Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability

Jorge L. Peña-Arancibia; Tim R. McVicar; Zahra Paydar; Lingtao Li; Juan P. Guerschman; Randall J. Donohue; Dushmanta Dutta; Geoff M. Podger; Albert I.J.M. van Dijk; Francis H.S. Chiew

doi:10.1016/j.rse.2014.08.016

Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability

Jorge L. Peña-Arancibia^*, Tim R. McVicar, Zahra Paydar, Lingtao Li, Juan P. Guerschman, Randall J. Donohue, Dushmanta Dutta, Geoff M. Podger, Albert I.J.M. van Dijk, Francis H.S. Chiew

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

42 Citations (Scopus)

Abstract

Globally, irrigation accounts for more than two thirds of freshwater demand. Despite their importance, the geographic distribution and water use of many important irrigated areas remains uncertain. This paper advances a methodology to map summer cropping irrigated areas experiencing extreme climate and different management practices in the Murray-Darling Basin (Australia). A Random Forest classification model was implemented to austral hemisphere summer irrigated areas for the water-years 2004/05 to 2010/11. The methodology used training samples from Landsat TM/ETM+ reflectance data and monthly time-series of remotely-sensed observations from the MODerate resolution Imaging Spectroradiometer (MODIS). The covariates included in the classification model characterised the monthly dynamics and rates of change of: (i) the vegetation phenology via the recurrent and persistent fractions of photosynthetically active radiation (fPAR_rec and fPAR_per, respectively); (ii) water use via remotely-sensed estimates of actual evapotranspiration (ET_a), precipitation (P) and the difference between ET_a and P. Observed agreement - in terms of the kappa coefficient - for correctly classified pixels in the training sample was 96%. Independent comparisons of yearly irrigated area estimates showed linear relationships with Pearson's correlation coefficients (r) generally greater than 0.7 for: (i) reported areas; (ii) areas with available metered water withdrawals; and (iii) estimates of agricultural yields. Sequential covariate optimisation suggested that the most important predictors to identify of irrigation areas included the emergence-senescence period (as determined by the fPAR_rec and corresponding rates of change) and the ET_a surplus over P during this period. The latter can be important when identifying supplementary irrigation due to periodically unreliable water supply in areas with otherwise high precipitation that are in-phase with summer crop growth.

Original language	English
Pages (from-to)	139-152
Number of pages	14
Journal	Remote Sensing of Environment
Volume	154
DOIs	https://doi.org/10.1016/j.rse.2014.08.016
Publication status	Published - 2014

Access to Document

10.1016/j.rse.2014.08.016

Cite this

Peña-Arancibia, J. L., McVicar, T. R., Paydar, Z., Li, L., Guerschman, J. P., Donohue, R. J., Dutta, D., Podger, G. M., van Dijk, A. I. J. M., & Chiew, F. H. S. (2014). Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability. Remote Sensing of Environment, 154, 139-152. https://doi.org/10.1016/j.rse.2014.08.016

@article{f5c80df72819424a983a1cf752c22435,

title = "Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability",

abstract = "Globally, irrigation accounts for more than two thirds of freshwater demand. Despite their importance, the geographic distribution and water use of many important irrigated areas remains uncertain. This paper advances a methodology to map summer cropping irrigated areas experiencing extreme climate and different management practices in the Murray-Darling Basin (Australia). A Random Forest classification model was implemented to austral hemisphere summer irrigated areas for the water-years 2004/05 to 2010/11. The methodology used training samples from Landsat TM/ETM+ reflectance data and monthly time-series of remotely-sensed observations from the MODerate resolution Imaging Spectroradiometer (MODIS). The covariates included in the classification model characterised the monthly dynamics and rates of change of: (i) the vegetation phenology via the recurrent and persistent fractions of photosynthetically active radiation (fPARrec and fPARper, respectively); (ii) water use via remotely-sensed estimates of actual evapotranspiration (ETa), precipitation (P) and the difference between ETa and P. Observed agreement - in terms of the kappa coefficient - for correctly classified pixels in the training sample was 96%. Independent comparisons of yearly irrigated area estimates showed linear relationships with Pearson's correlation coefficients (r) generally greater than 0.7 for: (i) reported areas; (ii) areas with available metered water withdrawals; and (iii) estimates of agricultural yields. Sequential covariate optimisation suggested that the most important predictors to identify of irrigation areas included the emergence-senescence period (as determined by the fPARrec and corresponding rates of change) and the ETa surplus over P during this period. The latter can be important when identifying supplementary irrigation due to periodically unreliable water supply in areas with otherwise high precipitation that are in-phase with summer crop growth.",

keywords = "Image classification, Irrigation, Mapping, Random Forest, Remote sensing",

author = "Pe{\~n}a-Arancibia, {Jorge L.} and McVicar, {Tim R.} and Zahra Paydar and Lingtao Li and Guerschman, {Juan P.} and Donohue, {Randall J.} and Dushmanta Dutta and Podger, {Geoff M.} and {van Dijk}, {Albert I.J.M.} and Chiew, {Francis H.S.}",

note = "Publisher Copyright: {\textcopyright} 2014.",

year = "2014",

doi = "10.1016/j.rse.2014.08.016",

language = "English",

volume = "154",

pages = "139--152",

journal = "Remote Sensing of Environment",

issn = "0034-4257",

publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability

AU - Peña-Arancibia, Jorge L.

AU - McVicar, Tim R.

AU - Paydar, Zahra

AU - Li, Lingtao

AU - Guerschman, Juan P.

AU - Donohue, Randall J.

AU - Dutta, Dushmanta

AU - Podger, Geoff M.

AU - van Dijk, Albert I.J.M.

AU - Chiew, Francis H.S.

PY - 2014

Y1 - 2014

N2 - Globally, irrigation accounts for more than two thirds of freshwater demand. Despite their importance, the geographic distribution and water use of many important irrigated areas remains uncertain. This paper advances a methodology to map summer cropping irrigated areas experiencing extreme climate and different management practices in the Murray-Darling Basin (Australia). A Random Forest classification model was implemented to austral hemisphere summer irrigated areas for the water-years 2004/05 to 2010/11. The methodology used training samples from Landsat TM/ETM+ reflectance data and monthly time-series of remotely-sensed observations from the MODerate resolution Imaging Spectroradiometer (MODIS). The covariates included in the classification model characterised the monthly dynamics and rates of change of: (i) the vegetation phenology via the recurrent and persistent fractions of photosynthetically active radiation (fPARrec and fPARper, respectively); (ii) water use via remotely-sensed estimates of actual evapotranspiration (ETa), precipitation (P) and the difference between ETa and P. Observed agreement - in terms of the kappa coefficient - for correctly classified pixels in the training sample was 96%. Independent comparisons of yearly irrigated area estimates showed linear relationships with Pearson's correlation coefficients (r) generally greater than 0.7 for: (i) reported areas; (ii) areas with available metered water withdrawals; and (iii) estimates of agricultural yields. Sequential covariate optimisation suggested that the most important predictors to identify of irrigation areas included the emergence-senescence period (as determined by the fPARrec and corresponding rates of change) and the ETa surplus over P during this period. The latter can be important when identifying supplementary irrigation due to periodically unreliable water supply in areas with otherwise high precipitation that are in-phase with summer crop growth.

AB - Globally, irrigation accounts for more than two thirds of freshwater demand. Despite their importance, the geographic distribution and water use of many important irrigated areas remains uncertain. This paper advances a methodology to map summer cropping irrigated areas experiencing extreme climate and different management practices in the Murray-Darling Basin (Australia). A Random Forest classification model was implemented to austral hemisphere summer irrigated areas for the water-years 2004/05 to 2010/11. The methodology used training samples from Landsat TM/ETM+ reflectance data and monthly time-series of remotely-sensed observations from the MODerate resolution Imaging Spectroradiometer (MODIS). The covariates included in the classification model characterised the monthly dynamics and rates of change of: (i) the vegetation phenology via the recurrent and persistent fractions of photosynthetically active radiation (fPARrec and fPARper, respectively); (ii) water use via remotely-sensed estimates of actual evapotranspiration (ETa), precipitation (P) and the difference between ETa and P. Observed agreement - in terms of the kappa coefficient - for correctly classified pixels in the training sample was 96%. Independent comparisons of yearly irrigated area estimates showed linear relationships with Pearson's correlation coefficients (r) generally greater than 0.7 for: (i) reported areas; (ii) areas with available metered water withdrawals; and (iii) estimates of agricultural yields. Sequential covariate optimisation suggested that the most important predictors to identify of irrigation areas included the emergence-senescence period (as determined by the fPARrec and corresponding rates of change) and the ETa surplus over P during this period. The latter can be important when identifying supplementary irrigation due to periodically unreliable water supply in areas with otherwise high precipitation that are in-phase with summer crop growth.

KW - Image classification

KW - Irrigation

KW - Mapping

KW - Random Forest

KW - Remote sensing

UR - http://www.scopus.com/inward/record.url?scp=84945337520&partnerID=8YFLogxK

U2 - 10.1016/j.rse.2014.08.016

DO - 10.1016/j.rse.2014.08.016

M3 - Article

SN - 0034-4257

VL - 154

SP - 139

EP - 152

JO - Remote Sensing of Environment

JF - Remote Sensing of Environment

ER -

Dynamic identification of summer cropping irrigated areas in a large basin experiencing extreme climatic variability

Abstract

Access to Document

Other files and links

Fingerprint

Cite this