,108 664,013 239,061 STANDARD DEVIATION

OBSERVED RECORD 132,015 219,503 81,599 MODEL 169,572 283,360 91,473 SKEW

OBSERVED RECORD 0.867 0.001 0.153 MODEL 0.953 0.333 0.291 CORRELATION COEFFICIENT

PRECIPITATION WITH 0.75 0.71 0.77 OBSERVED RECORD MODEL DISCHARGE WITH 0.91 0.83 0.84 OBSERVED RECORD

SAMPLE SIZE 35 65 65 37

The Conejos River and the Rio Grande models do a good job in predicting annual discharge. Salient discrepancies between predictions and observed flows occur in 1915 and 1937-1938. April 1915 was the wettest April on record in Chama and Saguache for the period

1915-1979. It was also quite wet in Hermit. Why neither the Magote gage on the Conejos River or the Del Norte gage on the Rio Grande reflect how wet 1915 was is unclear. The model prediction for 1928 is also at odds with the observed flow on the Conejos River. This may be an artifact of discretizing time. September 1927 was nearly twice as wet as any other September (for the period 1915-1979). Hermit and

Chama had 7-8 inches of rain, including approximately 2.5 inches in the last week of the month. Perhaps there was enough lag time that significant parts of water year 1927's precipitation appeared as 1928's streamflow. Since the model incorporates no lag, 1927 precipitation becomes 1927 runoff. Such a lag would neatly account for over- prediction for 1927 and under-prediction for 1928. The Annual Summaries of the Weather Bureau hold the explanation of the 1937-1938 discrepancies between model and gage. Although both years were slightly below average in precipitation, heavy snows occurred in the San Juan

Mountains. Hermit did not experience this heavy precipitation which seems to have been centered to the south of Cumbres, Colorado. In

February and March 1938, Cumbres had about 24 inches of precipitation while Hermit had but three. Quick runoff of the resulting deep snows

lead to high flows. Again, the cause of discrepancies is areal vari-

ation of precipitation. 38

The Rio Grande model's over-prediction for 1957 may be a

consequence of using monthly data. February 1957 was an extremely

mild month, 8 ° F above average in many parts of Colorado and New Mexico.

Daily high temperatures at Chama were, in the main, above 45 ° F for

both February and March. Quite possibly considerable snowmelt and

evaporation occurred as a result of these high temperatures. Night-

time lows remained cold, and monthly means when adjusted to simulate

11,000 feet conditions did not exceed 30.2 ° F. Thus the model missed

the suggested evaporation and over-predicted runoff in April and May.

Scenario Results

The success of the model predictions was reassuring

and lent confidence to further investigations of the scenarios. The

results of the climatic change scenarios are summarized in Table 4.

The results from the three watersheds show good agreement with each other. In particular, the most likely (and worst case) scenario

(warmer-drier) shows a 30 2 decline in mean annual discharge. That

translates into a 80,000AF decline on the Rio Chama, a 65,000AF decline on the Conejos River, and a 205,000AF decline on the Rio

Grande. Such a change would have severe impact on the people of the

Rio Grande Basin. The warmer-wetter scenario (+3.6 ° F, +10% precipitation) results reveal slight increases in water supply. Flohn (1981),

Kellogg (1982), and Revelle (1982) all describe this scenario as unlikely for inland areas at the upper Rio Grande's latitudes. It seems possible that any slight increase in water supply under this scenario might be offset by increased consumptive use by agricultural 39

Table 4 • Mean Annual Discharge of Scenario Investigations and the Percentage Change from Original Model.

MEAN DISCHARGE PERCENTAGE (ACRE-FEET) CHANGE

RIO CHAMA 266,648

MODEL 262,108 WARMER-DRIER 183,579 -30% WARMER-WETTER 289,251 COOLER-WETTER 362,186 +38%

CONEJOS RIVER 239,582

MODEL 239,061 WARMER-DRIER 175,366 -27% WARMER-WETTER 245,904 + 3 2 COOLER-WETTER 301,853 +26%

RIO GRANDE 640,160

MODEL 664,013 - WARMER-DRIER 451,264 -32% WARMER-WETTER 664,013 0% COOLER-WETTER 872,256 40 crops due to warmer temperatures. The Rio Chama shows the greatest

increase in runoff in the warmer-wetter scenario because, having the

lowest elevation for the center of runoff mass, it had the highest

ratio of evapotranspiration to potential evapotranspiration (ET/PE)

for the period of snowmelt runoff in the original model. With the change of temperature, ET/PE increased for the other two watersheds,

but runoff began a month earlier for the Chama model, and ET/PE actually decreased. It is interesting that the average change in

runoff for the cooler-wetter scenario is almost identical to the absolute value of that for the warmer-drier scenario. Were this

cooler scenario to come about, water supplies for the Rio Grande

Basin would be greatly improved, but, as cited earlier, atmospheric

carbon dioxide increases are not thought likely to lead to cooling.

The full impact of these changes is perhaps best under-

stood When cumulative departures from the mean are plotted basin

Figures 8, 9, and 10. Using the mean from each model, cumulative

departures from that mean were computed for each of the scenarios

derived from that model. Most striking is the 13.8 million acre-

feet (MAF) deficit on the Rio Grande in the warmer-drier scenario.

It should be remembered that the estimated natural flow for the Rio

Grande at the Texas-New Mexico border is only 2.2 MAF. The combined

65-year deficit for the Rio Grande and the Conejos River is 17.9 MAF,

over eight times the estimated natural flow. More dramatic, this

combined deficit is almost nineteen times the estimated natural flow

at the Colorado-New Mexico border. 41

C\.1 Cr") "C) (1) 1.) Q) "C) o G) o

1 1 1 I 1 I 1 1 1 1 1 1 1 1 1

.,--i 1.---4 CT-) CN1 D CN Cr) I I I - 4

C CO L) • CC a) 0 -- LJ 4-n L. MI D >"'"' c a) V) - L) - 17) s- faS V) 0— I 1— 1— su

s_E0 Ea) 1.- 4- CO I CD S•C) ..••n E _ LLCO s—

4.1 L IO 0 a) I_ CI CO C 0) 0)3 D > 00 - C\J 4-1 CT) COL E SD CO 4-)0) E 4-' s_ • 4-)

CD CT) e••••

N011 -11 A 143

1 1 1 1 I 1 1 1 1 I 1 1 1 1 LID LID 44

Were another drought similar to the period 1950-1957 to recur under the warmer-drier scenario, the consequences would be dire.

Actual total streamflow over those years constituted only 69% of mean discharge. Streamflow for the same period would be reduced to a mere

38% of the mean under warmer-drier conditions. In the same vein of natural climatic variation compounding the effects of anthropogenic change, it should be noted that tree-ring data show that the period

1890-1925 was the wettest such period in the last 300 years. Since the first ten years of data used in the Rio Grande and Conejos River models falls within that extreme period, it is quite possible that actual flow reductions in the warmer-drier scenario may be more severe than predicted. CHAPTER 5

CONCLUSION

Many climatologists now view as likely that increased concentrations of carbon dioxide in the atmosphere will lead to warmer, drier climates in areas like the upper Rio Grande Basin. The time envisioned for such change is not great. Kellogg (1982) states that the effects of change will begin to be evident within twenty years.

The conservative estimate of temperature and precipitation changes employed in this study produced a thirty percent decrease in streamflow for the upper Rio Grande Basin. In view of the present total allocation of surface water resources in the area, such a reduction would have profoundly adverse effects on the people, argiculture, and industry of the basin. Water allocation in the basin is strictly regulated by a complex web of international treaties, federal law, interstate compacts, and local law. Making changes in allocation to cope with climatic change under this system will be difficult and time-consuming. It is therefore imperative that the governmental bodies involved begin to study and plan for the possibility of man- induced climatic change.

45 LIST OF REFERENCES

Anderson, R., 1982. Conflict Between Establishment of lnstream Flows and Other Water Uses on Western Streams, Water Resources Bulletin, February 1982: pp. 53-60.

Budyko, M. I., 1956. Tepolovi Balans Zemnoi Poverkhnosti, Gidrometer- eologicheskoe lzdatel'stvo, Leningrad. (English translation: Stepanova, N.A., 1958. The Heat Balance of the Earth's Surface, Office of Technical Services, United States Department of Commerce.

Flohn, H., 1981. Major Climatic Events Associated with a Prolonged Oak Ridge CO2 -induced Warming, Institute for Energy Analysis, Associated Universities.

Gieger, R., 1965. The Climate Near the Ground, Harvard University Press, p. 611.

Hansen, J., Johnson, D., Lacis, A., Lebedeff, S., Lee, P., Rind, D., Russell, G., 1981. Climatic Impact of Increasing Atmospheric Carbon Dioxide, Science 213(4511), August 28, 1981: pp. 957-966.

Jones, E. B., Shafer, B. A., Rango, A., Frick, D. M., 1981. Application of a Snowmelt Model to Two Drainage Basins in Colorado, Proceedings ofIthe Western Snow Conference, Colorado State University: pp. 43-54.

Kellogg, W. W., Schware, R., 1982. Society, Science and Climate Change, Foreign Affairs, Summer 1982: pp. 1076-1109.

Mather, J. R., 1978. The Climatic Water Budget in Environmental Analysis, Lexington Books: p. 239.

, 1979. Use of the Climatic Water Budget to Estimate Streamflow, University of Delaware Water Resources Center: P. 52.

National Defence University, 1978. Climatic Change to the Year 2000, Government printing Office: p. 109.

New Mexico State Office of Engineering, 1967. Water Resources of New Mexico, State Planning Office: P. 321.

46 47

Revelle, R., 1982. Carbon Dioxide and World Climate, Scientific American, 247(2), August 1982: pp. 35-44.

Sellers, W. D. ,1965. Physical Climatology, The University of Chicago Press: p. 272.

Stone, L. A., Albrecht, J. C., Yoshioka, G. A., 1971. Computer Programs for the Climatic Water Balance, Publications in Climatology, 14(3).

Thornthwaite, C. W. ,Mather, J. R., 1957. Instructions and Tables for Computing Potential Evapotranspiration and the Water Balance, Publications in Climatology, 10(3).

United States Department of Commerce, 1968. Weather Atlas of the United States, Gale Research Co.: p. 262.

United States Geologic Survey, 1960. Compilation of Records of Surface Waters of the United States through September 1950 - Part 8. Western Gulf of Mexico Basins, Water Supply Paper 1312, Government Printing Office.

, 1964. Compilation of Records of Surface Waters of the United States, October 1950 to September 1960 - Part 8. Western Gulf of Mexico Basins, Water Supply Paper 1732, Government Printing Office.

, 1970. Surface Water Supply of the United States 1961-1965 - Part 8. Western Gulf of Mexico Basins, Water Supply Paper 1923, Government Printing Office.

, 1974. Surface Water Supply of the United States

19661970 - Part 8. Western . GUlf Of México Basins, Water Supply Paper 2123, Government Printing Office.

, 1971-1980. Water Resources Data for Colorado, Government Printing Office.

, 1971-1980. Water Resources Data for New Mexico, Government Printing Office.

United States Water Resources Council, 1977. Guidelines for determining Flood Frequency, Bull. 17-A, Hydrology Committee (Rev. 1477), Water Resources Council, Government Printing Office.

, 1978. The Nation's Water Resources - . Rio Grande Region, Government Printing Office: p. 117, 48 van Hylckama, T. E. A., 1956. The Water Balance of the Earth, Publications in Climatology, 9(2).

West, S. W., Broadhurst, W. L., 1975. Summary Appraisals of the

Nation's Ground-Water Resources - Rio Grande Region, United States Geologic Survey Professional Paper 813-D, Government Printing Office: p. 39.