Evaluation of soil carbon dynamics after forest cover change in CMIP6 land models using chronosequences
Boysen, Lena R. ; Brovkin, Victor ; Warlind, David LU
; Peano, Daniele
; Lanso, Anne Sofie
; Delire, Christine
; Burke, Eleanor
; Poeplau, Christopher
and Don, Axel
(2021)
In Environmental Research Letters
16(7).
- Abstract
Land surface models are used to provide global estimates of soil organic carbon (SOC) changes after past and future change land use change (LUC), in particular re-/deforestation. To evaluate how well the models capture decadal-scale changes in SOC after LUC, we provide the first consistent comparison of simulated time series of LUC by six land models all of which participated in the coupled model intercomparison project phase 6 (CMIP6) with soil carbon chronosequences (SCCs). For this comparison we use SOC measurements of adjacent plots at four high-quality data sites in temperate and tropical regions. We find that initial SOC stocks differ among models due to different approaches to represent SOC. Models generally meet the direction of... (More)
Land surface models are used to provide global estimates of soil organic carbon (SOC) changes after past and future change land use change (LUC), in particular re-/deforestation. To evaluate how well the models capture decadal-scale changes in SOC after LUC, we provide the first consistent comparison of simulated time series of LUC by six land models all of which participated in the coupled model intercomparison project phase 6 (CMIP6) with soil carbon chronosequences (SCCs). For this comparison we use SOC measurements of adjacent plots at four high-quality data sites in temperate and tropical regions. We find that initial SOC stocks differ among models due to different approaches to represent SOC. Models generally meet the direction of SOC change after reforestation of cropland but the amplitude and rate of changes vary strongly among them. The normalized root mean square errors of the multi model mean range from 0.5 to 0.8 across sites and 0.1-0.7 when excluding outliers. Further, models simulate SOC losses after deforestation for crop or grassland too slow due to the lack of crop harvest impacts in the models or an overestimation of the SOC recovery on grassland. The representation of management, especially nitrogen levels is important to capture drops in SOC after land abandonment for forest regrowth. Crop harvest and fire management are important to match SOC dynamics but more difficult to quantify as SCC rarely report on these events. Based on our findings, we identify strengths and propose potential improvements of the applied models in simulating SOC changes after LUC.
(Less)
- author
- Boysen, Lena R.
; Brovkin, Victor
; Warlind, David
LU
; Peano, Daniele
; Lanso, Anne Sofie
; Delire, Christine
; Burke, Eleanor
; Poeplau, Christopher
and Don, Axel
- organization
- publishing date
- 2021
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- chronosequences, land use change, model-data intercomparison, soil carbon
- in
- Environmental Research Letters
- volume
- 16
- issue
- 7
- article number
- 074030
- publisher
- IOP Publishing
- external identifiers
-
- scopus:85110462505
- ISSN
- 1748-9318
- DOI
- 10.1088/1748-9326/ac0be1
- language
- English
- LU publication?
- yes
- id
- 6c50cf85-f795-413b-a6ea-562af5ffcf22
- date added to LUP
- 2021-09-08 16:41:52
- date last changed
- 2023-03-08 00:40:16
@article{6c50cf85-f795-413b-a6ea-562af5ffcf22,
abstract = {{<p>Land surface models are used to provide global estimates of soil organic carbon (SOC) changes after past and future change land use change (LUC), in particular re-/deforestation. To evaluate how well the models capture decadal-scale changes in SOC after LUC, we provide the first consistent comparison of simulated time series of LUC by six land models all of which participated in the coupled model intercomparison project phase 6 (CMIP6) with soil carbon chronosequences (SCCs). For this comparison we use SOC measurements of adjacent plots at four high-quality data sites in temperate and tropical regions. We find that initial SOC stocks differ among models due to different approaches to represent SOC. Models generally meet the direction of SOC change after reforestation of cropland but the amplitude and rate of changes vary strongly among them. The normalized root mean square errors of the multi model mean range from 0.5 to 0.8 across sites and 0.1-0.7 when excluding outliers. Further, models simulate SOC losses after deforestation for crop or grassland too slow due to the lack of crop harvest impacts in the models or an overestimation of the SOC recovery on grassland. The representation of management, especially nitrogen levels is important to capture drops in SOC after land abandonment for forest regrowth. Crop harvest and fire management are important to match SOC dynamics but more difficult to quantify as SCC rarely report on these events. Based on our findings, we identify strengths and propose potential improvements of the applied models in simulating SOC changes after LUC. </p>}},
author = {{Boysen, Lena R. and Brovkin, Victor and Warlind, David and Peano, Daniele and Lanso, Anne Sofie and Delire, Christine and Burke, Eleanor and Poeplau, Christopher and Don, Axel}},
issn = {{1748-9318}},
keywords = {{chronosequences; land use change; model-data intercomparison; soil carbon}},
language = {{eng}},
number = {{7}},
publisher = {{IOP Publishing}},
series = {{Environmental Research Letters}},
title = {{Evaluation of soil carbon dynamics after forest cover change in CMIP6 land models using chronosequences}},
url = {{http://dx.doi.org/10.1088/1748-9326/ac0be1}},
doi = {{10.1088/1748-9326/ac0be1}},
volume = {{16}},
year = {{2021}},
}