SOIL MOISTURE MONITORIZATION USING GNSS REFLECTED SIGNALS

Transcription

1 SOIL MOISTURE MONITORIZATION USING GNSS REFLECTED SIGNALS A. Egido, G. Ruffini, M. Capaini, C. Matín, E. Faés, X. Banqué Stalab Bacelona Edifici de l'obsevatoi Faba, Muntanya del Tibidabo Camí de l Obsevatoi Faba s/n Bacelona - Spain alejando.egido@stalab.es 1. INTRODUCTION The use of GNSS signals as a souce of oppotunity fo emote sensing applications has been a eseach aea of geat inteest since 1993, when M. Matin Neia (ESA) poposed that GPS signals eflected fom the Eath s suface could be detected to etieve ocean altimety infomation accounting fo the existent delay between the diect and the eflected signals (the PARIS concept, PAssive Reflectomety and Intefeomety System) [1]. Since then, seveal applications based on a GNSS bistatic ada configuation have been developed taking advantage of the high availability and stability of GNSS signals. This technique is commonly known as GNSS-R (Global Navigation Satellite System Reflections). GNSS-R studies and investigations have been mainly focused on sea suface topogaphy. Within this fame, Stalab Bacelona has developed Oceanpal, a fully opeational system that can povide GNSS-R data and highe level poducts, such as eal time significant wave height (SWH) and altimety data. The application of GNSS-R to land emote sensing has been lagely ovelooked. Nevetheless, thee is expeimental evidence that GPS eflected signals fom the gound can be detected and pocessed in ode to obtain soil moistue estimates, [][3][4]. The impotance of soil moistue elies in the fact that it is a pime paamete fo the suface hydology cycle, which is one of the keys fo the undestanding of the inteaction between continental sufaces and the atmosphee in envionmental studies. Wate stoage in the soil, eithe in the suface laye o in deepe levels, affects not only the evapotanspiation but also the heat stoage ability of the soil, its themal conductivity, and the patitioning of enegy between latent and sensible heat fluxes. In addition, the value of the suface laye volumetic soil moistue diect evapoation fom soil, and detemines the possibility of suface unoff afte ainfalls. Despite the ecognised elevance of soil moistue, poviding such paamete on global scales emains a significant challenge. Sensos based on GNSS-R offe this possibility and could epesent a vey impotant milestone in the development of a global soil moistue model. Stalab expects to develop an opeational GNSS-R senso oiented to soil moistue etieval, which will be based in the Oceanpal instument s achitectue. The pesent pape eviews the most impotant theoetical aspects to take into consideation fo the development of a GNSS-R soil moistue senso, and suggests how the use of the fothcoming Galileo signals might help in this task.. SOIL MOISURE ESTIMATION WITH GNSS SIGNALS The basis fo the etieval of soil moistue with GNSS-R systems lays in the vaiability of the gound dielectic popeties associated to soil moistue. Highe concentations of wate in the soil yield a highe dielectic constant and eflectivity. Consequently, the eflected signal's peak powe can be elated to soil moistue. Pevious investigations [-7] have demonstated the capability of GPS bistatic scatteometes to obtain signal to noise atios high enough to sense small changes in suface eflectivity. Futhemoe, these systems pesent some advantages with espect to those cuently used to etieve soil moistue. Fist, GPS signals lie in L band, which is the most sensitive band fo soil moistue micowave emote sensing. Secondly, in contast to micowave adiomety, vaiations on themal backgound do not damatically contaminate the GPS eflected signals. As will be seen below, themal backgound influences soil moistue obsevables, but this effect is not as impotant as fo micowave adiomety. Thidly, GPS scatteomety fom space has a potential highe spatial esolution than micowave adiomety, due to the highly stable caie and code modulations of the incident signals which enables the use of Delay Dopple mapping. Nevetheless, in ode to obtain pecise soil moistue estimates thee ae seveal phenomena that need to be taken into consideation, mainly the effects of diffuse scatteing ove the soil suface: soil oughness and vegetation canopy.

2 3. THE OCEANPAL INSTRUMENT Oceanpal is a GNSS-R based senso designed fo opeational coastal monitoing. It is an inexpensive, all-weathe, dy and passive instument which can be deployed on multiple platfoms, static (coasts, habous, off-shoe), and slowly moving (boats, floating platfoms, buoys). In its pesent fom, Oceanpal can delive two kinds of Level- poducts: sea-suface height and significant wave height (SWH). Howeve, due to its flexibility in tems of data acquisition, the Oceanpal instument can be applied to the etieval of soil moistue in a staightfowad way Instument s Achitectue Oceanpal compises thee subsystems: a adio fequency (RF) section, an intemediate fequency section and a data pocessing section. The basic system achitectue is illustated in Fig. 1. The RF section featues a pai of low gain L- band antennas. An RHCP (Right-hand cicula polaized) zenith antenna collects the diect GNSS signals while an LHCP (left-hand cicula polaized) nadi antenna collects the sea-suface eflected GNSS signals. Data busts of some minutes ae acquied fom each channel using two adio fequency font-ends that down-convet the signal to intemediate fequency (IF). The acquisition time is a paamete that can be specified by the use. Within the IF section, the signal is one-bit sampled and stoed on a had disk. Afte the acquisition pocess, these diect and eflected aw data ae then fed into the pocessing section of the instument whee a pai of softwae GNSS eceives detects and tacks the available signals in the diect channel (which woks as maste) and blindly dispeads the eflected signals in the eflected/slave channel. The esult of this pocessing is a set of diect and eflected electomagnetic field time seies (complex wavefoms) fo each satellite in view, plus some ancillay infomation. The complex wavefoms ae then used to poduce highe level poducts by the data pocessing algoithms. It must be noted at this point that Oceanpal is cuently a GPS based instument. Howeve, the inteopeability of the GPS and Galileo L1 signals and the fact that Oceanpal is implemented as a softwae eceive (afte the digitization of the signal) the evolution of the system towads a GPS and Galileo instument is elatively easy. Stalab Bacelona expects to have this GNSS-R instument woking by the beginning of 008. Fig. 1. Oceanpal basic setup.

3 3.. The Intefeometic Complex Field The fundamental poduct of the instument data pocessing chain is the so called Intefeometic Complex Field (ICF), which is a time seies calculated as the atio between the eflected and diect wavefom peaks, (1) ICF(t) = p R (t) p D (t) (1) whee p R and p D epesent the time seies of wavefom peaks fo the eflected and diect signals, espectively. The ICF is the basis fo the algoithms used to calculate diffeent highe ode level poducts, and epesents also the fundamental magnitude fo soil moistue estimation. 4. SOIL MOISTURE ESTIMATION USING OCEANPAL In the pevious section, the intefeometic complex field has been defined. It must be noted in (1) that the squaed absolute value of the ICF can be consideed as the peak powe atio of the eflected and the diect signal wavefoms in the lapse of time t. This means that it epesents a measue of the suface eflectivity, which in tun, as mentioned in section, can be elated to the soil moistue volumetic content. Howeve, thee ae seveal paametes such as suface oughness, vegetation canopy, and themal backgound, which affect the detemination of soil moistue. Thei effects ae eviewed moe in depth in the next section. Consideing, as a fist appoximation, that the only paamete affecting the eflected signal is the soil eflectivity, the following incoheent aveaging can be pefomed: N 1 Γ = ICF( t) () av N i = 1 whee N is the numbe of wavefoms computed duing one data acquisition. With () an aveaged value of the soil s eflectivity Γ is be obtained fo the whole acquisition time span, typically one minute. Fom the Fesnel equations of av eflection, the eflection coefficients fo vetical and hoizontal polaization can be obtained: and Γ = v ε sin γ ε + ε sin γ ε sin Γ h = sin γ ε + γ ε (3) (4) wheeγ is the incidence angle. The GPS signals is mostly ight hand cicula polaized (RHCP), which means that it pesents a vetical and a hoizontal polaization component. Howeve, fo high incidence angles, e.g. above 60º, the diffeence between the eflections coefficients fo vetical and hoizontal polaization can be consideed negligible, in a fist appoach, and theefoe, just the eflection coefficient fo vetical polaization can be taken into account. As noted in [4], the eo in using just the vetical value is 5% of the eflectivity at the wost incidence angle. Identifyingε withε and solving the pevious equation fo the pemittivity one can show that soil ε soil 1 Γ ± γ γ 1 1 4sin cos 1+ Γ = 1 Γ sin γ 1+ Γ (5)

4 In ode to elate the soil pemittivity to soil moistue a semi-empiical model pesented in [11] can be used. The authos suggest that the polynomial that descibes the elationship between soil moistue and dielectic constant fo a fequency aound 1.4 GHz is given by ( ) ( ) ε = S C C m S C m (6) soil v v whee S and C ae the sand and clay textual compositions of a soil in pecent by weight, and m is the volumetic soil moistue. This model has been used as a semi-empiical appoach in [5] with acceptable esults. Anothe patially diffeent appoach fo detemine the complex pemittivity of the soil using navigation eflected signals has been poposed in [][1]. The main diffeence in this appoach is the fact that only one antenna is used to collect both the diect and the eflected signal, thus obtaining an intefeometic field as sum of the two EM waves. Othe moe complicated and accuate models than those pesented above have also been used. Fo instance, in [6] a model based on Kichoff Appoximation and Geometic Optics is adopted. This model has been successfully applied in modelling vaious GNSS bistatic ada scenaios (see fo instance [7] which is the most commonly used model in GNSS scatteing fom ocean sufaces). 5. CONSTRAINT IN THE DETERMINATION OF SOIL MOISTURE As mentioned in the pevious section, the estimation of soil moistue with L band signals is affected by seveal ancillay phenomena that distot the intefeometic complex field and bias the measuements of the soil eflectivity. The pesent section eviews the most impotant ones and thei effects in the scatteed signals Suface Roughness and Vegetation Canopy Unde the assumption of a flat scatteing suface, the GPS signal emitted by a cetain space vehicle would be eflected basically fom a specula point ove the suface (moe pecisely, fom the fist Fesnel zone [8]). It is known that the specula point is detemines the shotest path between the emitte and the eceive, though the eflecting suface. Howeve, if a ough suface is consideed and accoding to the Geometical Optics model, slopes may exist with the pope oientation to ediect the incoming adiation to the eceive antenna fom sites away fom the specula point, as depicted in Fig.1. The locus of points ove the suface fom which the eflected signal aives at the same delay, with espect to the specula point, is given by ellipses which ae called iso-delay lines. The signals will also be affected by Dopple shifts due to the changing geomety of the scenaio and the elative motion of emitte and eceive, howeve this effect will not be consideed fo the moment fo soil moistue etieval. Note that this appoximation is pefectly valid in the case of a fix eceive. The powe signal in the eceive at any delay is the esult of summing up the eflected field fom each individual scattee within the coesponding iso-delay ellipse. Assuming that natual scenes ae composed of independently phased scattees, the esulting composite signal is stochastic with Rayleigh distibution (i.e., affected by speckle [8]). Note that the signal stength eceived fom an individual scattee depends on the eflection coefficient of each suface element, as well as the incidence angle. In addition, futhe scatteing and attenuation occus when the signal path includes vegetation canopy. GNSS eceives ely on the spead spectum popeties of the pseudo-andom noise code (PRN), which modulates the caies, in ode to tack the signals. The tacking is pefomed though the coelation of the eceived signals with a clean eplica of itself (matched filteing), obtaining a wavefom whose shape esembles, in an aveage sense, the autocoelation of the oiginal PRN code. Howeve, fo the case of the eflected signal, in addition to the delay intoduced as a consequence of a longe signal path, the wavefom is in geneal distoted and its peak powe diminished due to the scatteing pocess. In Fig., ideal wavefoms of the diect and eflected signals ae sketched. As mentioned befoe, the peak powe of the eflected signal wavefom depends on the suface eflectivity and theefoe can be elated to soil moistue. Howeve, since the eflected signal becomes a stochastic pocess though the scatteing pocess, measuing the peak of the wavefom is not a staightfowad task. The effects of vegetation canopy in the scatteing pocess of GNSS signals ae a vey impotant facto to be taken into consideation [4]. Specifically, fo soil moistue emote sensing, vegetation is often modelled sepaately fom the bae soil suface as a signal attenuation which is popotional to vegetation wate content [9]. v

5 Fig.1. GPS eflections ove a ough suface. Scattees outside the specula point exist that have the pope inclination to ediect the incident signal towads the eceive. Fig.. Diect and eflected signal wavefoms 5.3. System Noise Back to the assumption of a pefectly smooth eflecting suface, the only changes affecting in the eflected signal with espect to the diect one would be the amplitude, deceased by the eflectivity of the suface, and a phase shift. Thus, the peak powe of the eflected signal nomalized by the peak powe of the diect signal povides an obsevable that is popotional to the soil s eflectivity. Howeve, the only diect obsevable of GPS signals is the wavefom that esults fom the coss-coelation of the incoming signal with the locally geneated PRN code, as explained above. It can be demonstated, although it is not within the scope of this eview document, that the wavefom s peak powes of the diect and eflected signals ae popotional to the signal to noise atio of the eceived signal. In the ealistic case of a ough scatteing suface, the eflected signal is affected by additive Gaussian white noise of themal oigin, plus speckle. As a consequence of the speckle noise, the wavefom shape is distoted and the peak powe fluctuates due to fading effects. Thus, estimation of the peak powe equies a cetain amount of incoheent aveaging in ode to educe uncetainty in the measuements. The equation that models the mean complex wavefom powe can be appoximated as whee s is the signal SNR, Λ ( τ ) ( ) ( ) Cˆ τ = s Λ τ + f ( T, τ ) (7) k n indicates the tiangle coelation function of the PRN code, and f is a tem which is function of the coheent integation time and the noise coheence time. The caet denotes independent stochastic vaiables that have some pobability distibution function (a combination of Gaussian themal noise and Rayleigh

6 speckle). The lag vaiable τ identifies diffeent vaiables. It can be infeed fom the pevious elation that the soil moistue obsevable will be affected by the vaiations of themal backgound though vaiations in the signal to noise atio, which will need to be accounted fo in the invesion pocess fo soil moistue estimates. Anothe significant effect of the system noise is the maximum allowed pecision with which the wavefom peak powe can be detemined. The minimum vaiance associated with an estimato is defined by the Came-Rao lowe bound. The Came-Rao lowe bound fo estimating a peak powe α is given by [10] α p α p α = α p Nϒ whee N is the numbe of independent samples aveaged and ϒ is the detected signal enegy to noise atio. In ode to incease the signal enegy to noise atio, longe integation times should be used when pefoming the coelation of the incoming signal with the clean eplica. Howeve, the coheent integation time has a natual pactical uppe limit, fo the case of GPS signals, which coesponds to the duation of a navigation bit, e.g., 0ms fo GPS C/A. In the case of Galileo signals, the existence of pilot signals which ae not modulated by a navigation message eliminates the uppe limit estiction fo the integation time. Theefoe the vaiance of the wavefom peak powe estimation should be significantly educed by using this new signal. The othe way to futhe educe the uncetainty in the final estimation is by aveaging independent samples. Independency of obsevations implies that the phases of the scattees ae uncoelated in each data set to be aveaged. This independency tanslates into a sufficient change of the geomety of the eflection in subsequent data takes. Consideing a gound based instument, since the soil suface is static, the independency condition is accomplished by allowing enough time to pass between consecutive samples, so that the geomety of the obsevations is sufficiently diffeent because of the movement of the tansmitting GPS space vehicles. The fact that a GNSS-R based instument (such as Oceanpal ) is capable to simultaneously tack and pocess signals fom seveal satellites, povides additional independent measues of the scatteing suface, which will also contibute to educe the vaiance of the estimation. It is impotant to stess that in the appoach to soil moistue estimation pesented, vaiations in the ICF which can be caused by the diffeent themal backgound fluctuations in the diect and eflected signals, as well as any mismatching in the diect and eflected eceiving chains have to be accounted fo since they diectly impact the magnitude used fo the estimation. Similaly, since most GNSS-R systems use one zenith and one nadi antenna, a tempeatue gadient in the system is likely to occu, which can esult in an additional diffeence of the noise affecting the diect and the eflected eceiving chains. Hence, both eceiving chains will need to be calibated to pefom effective soil moistue estimation. (8) 6. CONCLUSIONS The most impotant aspects and constaints fo soil moistue etieval with a GNSS-R based instument, as well as a simple fist-appoach scatteing model have been eviewed in this aticle. In addition, a GPS-R instument by Stalab (Oceanpal ), suitable fo soil moistue etieval, has been pesented. The state of the at does not povide the capability to pefom GNSS-R soil moistue emote sensing in an accuate and pecise way. Futhe investigations ae needed to elate in a moe pecise way the effects of soil moistue to the GNSS bistatic scatteing pocess; diffeences between vetical and hoizontal eflection coefficients will need to be accounted fo, as well as diffuse scatteing effects due to suface oughness, advese effects caused by tempeatue vaiations, and vegetation canopy should also be included in fothcoming scatteing models. Extensive validation and calibation campaigns will need to be pefomed, so that soil moistue estimates can be compaed and elated to in-situ soil moistue measuements with pecise knowledge of suface oughness and vegetation canopy. Concening the use of GALILEO, wheeas fo othe emote sensing applications it has been poven to epesent a big step fowad (fo example in ocean mesoscale altimety [13]), with espect to soil moistue estimation, at the moment, the specific contibution of these new signals seems to be limited to the possibility of longe integation time and longe codes, fo noise eduction, and to the use of moe fequencies and satellites to incease the numbe of measuements. Notwithstanding the highlighted difficulties, soil moistue emote sensing with GNSS-R emains an inteesting goal, taking into account the huge advance it would epesent in obtaining global estimations of such an impotant paamete fo the hydologic cycle.

7 7. REFERENCES [1] M.Matin-Neia. A Passive eflectomety and intefeomety system (PARIS): Application to ocean altimety. ESA J., 17: , 1993 [] A. Kavak, G. Xu, W.J. Vogel, GPS Multipath Fade Meassuements to Detemine L-Band Goud Reflectivity Popeties, Univesity of Texas, Austin, [3] D. Mastes, V. Zavootny, S. Katzbeg, W. Emey GPS Signal Scatteing fom Land fo Moistue Content Detemination IGARSS Poceedings, July 4-8, 000. [4] S. J. Katzbeg, O. Toes, M. S. Gant, and D. Mastes, Utilizing calibated GPS eflected signals to estimate soil eflectivity and dielectic constant: Results fom SMEX0, Remote Sens. Envion., vol. 100, no. 1, pp. 17 8, Jan [5] O.Toes, Analysis of Reflected Global Positioning System Signals as a Method fo the Detemination of Soil Moistue, Mastes Thesis. [6] D. Mastes, Suface Remote Sensing Applications of GNSS Bistatic Rada, Soil Moistue and Aicaft Altimety [7] V.U. Zavootny, A.G. Voonovich, Bistatic GPS Signal Reflections at Vaious Polaizations fom Rough Land Suface with Moistue Content. Poc. IEEE Int. Geoscience Remote Sensing Symp., volume 7, 000 [8] P. Beckman and A. Spizzichino. The scatteing of Electomagnetic Waves fom Rough Sufaces. Atech House, Nowood, MA, [9] F. Ulaby, R. Mooe, A. Fung. Micowave Remote Sensing. Atech House, [10] P. Peebles. Rada Pinciples. John Wiley & Sons, Inc, [11] Hallikainen and Ulaby, Micowave Dielectic Behavio of Wet Soil-Pat 1: Empiical Models and Expeimental Obsevations, IEEE Tansactions on Geoscience and Remote Sensing Vol. GE-3, No. 1, pp. 5-34, Jan [1] Kavak, A., Vogel, W.J., Guanghan, X., Using GPS to measue gound complex pemittivity, Electonic Lettes, vol 34, no. 3, p. 54, Feb 5th [13] O. Gemain, G. Ruffini, A evisit to the GNSS-R code ange pecision, Poceedings of the GNSS-R '06 Wokshop, June 006, ESA/ESTEC, Noodwijk, The Nethelands