Evaluation and merging of multiple gridded surface soil moisture products in Europe using ICOS measurements and triple collocation analysis
(2023) In Student thesis series INES NGEM01 20231Dept of Physical Geography and Ecosystem Science
- Abstract
- Surface soil moisture (SM) is an essential climate variable that plays a key role in ecosystems and the energy, water, and carbon cycles. SM can be accurately measured using in situ measurements. However, these measurements are globally not densely located over large areas, which would be required for accurate large-scale SM estimation due to the high spatial variability of SM. Instead, global atmospheric models and satellite remote sensing in the microwave range are commonly utilised for large-scale SM monitoring. Both model and satellite approaches have resulted in multiple gridded SM products at regional or global scales at various spatial resolutions (typically between 1 and 40 km). The accuracy of the gridded products varies over... (More)
- Surface soil moisture (SM) is an essential climate variable that plays a key role in ecosystems and the energy, water, and carbon cycles. SM can be accurately measured using in situ measurements. However, these measurements are globally not densely located over large areas, which would be required for accurate large-scale SM estimation due to the high spatial variability of SM. Instead, global atmospheric models and satellite remote sensing in the microwave range are commonly utilised for large-scale SM monitoring. Both model and satellite approaches have resulted in multiple gridded SM products at regional or global scales at various spatial resolutions (typically between 1 and 40 km). The accuracy of the gridded products varies over different regions, climates, and land covers, necessitating their evaluation. Evaluation with in situ data is limited to areas where measurements are available. Over the past 15 years, triple collocation analysis (TCA) has been extensively applied to evaluate gridded SM products among different geophysical variables, as it can estimate the error structure of three independent datasets without the need for in situ measurements. TCA has also been used to successfully merge gridded products to generate more accurate SM estimates. This study evaluated and ranked eight gridded SM products, including SMOS L4, SMAP L3E, SMAP L4, Sentinel-1, ASCAT, ESA CCI SM, ERA5-Land, and GLDAS Noah, using in situ measurements of SM taken during 2020-2021 from the Integrated Carbon Observation System (ICOS) station network. SMAP L4 and ERA5-Land generally performed the best with similar statistical scores. When comparing the products against absolute SM on collocated dates, SMAP L4 had a median ubRMSD of ca 0.047 m3/m3 and a median correlation coefficient of 0.73. ESA CCI SM and SMAP L3 gave slightly worse scores, while GLDAS Noah showed a relatively poor correlation against short-term SM anomalies. Sentinel-1 generally had the worst performance, together with ASCAT and SMOS L4. In addition, TCA was performed using a triplet consisting of SMAP L3E, ASCAT, and GLDAS Noah. The TCA reinforced the results of the in situ-based evaluation, with the lowest ubRMSE and highest SNR found for SMAP L3E among the three products. The TCA of this triplet was also used to produce a weighted merged dataset of SM estimates. When compared to ICOS measurements, the merged product performed better than GLDAS Noah and ASCAT but similar to or worse than SMAP L3E. Additionally, the TCA-weighted product performed similarly to a simple arithmetic mean, indicating the merging process was not worthwhile in the study area. (Less)
- Popular Abstract (Swedish)
- Mängden vatten i markens ytskikt, markfuktigheten, är viktig för ekosystem, jordbruk samt vattnets- och kolets kretslopp. Markfuktighet är även en indikator för torka och eftersom extremt väder ökar på grund av de globala klimatförändringarna läggs allt mer resurser på dess övervakning.
Övervakning av markfuktighet görs vanligen antingen via punktbaserad mätning direkt i marken eller via satelliter som mäter mängden vatten via naturliga (passiva system) eller satellitutskickade (aktiva system) mikrovågor. Flera satellitbaserade uppskattningar sammanställs till så kallade markfuktighetsprodukter som publiceras regelbundet. Det finns även produkter som istället för satellitdata utnyttjar atmosfäriska modeller som bygger på uppmätt data,... (More) - Mängden vatten i markens ytskikt, markfuktigheten, är viktig för ekosystem, jordbruk samt vattnets- och kolets kretslopp. Markfuktighet är även en indikator för torka och eftersom extremt väder ökar på grund av de globala klimatförändringarna läggs allt mer resurser på dess övervakning.
Övervakning av markfuktighet görs vanligen antingen via punktbaserad mätning direkt i marken eller via satelliter som mäter mängden vatten via naturliga (passiva system) eller satellitutskickade (aktiva system) mikrovågor. Flera satellitbaserade uppskattningar sammanställs till så kallade markfuktighetsprodukter som publiceras regelbundet. Det finns även produkter som istället för satellitdata utnyttjar atmosfäriska modeller som bygger på uppmätt data, till exempel lufttemperatur och regn. Då olika produkter utnyttjar olika metoder, sensorer och algoritmer genererar de olika markfuktighetsvärden vilket innebär att varje produkts riktighet och precision bör utvärderas.
I det här examensarbetet utvärderades åtta markfuktighetsprodukter i Europa med två metoder. De åtta produkterna inkluderade fem som huvudsakligen använde data från endast en typ av sensor, där sensorerna skiljer sig mellan olika produkter. Dessa var SMAP L3E, SMAP L4 (varav båda använder passiv SMAP-data), SMOS L4 (passiv), Sentinel-1 (aktiv), och ASCAT (aktiv). Dessutom utvärderades produkten ESA CCI SM som använder satellitdata från flera olika sensorer, både aktiva och passiva. Därutöver utvärderades två modellbaserade produkter; ERA5-Land och GLDAS Noah. De åtta produkterna utvärderades först med punktbaserade mätningar från mätstationsnätverket Integrated Carbon Observation System (ICOS). SMAP L4 och ERA5-Land hade bäst resultat medan ESA CCI SM och SMAP L3E också presterade väl. GLDAS Noah hade vissa problem med att fånga kortare väderhändelser. SMOS L4 och ASCAT presterade generellt sämre och allra sämst uppskattningar genererades av Sentinel-1.
Den andra utvärderingsmetoden var ”Triple Collocation Analysis” (TCA) vilket baseras på tre oberoende produkter. Inom TCA är punktbaserad mätning ej nödvändig för produktutvärdering. TCA kräver att de tre produkterna är självständiga vilket begränsar de möjliga kombinationerna. Därför valdes bara en trippel av produkter: SMAP L3E, ASCAT och GLDAS Noah. Resultaten visade att SMAP L3E var mer pålitlig än de andra vilket överrensstämde med de föregående resultaten. Med TCA sammanslogs även de tre produkterna till en ny produkt som kunde utvärderas med ICOS-datan. Den sammanslagna produkten presterade bättre än GLDAS Noah och ASCAT, men inte bättre än SMAP L3E eller ett medelvärde av de tre produkterna. Detta tyder på att TCA-baserad sammanslagning av dessa tre produkter är onödig. Framtida studier rekommenderas att testa andra produktkombinationer, speciellt med ERA5-Land istället för GLDAS Noah. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9129044
- author
- Bergman, Hugo LU
- supervisor
-
- Zheng Duan LU
- organization
- course
- NGEM01 20231
- year
- 2023
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- physical geography and ecosystem analysis, evaluation, soil moisture, remote sensing, ICOS, triple collocation analysis, drought monitoring, geomatics
- publication/series
- Student thesis series INES
- report number
- 607
- language
- English
- id
- 9129044
- date added to LUP
- 2023-06-22 11:39:49
- date last changed
- 2024-06-20 03:52:30
@misc{9129044,
abstract = {{Surface soil moisture (SM) is an essential climate variable that plays a key role in ecosystems and the energy, water, and carbon cycles. SM can be accurately measured using in situ measurements. However, these measurements are globally not densely located over large areas, which would be required for accurate large-scale SM estimation due to the high spatial variability of SM. Instead, global atmospheric models and satellite remote sensing in the microwave range are commonly utilised for large-scale SM monitoring. Both model and satellite approaches have resulted in multiple gridded SM products at regional or global scales at various spatial resolutions (typically between 1 and 40 km). The accuracy of the gridded products varies over different regions, climates, and land covers, necessitating their evaluation. Evaluation with in situ data is limited to areas where measurements are available. Over the past 15 years, triple collocation analysis (TCA) has been extensively applied to evaluate gridded SM products among different geophysical variables, as it can estimate the error structure of three independent datasets without the need for in situ measurements. TCA has also been used to successfully merge gridded products to generate more accurate SM estimates. This study evaluated and ranked eight gridded SM products, including SMOS L4, SMAP L3E, SMAP L4, Sentinel-1, ASCAT, ESA CCI SM, ERA5-Land, and GLDAS Noah, using in situ measurements of SM taken during 2020-2021 from the Integrated Carbon Observation System (ICOS) station network. SMAP L4 and ERA5-Land generally performed the best with similar statistical scores. When comparing the products against absolute SM on collocated dates, SMAP L4 had a median ubRMSD of ca 0.047 m3/m3 and a median correlation coefficient of 0.73. ESA CCI SM and SMAP L3 gave slightly worse scores, while GLDAS Noah showed a relatively poor correlation against short-term SM anomalies. Sentinel-1 generally had the worst performance, together with ASCAT and SMOS L4. In addition, TCA was performed using a triplet consisting of SMAP L3E, ASCAT, and GLDAS Noah. The TCA reinforced the results of the in situ-based evaluation, with the lowest ubRMSE and highest SNR found for SMAP L3E among the three products. The TCA of this triplet was also used to produce a weighted merged dataset of SM estimates. When compared to ICOS measurements, the merged product performed better than GLDAS Noah and ASCAT but similar to or worse than SMAP L3E. Additionally, the TCA-weighted product performed similarly to a simple arithmetic mean, indicating the merging process was not worthwhile in the study area.}},
author = {{Bergman, Hugo}},
language = {{eng}},
note = {{Student Paper}},
series = {{Student thesis series INES}},
title = {{Evaluation and merging of multiple gridded surface soil moisture products in Europe using ICOS measurements and triple collocation analysis}},
year = {{2023}},
}