A Mass-Balance, Watershed-Scale Analysis of

the Chemistry of Adirondack Lakes

Discussion - Day 5

Patterns and Consequences of Variation in Lake DOC

DOC in 1470 lakes sampled by the Adirondack Lake Survey Corporation (ALSC) in the 1980’s ranged from 0.2 - 35 mg/L.

• Lakes within a given region can vary dramatically in dissolved organic carbon (DOC) concentrations.

Patterns and Consequences of Variation in Lake DOC

Low DOC lakes typically have- high light penetration,

- higher pH (unless acidified by mineral acids), and

- richer oxygen conditions

Low DOC lakes are also more susceptible to - acidification,

- eutrophication, and

- UV-light effects

Acid deposition has reduced DOC in many north temperature lakes, with attendant increases in UV light penetration.

High DOC lakes appear to have higher levels of contamination with mercury

Patterns and Consequences of Variation in Lake DOC

Sources and Fates of Lake DOC

The bulk of the DOC in lakes originates from decomposition in wetland and upland ecosystems within the watershed.

As it moves from uplands to lakes, DOC links terrestrial, wetland, littoral, and open water habitats.

Lakes export far less DOC than they import. Fates of DOC in lakes include photolysis, decomposition, flocculation and sedimentation.

Objectives

 

Most previous studies of variation in lake DOC have relied on multiple regression models.

We developed an alternative approach, based on mass balance principles, that focuses on the inputs and losses of DOC.

Our most basic objective was to understand how the composition and spatial configuration of the upland and wetland vegetation within a watershed influences DOC concentrations within each lake.

Approach

Our approach takes advantage of data sets available for the watersheds of over 600 lakes in the Adirondack Mountains of New York.

The approach is spatially-explicit, and divides each watershed into 10 x 10 m grid cells.

Our analysis estimates the loading of DOC to the lakes as a function of:- the type of vegetation in each grid cell, and- the flow-path distance from the cell to the

lakeshore.

Map of the roads and boundary of the Adirondack Park. Upland vegetation types of the 610 sampled watersheds are indicated in shades of green. Sampled lakes are shown in blue.

Within Individual Watersheds

Vegetation Cover Type

Relative Cover over all

Watersheds (%)Median

Cover (%)Minimum

Cover (%)Maximum Cover (%)

Upland Vegetation

Deciduous Forest 30.20 25.9 0.0 84.5

Mixed Forest 41.67 39.2 0.0 94.2

Conifer Forest 12.13 7.4 0.0 93.0

Deciduous / Open 1.73 0.0 0.0 56.0

Open Uplands 4.36 1.4 0.0 67.7

Wetlands

Open Water 0.08 0.0 0.0 7.6

Emergent Marsh 0.67 0.0 0.0 22.0

Deciduous Forest Swamp 1.01 0.0 0.0 17.1

Conifer Forest Swamp 4.43 2.4 0.0 79.9

Dead Tree Swamp 0.16 0.0 0.0 15.1

Deciduous Shrub Swamp 1.99 0.4 0.0 58.8

Broadleaved Evergreen Shrub Swamp 0.70 0.0 0.0 56.5

Needle-leaved Evergreen Shrub Swamp 0.85 0.0 0.0 96.9

Total Area (ha) 42,172.75

A Mass Balance Model of Variation in Lake DOC

•We assume that lake DOC is in approximate steady state from year-to year. Thus, inputs to a lake should approximately equal outputs.

•Inputs:

• Within-lake annual net production (assumed to be a linear function of lake area)

•Input from wetlands and upland vegetation within the watershed

•Outputs:

• Lake discharge

• Within-lake degradation

At Steady State:

)))

1-33

(yr k) Rate (Flushing * (m Volume Lake

(g/yr) Loading (g/m DOC

Inputs to Headwater Lakes

for j = 1..n pixels (100 m2) of i = 1..c vegetation types, where Dij is the flowpath distance from pixelij to the lakeshore LakeArea is lake surface area (in m2), p1 is the estimated within-lake DOC production (g/m2), Ei is the estimated DOC export (g/100 m2) of vegetation type i,

and i, and i are estimated parameters that

determine the decline in DOC loadingof vegetation type i with distance from the lake

iijiD

c

i

n

ji e*E)LakeArea*p(Inputs

1 1

1

0

20

40

60

80

100

120

0 50 100 150 200

Distance from Lake (m)

Lo

adin

g (

kg/h

a)

Lake DOC Outputs

• Lake Discharge• Discharge = DOC concentration * lake volume * flushing

rate

• Within-Lake Degradation• Degradation = DOC concentration * lake volume * k

• Also allow k to vary as a function of

•ANC: k = a + b*ANC

•Depth: k = a * exp(-b*depth)

Adding Upstream Lakes – a recursive model

Include watersheds that have embedded ponds within the watershed

For sampled ponds that have other ponds immediately upstream, add inputs from the immediately upstream pond to the loading term (number of immediately upstream ponds is as high as 8, along a branching stream)

Add a term to the model to estimate the percentage of upstream inputs that actually make it to the lake

Parameters Estimated by the Analysis

• 3 parameters for each vegetation type

• 1 parameter for within-lake production

• 1-2 parameter(s) for within-lake decay (k)

• 3 parameters to account for interannual variation in total loading

• = total of 41-42 parameters when using 12 cover types

• Solve for the parameter values that provide the best fit to the observed variation in lake DOC (i.e. maximize the likelihood of observing the dataset) using simulated annealing (a global optimization procedure)

Model comparison

Model # Lakes # Parameters Likelihood AICcorr R2 SlopeHEADWATER LAKESBasic Model: Total Distance 355 41 -818.67 1730.34 0.551 1.012

Basic Model: Ground Distance 355 41 -823.13 1739.26 0.538 0.999

Basic Model: Stream Distance 355 41 -832.52 1758.05 0.509 1.005

Basic Model + Depth 355 42 -814.92 1725.41 0.555 0.994

Basic Model + Wetland Loading 355 42 -816.11 1727.79 0.551 1.001

Basic Model + ANC1 348 42 -782.53 0.542 1.003

Reduced Model: No Distance Decay 355 17 -838.26 1712.33 0.498 0.997

Reduced Model: 5 Types Vary 355 22 -824.60 1696.26 0.530 1.011

ALL LAKESBasic Model: Total Distance 428 29 -1040.21 2142.79 0.477 0.995

Reduced Model: No Distance Decay 428 18 -1046.16 2129.99 0.461 0.996

1 ANC model compared against basic model with just 348 lakes

Headwater Lakes - Reduced Model

0

100

200

300

400

500

0 50 100 150 200 250

Distance from Lake (m)

DO

C L

oadi

ng (

kg/h

a/yr

)

Deciduous Forest

Mixed Forest

Conifer Forest

Deciduous / Open

Open Vegetation

Emergent Marsh

Deciduous Forest Swamp

Conifer Forest Swamp

Dead Tree Swamp

Deciduous Shrub Swamp

Broadleaved Evergreen ShrubSwamp

Needle-leaved EvergreenShrub Swamp

Predicted loading as a function of distance

Headwater Lakes (n=355)

y = 1.0122x

R2 = 0.5459

0

5

10

15

20

25

30

0 5 10 15 20

Predicted DOC

Ob

serv

ed D

OC

Goodness of fit

0

50

100

150

200

250

300

350

400

DF MF CF

DO OVEM

DFSCFS

DTSDSS

BESS

NESS

Cover Type

DO

C L

oa

din

g (

kg/h

a/y

r)

532.1

Upland cover types wetlands

Estimated loading from different cover types (all lakes)

Error bars are 2-unit support limits

Results from Alternative Models

• Significantly worse fits produced when:

• assuming no decline in loading with distance for all cover types

• Using a “topographic index” to identify and limit inputs to areas likely to have saturated soils

• Limiting distance to specified distances from the lake

• Calculating distance to nearest open water (stream or lakeshore) rather than all the way to the lakeshore

Effects of In-Lake Processes

Net In-Lake DOC Production:

- basic model estimate: 12.4 kg/ha (95% S.I. = 0 – 26.9)

In-Lake Decay

- basic model estimate: 0.82 (95% S.I. = 0.69 – 1.00)

- declines significantly with depth

- marginal increase with fraction of loading from wetlands

- hint of increase with ANC

Effect of Lake Depth on K

0.0

0.5

1.0

1.5

2.0

0 5 10 15 20

Mean Lake Depth (m)

In-L

ake D

ecay C

oeff

icie

nt

(k)

Effect of Wetland Loading on In-Lake Decay

0.6

0.7

0.8

0.9

1.0

1.1

1.2

0 20 40 60 80 100

% of DOC Loading from Wetlands

k (

/yr)

CONCLUSIONS

• Loading of DOC from upland vegetation types associated with disturbance (logging, beech bark disease, and limited lakeshore development)

•was very high when the disturbance was immediately adjacent to the lake,

•but declined dramatically with distance from the lake

• Similar patterns for wetlands dominated by dead trees (beaver ponds)

CONCLUSIONS

• However, these small headwater watersheds (most with thin, glacial till soils) are very well “plumbed”. Inputs do not decline significantly with distance from the lake for the major wetland and forest types

• There was significant variation in DOC loading from different vegetation types, however:• The main closed forest types (deciduous, mixed, conifer)

had approximately equal predicted loading of ~ 50 kg/ha/yr, and

• 4 of the most common wetland types also had approximately equal predicted loading of ~ 200 kg/ha/yr

• Thus, as expected, wetlands generally export much more DOC per unit area than uplands

• However, as a result of the much larger area of uplands (~ 90% of drainage area), upland forests are the dominant source of DOC in these lakes

CONCLUSIONS